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 1.334 by root, Mon Oct 25 10:30:23 2010 UTC

…		…
2		2
3	libev - a high performance full-featured event loop written in C	3	libev - a high performance full-featured event loop written in C
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	#include <ev.h>	7	#include <ev.h>
8		8
9	=head1 DESCRIPTION	9	=head2 EXAMPLE PROGRAM
		10
		11	// a single header file is required
		12	#include <ev.h>
		13
		14	#include <stdio.h> // for puts
		15
		16	// every watcher type has its own typedef'd struct
		17	// with the name ev_TYPE
		18	ev_io stdin_watcher;
		19	ev_timer timeout_watcher;
		20
		21	// all watcher callbacks have a similar signature
		22	// this callback is called when data is readable on stdin
		23	static void
		24	stdin_cb (EV_P_ ev_io *w, int revents)
		25	{
		26	puts ("stdin ready");
		27	// for one-shot events, one must manually stop the watcher
		28	// with its corresponding stop function.
		29	ev_io_stop (EV_A_ w);
		30
		31	// this causes all nested ev_run's to stop iterating
		32	ev_break (EV_A_ EVBREAK_ALL);
		33	}
		34
		35	// another callback, this time for a time-out
		36	static void
		37	timeout_cb (EV_P_ ev_timer *w, int revents)
		38	{
		39	puts ("timeout");
		40	// this causes the innermost ev_run to stop iterating
		41	ev_break (EV_A_ EVBREAK_ONE);
		42	}
		43
		44	int
		45	main (void)
		46	{
		47	// use the default event loop unless you have special needs
		48	struct ev_loop *loop = EV_DEFAULT;
		49
		50	// initialise an io watcher, then start it
		51	// this one will watch for stdin to become readable
		52	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		53	ev_io_start (loop, &stdin_watcher);
		54
		55	// initialise a timer watcher, then start it
		56	// simple non-repeating 5.5 second timeout
		57	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		58	ev_timer_start (loop, &timeout_watcher);
		59
		60	// now wait for events to arrive
		61	ev_run (loop, 0);
		62
		63	// unloop was called, so exit
		64	return 0;
		65	}
		66
		67	=head1 ABOUT THIS DOCUMENT
		68
		69	This document documents the libev software package.
		70
		71	The newest version of this document is also available as an html-formatted
		72	web page you might find easier to navigate when reading it for the first
		73	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		74
		75	While this document tries to be as complete as possible in documenting
		76	libev, its usage and the rationale behind its design, it is not a tutorial
		77	on event-based programming, nor will it introduce event-based programming
		78	with libev.
		79
		80	Familiarity with event based programming techniques in general is assumed
		81	throughout this document.
		82
		83	=head1 WHAT TO READ WHEN IN A HURRY
		84
		85	This manual tries to be very detailed, but unfortunately, this also makes
		86	it very long. If you just want to know the basics of libev, I suggest
		87	reading L<ANATOMY OF A WATCHER>, then the L<EXAMPLE PROGRAM> above and
		88	look up the missing functions in L<GLOBAL FUNCTIONS> and the C<ev_io> and
		89	C<ev_timer> sections in L<WATCHER TYPES>.
		90
		91	=head1 ABOUT LIBEV
10		92
11	Libev is an event loop: you register interest in certain events (such as a	93	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	94	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	95	these event sources and provide your program with events.
14		96
15	To do this, it must take more or less complete control over your process	97	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	98	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	99	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	101	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	102	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	103	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	104	watcher.
23		105
24	=head1 FEATURES	106	=head2 FEATURES
25		107
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	108	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	109	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	110	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
		111	(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		112	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		113	timers (C<ev_timer>), absolute timers with customised rescheduling
		114	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	115	change events (C<ev_child>), and event watchers dealing with the event
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	116	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		117	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		118	limited support for fork events (C<ev_fork>).
		119
		120	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	121	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	122	for example).
33		123
34	=head1 CONVENTIONS	124	=head2 CONVENTIONS
35		125
36	Libev is very configurable. In this manual the default configuration	126	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	127	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	128	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	129	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	130	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 *>)	131	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	132	this argument.
43		133
44	=head1 TIME REPRESENTATION	134	=head2 TIME REPRESENTATION
45		135
46	Libev represents time as a single floating point number, representing the	136	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	137	the (fractional) number of seconds since the (POSIX) epoch (in practice
48	the beginning of 1970, details are complicated, don't ask). This type is	138	somewhere near the beginning of 1970, details are complicated, don't
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	139	ask). This type is called C<ev_tstamp>, which is what you should use
50	to the double type in C.	140	too. It usually aliases to the C<double> type in C. When you need to do
		141	any calculations on it, you should treat it as some floating point value.
		142
		143	Unlike the name component C<stamp> might indicate, it is also used for
		144	time differences (e.g. delays) throughout libev.
		145
		146	=head1 ERROR HANDLING
		147
		148	Libev knows three classes of errors: operating system errors, usage errors
		149	and internal errors (bugs).
		150
		151	When libev catches an operating system error it cannot handle (for example
		152	a system call indicating a condition libev cannot fix), it calls the callback
		153	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		154	abort. The default is to print a diagnostic message and to call C<abort
		155	()>.
		156
		157	When libev detects a usage error such as a negative timer interval, then
		158	it will print a diagnostic message and abort (via the C<assert> mechanism,
		159	so C<NDEBUG> will disable this checking): these are programming errors in
		160	the libev caller and need to be fixed there.
		161
		162	Libev also has a few internal error-checking C<assert>ions, and also has
		163	extensive consistency checking code. These do not trigger under normal
		164	circumstances, as they indicate either a bug in libev or worse.
		165
51		166
52	=head1 GLOBAL FUNCTIONS	167	=head1 GLOBAL FUNCTIONS
53		168
54	These functions can be called anytime, even before initialising the	169	These functions can be called anytime, even before initialising the
55	library in any way.	170	library in any way.
…		…
58		173
59	=item ev_tstamp ev_time ()	174	=item ev_tstamp ev_time ()
60		175
61	Returns the current time as libev would use it. Please note that the	176	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	177	C<ev_now> function is usually faster and also often returns the timestamp
63	you actually want to know.	178	you actually want to know. Also interesting is the combination of
		179	C<ev_update_now> and C<ev_now>.
		180
		181	=item ev_sleep (ev_tstamp interval)
		182
		183	Sleep for the given interval: The current thread will be blocked until
		184	either it is interrupted or the given time interval has passed. Basically
		185	this is a sub-second-resolution C<sleep ()>.
64		186
65	=item int ev_version_major ()	187	=item int ev_version_major ()
66		188
67	=item int ev_version_minor ()	189	=item int ev_version_minor ()
68		190
69	You can find out the major and minor version numbers of the library	191	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	192	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	193	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	194	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
73	version of the library your program was compiled against.	195	version of the library your program was compiled against.
74		196
		197	These version numbers refer to the ABI version of the library, not the
		198	release version.
		199
75	Usually, it's a good idea to terminate if the major versions mismatch,	200	Usually, it's a good idea to terminate if the major versions mismatch,
76	as this indicates an incompatible change. Minor versions are usually	201	as this indicates an incompatible change. Minor versions are usually
77	compatible to older versions, so a larger minor version alone is usually	202	compatible to older versions, so a larger minor version alone is usually
78	not a problem.	203	not a problem.
79		204
		205	Example: Make sure we haven't accidentally been linked against the wrong
		206	version (note, however, that this will not detect other ABI mismatches,
		207	such as LFS or reentrancy).
		208
		209	assert (("libev version mismatch",
		210	ev_version_major () == EV_VERSION_MAJOR
		211	&& ev_version_minor () >= EV_VERSION_MINOR));
		212
		213	=item unsigned int ev_supported_backends ()
		214
		215	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		216	value) compiled into this binary of libev (independent of their
		217	availability on the system you are running on). See C<ev_default_loop> for
		218	a description of the set values.
		219
		220	Example: make sure we have the epoll method, because yeah this is cool and
		221	a must have and can we have a torrent of it please!!!11
		222
		223	assert (("sorry, no epoll, no sex",
		224	ev_supported_backends () & EVBACKEND_EPOLL));
		225
		226	=item unsigned int ev_recommended_backends ()
		227
		228	Return the set of all backends compiled into this binary of libev and
		229	also recommended for this platform, meaning it will work for most file
		230	descriptor types. This set is often smaller than the one returned by
		231	C<ev_supported_backends>, as for example kqueue is broken on most BSDs
		232	and will not be auto-detected unless you explicitly request it (assuming
		233	you know what you are doing). This is the set of backends that libev will
		234	probe for if you specify no backends explicitly.
		235
		236	=item unsigned int ev_embeddable_backends ()
		237
		238	Returns the set of backends that are embeddable in other event loops. This
		239	value is platform-specific but can include backends not available on the
		240	current system. To find which embeddable backends might be supported on
		241	the current system, you would need to look at C<ev_embeddable_backends ()
		242	& ev_supported_backends ()>, likewise for recommended ones.
		243
		244	See the description of C<ev_embed> watchers for more info.
		245
80	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	246	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
81		247
82	Sets the allocation function to use (the prototype is similar to the	248	Sets the allocation function to use (the prototype is similar - the
83	realloc C function, the semantics are identical). It is used to allocate	249	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	250	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	251	when memory needs to be allocated (C<size != 0>), the library might abort
86	destructive action. The default is your system realloc function.	252	or take some potentially destructive action.
		253
		254	Since some systems (at least OpenBSD and Darwin) fail to implement
		255	correct C<realloc> semantics, libev will use a wrapper around the system
		256	C<realloc> and C<free> functions by default.
87		257
88	You could override this function in high-availability programs to, say,	258	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,	259	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.	260	or even to sleep a while and retry until some memory is available.
91		261
		262	Example: Replace the libev allocator with one that waits a bit and then
		263	retries (example requires a standards-compliant C<realloc>).
		264
		265	static void *
		266	persistent_realloc (void *ptr, size_t size)
		267	{
		268	for (;;)
		269	{
		270	void *newptr = realloc (ptr, size);
		271
		272	if (newptr)
		273	return newptr;
		274
		275	sleep (60);
		276	}
		277	}
		278
		279	...
		280	ev_set_allocator (persistent_realloc);
		281
92	=item ev_set_syserr_cb (void (*cb)(const char *msg));	282	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
93		283
94	Set the callback function to call on a retryable syscall error (such	284	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	285	as failed select, poll, epoll_wait). The message is a printable string
96	indicating the system call or subsystem causing the problem. If this	286	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	287	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	288	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	289	requested operation, or, if the condition doesn't go away, do bad stuff
100	(such as abort).	290	(such as abort).
101		291
		292	Example: This is basically the same thing that libev does internally, too.
		293
		294	static void
		295	fatal_error (const char *msg)
		296	{
		297	perror (msg);
		298	abort ();
		299	}
		300
		301	...
		302	ev_set_syserr_cb (fatal_error);
		303
102	=back	304	=back
103		305
104	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	306	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
105		307
106	An event loop is described by a C<struct ev_loop *>. The library knows two	308	An event loop is described by a C<struct ev_loop *> (the C<struct> is
107	types of such loops, the I<default> loop, which supports signals and child	309	I<not> optional in this case unless libev 3 compatibility is disabled, as
108	events, and dynamically created loops which do not.	310	libev 3 had an C<ev_loop> function colliding with the struct name).
109		311
110	If you use threads, a common model is to run the default event loop	312	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	313	supports child process events, and dynamically created event loops which
112	create, you also create another event loop. Libev itself does no locking	314	do 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		315
117	=over 4	316	=over 4
118		317
119	=item struct ev_loop *ev_default_loop (unsigned int flags)	318	=item struct ev_loop *ev_default_loop (unsigned int flags)
120		319
121	This will initialise the default event loop if it hasn't been initialised	320	This returns the "default" event loop object, which is what you should
122	yet and return it. If the default loop could not be initialised, returns	321	normally use when you just need "the event loop". Event loop objects and
123	false. If it already was initialised it simply returns it (and ignores the	322	the C<flags> parameter are described in more detail in the entry for
124	flags).	323	C<ev_loop_new>.
		324
		325	If the default loop is already initialised then this function simply
		326	returns it (and ignores the flags. If that is troubling you, check
		327	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		328	flags, which should almost always be C<0>, unless the caller is also the
		329	one calling C<ev_run> or otherwise qualifies as "the main program".
125		330
126	If you don't know what event loop to use, use the one returned from this	331	If you don't know what event loop to use, use the one returned from this
127	function.	332	function (or via the C<EV_DEFAULT> macro).
		333
		334	Note that this function is I<not> thread-safe, so if you want to use it
		335	from multiple threads, you have to employ some kind of mutex (note also
		336	that this case is unlikely, as loops cannot be shared easily between
		337	threads anyway).
		338
		339	The default loop is the only loop that can handle C<ev_child> watchers,
		340	and to do this, it always registers a handler for C<SIGCHLD>. If this is
		341	a problem for your application you can either create a dynamic loop with
		342	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
		343	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
		344
		345	Example: This is the most typical usage.
		346
		347	if (!ev_default_loop (0))
		348	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		349
		350	Example: Restrict libev to the select and poll backends, and do not allow
		351	environment settings to be taken into account:
		352
		353	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		354
		355	=item struct ev_loop *ev_loop_new (unsigned int flags)
		356
		357	This will create and initialise a new event loop object. If the loop
		358	could not be initialised, returns false.
		359
		360	Note that this function I<is> thread-safe, and one common way to use
		361	libev with threads is indeed to create one loop per thread, and using the
		362	default loop in the "main" or "initial" thread.
128		363
129	The flags argument can be used to specify special behaviour or specific	364	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).	365	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
131		366
132	It supports the following flags:	367	The following flags are supported:
133		368
134	=over 4	369	=over 4
135		370
136	=item C<EVFLAG_AUTO>	371	=item C<EVFLAG_AUTO>
137		372
138	The default flags value. Use this if you have no clue (it's the right	373	The default flags value. Use this if you have no clue (it's the right
139	thing, believe me).	374	thing, believe me).
140		375
141	=item C<EVFLAG_NOENV>	376	=item C<EVFLAG_NOENV>
142		377
143	If this flag bit is ored into the flag value (or the program runs setuid	378	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	379	or setgid) then libev will I<not> look at the environment variable
145	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	380	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
146	override the flags completely if it is found in the environment. This is	381	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	382	useful to try out specific backends to test their performance, or to work
148	around bugs.	383	around bugs.
149		384
		385	=item C<EVFLAG_FORKCHECK>
		386
		387	Instead of calling C<ev_loop_fork> manually after a fork, you can also
		388	make libev check for a fork in each iteration by enabling this flag.
		389
		390	This works by calling C<getpid ()> on every iteration of the loop,
		391	and thus this might slow down your event loop if you do a lot of loop
		392	iterations and little real work, but is usually not noticeable (on my
		393	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		394	without a system call and thus I<very> fast, but my GNU/Linux system also has
		395	C<pthread_atfork> which is even faster).
		396
		397	The big advantage of this flag is that you can forget about fork (and
		398	forget about forgetting to tell libev about forking) when you use this
		399	flag.
		400
		401	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		402	environment variable.
		403
		404	=item C<EVFLAG_NOINOTIFY>
		405
		406	When this flag is specified, then libev will not attempt to use the
		407	I<inotify> API for it's C<ev_stat> watchers. Apart from debugging and
		408	testing, this flag can be useful to conserve inotify file descriptors, as
		409	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		410
		411	=item C<EVFLAG_SIGNALFD>
		412
		413	When this flag is specified, then libev will attempt to use the
		414	I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This API
		415	delivers signals synchronously, which makes it both faster and might make
		416	it possible to get the queued signal data. It can also simplify signal
		417	handling with threads, as long as you properly block signals in your
		418	threads that are not interested in handling them.
		419
		420	Signalfd will not be used by default as this changes your signal mask, and
		421	there are a lot of shoddy libraries and programs (glib's threadpool for
		422	example) that can't properly initialise their signal masks.
		423
150	=item C<EVMETHOD_SELECT> (portable select backend)	424	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
151		425
		426	This is your standard select(2) backend. Not I<completely> standard, as
		427	libev tries to roll its own fd_set with no limits on the number of fds,
		428	but if that fails, expect a fairly low limit on the number of fds when
		429	using this backend. It doesn't scale too well (O(highest_fd)), but its
		430	usually the fastest backend for a low number of (low-numbered :) fds.
		431
		432	To get good performance out of this backend you need a high amount of
		433	parallelism (most of the file descriptors should be busy). If you are
		434	writing a server, you should C<accept ()> in a loop to accept as many
		435	connections as possible during one iteration. You might also want to have
		436	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		437	readiness notifications you get per iteration.
		438
		439	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		440	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		441	C<exceptfds> set on that platform).
		442
152	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	443	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
153		444
154	=item C<EVMETHOD_EPOLL> (linux only)	445	And this is your standard poll(2) backend. It's more complicated
		446	than select, but handles sparse fds better and has no artificial
		447	limit on the number of fds you can use (except it will slow down
		448	considerably with a lot of inactive fds). It scales similarly to select,
		449	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		450	performance tips.
155		451
156	=item C<EVMETHOD_KQUEUE> (some bsds only)	452	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		453	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
157		454
158	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	455	=item C<EVBACKEND_EPOLL> (value 4, Linux)
159		456
160	=item C<EVMETHOD_PORT> (solaris 10 only)	457	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		458	kernels).
161		459
162	If one or more of these are ored into the flags value, then only these	460	For few fds, this backend is a bit little slower than poll and select,
163	backends will be tried (in the reverse order as given here). If one are	461	but it scales phenomenally better. While poll and select usually scale
164	specified, any backend will do.	462	like O(total_fds) where n is the total number of fds (or the highest fd),
		463	epoll scales either O(1) or O(active_fds).
		464
		465	The epoll mechanism deserves honorable mention as the most misdesigned
		466	of the more advanced event mechanisms: mere annoyances include silently
		467	dropping file descriptors, requiring a system call per change per file
		468	descriptor (and unnecessary guessing of parameters), problems with dup and
		469	so on. The biggest issue is fork races, however - if a program forks then
		470	I<both> parent and child process have to recreate the epoll set, which can
		471	take considerable time (one syscall per file descriptor) and is of course
		472	hard to detect.
		473
		474	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		475	of course I<doesn't>, and epoll just loves to report events for totally
		476	I<different> file descriptors (even already closed ones, so one cannot
		477	even remove them from the set) than registered in the set (especially
		478	on SMP systems). Libev tries to counter these spurious notifications by
		479	employing an additional generation counter and comparing that against the
		480	events to filter out spurious ones, recreating the set when required. Last
		481	not least, it also refuses to work with some file descriptors which work
		482	perfectly fine with C<select> (files, many character devices...).
		483
		484	While stopping, setting and starting an I/O watcher in the same iteration
		485	will result in some caching, there is still a system call per such
		486	incident (because the same I<file descriptor> could point to a different
		487	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		488	file descriptors might not work very well if you register events for both
		489	file descriptors.
		490
		491	Best performance from this backend is achieved by not unregistering all
		492	watchers for a file descriptor until it has been closed, if possible,
		493	i.e. keep at least one watcher active per fd at all times. Stopping and
		494	starting a watcher (without re-setting it) also usually doesn't cause
		495	extra overhead. A fork can both result in spurious notifications as well
		496	as in libev having to destroy and recreate the epoll object, which can
		497	take considerable time and thus should be avoided.
		498
		499	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		500	faster than epoll for maybe up to a hundred file descriptors, depending on
		501	the usage. So sad.
		502
		503	While nominally embeddable in other event loops, this feature is broken in
		504	all kernel versions tested so far.
		505
		506	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		507	C<EVBACKEND_POLL>.
		508
		509	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		510
		511	Kqueue deserves special mention, as at the time of this writing, it
		512	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		513	with anything but sockets and pipes, except on Darwin, where of course
		514	it's completely useless). Unlike epoll, however, whose brokenness
		515	is by design, these kqueue bugs can (and eventually will) be fixed
		516	without API changes to existing programs. For this reason it's not being
		517	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		518	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		519	system like NetBSD.
		520
		521	You still can embed kqueue into a normal poll or select backend and use it
		522	only for sockets (after having made sure that sockets work with kqueue on
		523	the target platform). See C<ev_embed> watchers for more info.
		524
		525	It scales in the same way as the epoll backend, but the interface to the
		526	kernel is more efficient (which says nothing about its actual speed, of
		527	course). While stopping, setting and starting an I/O watcher does never
		528	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		529	two event changes per incident. Support for C<fork ()> is very bad (but
		530	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		531	cases
		532
		533	This backend usually performs well under most conditions.
		534
		535	While nominally embeddable in other event loops, this doesn't work
		536	everywhere, so you might need to test for this. And since it is broken
		537	almost everywhere, you should only use it when you have a lot of sockets
		538	(for which it usually works), by embedding it into another event loop
		539	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		540	also broken on OS X)) and, did I mention it, using it only for sockets.
		541
		542	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		543	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		544	C<NOTE_EOF>.
		545
		546	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		547
		548	This is not implemented yet (and might never be, unless you send me an
		549	implementation). According to reports, C</dev/poll> only supports sockets
		550	and is not embeddable, which would limit the usefulness of this backend
		551	immensely.
		552
		553	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		554
		555	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		556	it's really slow, but it still scales very well (O(active_fds)).
		557
		558	Please note that Solaris event ports can deliver a lot of spurious
		559	notifications, so you need to use non-blocking I/O or other means to avoid
		560	blocking when no data (or space) is available.
		561
		562	While this backend scales well, it requires one system call per active
		563	file descriptor per loop iteration. For small and medium numbers of file
		564	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		565	might perform better.
		566
		567	On the positive side, with the exception of the spurious readiness
		568	notifications, this backend actually performed fully to specification
		569	in all tests and is fully embeddable, which is a rare feat among the
		570	OS-specific backends (I vastly prefer correctness over speed hacks).
		571
		572	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		573	C<EVBACKEND_POLL>.
		574
		575	=item C<EVBACKEND_ALL>
		576
		577	Try all backends (even potentially broken ones that wouldn't be tried
		578	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		579	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		580
		581	It is definitely not recommended to use this flag.
165		582
166	=back	583	=back
167		584
168	=item struct ev_loop *ev_loop_new (unsigned int flags)	585	If one or more of the backend flags are or'ed into the flags value,
		586	then only these backends will be tried (in the reverse order as listed
		587	here). If none are specified, all backends in C<ev_recommended_backends
		588	()> will be tried.
169		589
170	Similar to C<ev_default_loop>, but always creates a new event loop that is	590	Example: Try to create a event loop that uses epoll and nothing else.
171	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
173	undefined behaviour (or a failed assertion if assertions are enabled).
174		591
175	=item ev_default_destroy ()	592	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		593	if (!epoller)
		594	fatal ("no epoll found here, maybe it hides under your chair");
176		595
177	Destroys the default loop again (frees all memory and kernel state	596	Example: Use whatever libev has to offer, but make sure that kqueue is
178	etc.). This stops all registered event watchers (by not touching them in	597	used if available.
179	any way whatsoever, although you cannot rely on this :).	598
		599	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
180		600
181	=item ev_loop_destroy (loop)	601	=item ev_loop_destroy (loop)
182		602
183	Like C<ev_default_destroy>, but destroys an event loop created by an	603	Destroys an event loop object (frees all memory and kernel state
184	earlier call to C<ev_loop_new>.	604	etc.). None of the active event watchers will be stopped in the normal
		605	sense, so e.g. C<ev_is_active> might still return true. It is your
		606	responsibility to either stop all watchers cleanly yourself I<before>
		607	calling this function, or cope with the fact afterwards (which is usually
		608	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		609	for example).
185		610
186	=item ev_default_fork ()	611	Note that certain global state, such as signal state (and installed signal
		612	handlers), will not be freed by this function, and related watchers (such
		613	as signal and child watchers) would need to be stopped manually.
187		614
		615	This function is normally used on loop objects allocated by
		616	C<ev_loop_new>, but it can also be used on the default loop returned by
		617	C<ev_default_loop>, in which case it is not thread-safe.
		618
		619	Note that it is not advisable to call this function on the default loop
		620	except in the rare occasion where you really need to free it's resources.
		621	If you need dynamically allocated loops it is better to use C<ev_loop_new>
		622	and C<ev_loop_destroy>.
		623
		624	=item ev_loop_fork (loop)
		625
		626	This function sets a flag that causes subsequent C<ev_run> iterations to
188	This function reinitialises the kernel state for backends that have	627	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	628	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	629	the child process. You I<must> call it (or use C<EVFLAG_FORKCHECK>) in the
191	again makes little sense).	630	child before resuming or calling C<ev_run>.
192		631
193	You I<must> call this function after forking if and only if you want to	632	Again, you I<have> to call it on I<any> loop that you want to re-use after
194	use the event library in both processes. If you just fork+exec, you don't	633	a fork, I<even if you do not plan to use the loop in the parent>. This is
195	have to call it.	634	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		635	during fork.
		636
		637	On the other hand, you only need to call this function in the child
		638	process if and only if you want to use the event loop in the child. If
		639	you just fork+exec or create a new loop in the child, you don't have to
		640	call it at all (in fact, C<epoll> is so badly broken that it makes a
		641	difference, but libev will usually detect this case on its own and do a
		642	costly reset of the backend).
196		643
197	The function itself is quite fast and it's usually not a problem to call	644	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	645	it just in case after a fork.
199	quite nicely into a call to C<pthread_atfork>:
200		646
		647	Example: Automate calling C<ev_loop_fork> on the default loop when
		648	using pthreads.
		649
		650	static void
		651	post_fork_child (void)
		652	{
		653	ev_loop_fork (EV_DEFAULT);
		654	}
		655
		656	...
201	pthread_atfork (0, 0, ev_default_fork);	657	pthread_atfork (0, 0, post_fork_child);
202		658
203	=item ev_loop_fork (loop)	659	=item int ev_is_default_loop (loop)
204		660
205	Like C<ev_default_fork>, but acts on an event loop created by	661	Returns true when the given loop is, in fact, the default loop, and false
206	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	662	otherwise.
207	after fork, and how you do this is entirely your own problem.
208		663
		664	=item unsigned int ev_iteration (loop)
		665
		666	Returns the current iteration count for the event loop, which is identical
		667	to the number of times libev did poll for new events. It starts at C<0>
		668	and happily wraps around with enough iterations.
		669
		670	This value can sometimes be useful as a generation counter of sorts (it
		671	"ticks" the number of loop iterations), as it roughly corresponds with
		672	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		673	prepare and check phases.
		674
209	=item unsigned int ev_method (loop)	675	=item unsigned int ev_depth (loop)
210		676
		677	Returns the number of times C<ev_run> was entered minus the number of
		678	times C<ev_run> was exited, in other words, the recursion depth.
		679
		680	Outside C<ev_run>, this number is zero. In a callback, this number is
		681	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		682	in which case it is higher.
		683
		684	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread
		685	etc.), doesn't count as "exit" - consider this as a hint to avoid such
		686	ungentleman-like behaviour unless it's really convenient.
		687
		688	=item unsigned int ev_backend (loop)
		689
211	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	690	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
212	use.	691	use.
213		692
214	=item ev_tstamp ev_now (loop)	693	=item ev_tstamp ev_now (loop)
215		694
216	Returns the current "event loop time", which is the time the event loop	695	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	696	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	697	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	698	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).	699	event occurring (or more correctly, libev finding out about it).
221		700
		701	=item ev_now_update (loop)
		702
		703	Establishes the current time by querying the kernel, updating the time
		704	returned by C<ev_now ()> in the progress. This is a costly operation and
		705	is usually done automatically within C<ev_run ()>.
		706
		707	This function is rarely useful, but when some event callback runs for a
		708	very long time without entering the event loop, updating libev's idea of
		709	the current time is a good idea.
		710
		711	See also L<The special problem of time updates> in the C<ev_timer> section.
		712
		713	=item ev_suspend (loop)
		714
		715	=item ev_resume (loop)
		716
		717	These two functions suspend and resume an event loop, for use when the
		718	loop is not used for a while and timeouts should not be processed.
		719
		720	A typical use case would be an interactive program such as a game: When
		721	the user presses C<^Z> to suspend the game and resumes it an hour later it
		722	would be best to handle timeouts as if no time had actually passed while
		723	the program was suspended. This can be achieved by calling C<ev_suspend>
		724	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		725	C<ev_resume> directly afterwards to resume timer processing.
		726
		727	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		728	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		729	will be rescheduled (that is, they will lose any events that would have
		730	occurred while suspended).
		731
		732	After calling C<ev_suspend> you B<must not> call I<any> function on the
		733	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		734	without a previous call to C<ev_suspend>.
		735
		736	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		737	event loop time (see C<ev_now_update>).
		738
222	=item ev_loop (loop, int flags)	739	=item ev_run (loop, int flags)
223		740
224	Finally, this is it, the event handler. This function usually is called	741	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	742	after you have initialised all your watchers and you want to start
226	events.	743	handling events. It will ask the operating system for any new events, call
		744	the watcher callbacks, an then repeat the whole process indefinitely: This
		745	is why event loops are called I<loops>.
227		746
228	If the flags argument is specified as 0, it will not return until either	747	If the flags argument is specified as C<0>, it will keep handling events
229	no event watchers are active anymore or C<ev_unloop> was called.	748	until either no event watchers are active anymore or C<ev_break> was
		749	called.
230		750
		751	Please note that an explicit C<ev_break> is usually better than
		752	relying on all watchers to be stopped when deciding when a program has
		753	finished (especially in interactive programs), but having a program
		754	that automatically loops as long as it has to and no longer by virtue
		755	of relying on its watchers stopping correctly, that is truly a thing of
		756	beauty.
		757
231	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	758	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
232	those events and any outstanding ones, but will not block your process in	759	those events and any already outstanding ones, but will not wait and
233	case there are no events and will return after one iteration of the loop.	760	block your process in case there are no events and will return after one
		761	iteration of the loop. This is sometimes useful to poll and handle new
		762	events while doing lengthy calculations, to keep the program responsive.
234		763
235	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	764	A flags value of C<EVRUN_ONCE> will look for new events (waiting if
236	neccessary) and will handle those and any outstanding ones. It will block	765	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	766	will block your process until at least one new event arrives (which could
		767	be an event internal to libev itself, so there is no guarantee that a
		768	user-registered callback will be called), and will return after one
238	one iteration of the loop.	769	iteration of the loop.
239		770
240	This flags value could be used to implement alternative looping	771	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	772	with something not expressible using other libev watchers (i.e. "roll your
242	more generic mechanism.	773	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		774	usually a better approach for this kind of thing.
243		775
244	Here are the gory details of what ev_loop does:	776	Here are the gory details of what C<ev_run> does:
245		777
246	1. If there are no active watchers (reference count is zero), return.	778	- Increment loop depth.
		779	- Reset the ev_break status.
		780	- Before the first iteration, call any pending watchers.
		781	LOOP:
		782	- If EVFLAG_FORKCHECK was used, check for a fork.
		783	- If a fork was detected (by any means), queue and call all fork watchers.
247	2. Queue and immediately call all prepare watchers.	784	- Queue and call all prepare watchers.
		785	- If ev_break was called, goto FINISH.
248	3. If we have been forked, recreate the kernel state.	786	- If we have been forked, detach and recreate the kernel state
		787	as to not disturb the other process.
249	4. Update the kernel state with all outstanding changes.	788	- Update the kernel state with all outstanding changes.
250	5. Update the "event loop time".	789	- Update the "event loop time" (ev_now ()).
251	6. Calculate for how long to block.	790	- Calculate for how long to sleep or block, if at all
		791	(active idle watchers, EVRUN_NOWAIT or not having
		792	any active watchers at all will result in not sleeping).
		793	- Sleep if the I/O and timer collect interval say so.
		794	- Increment loop iteration counter.
252	7. Block the process, waiting for events.	795	- Block the process, waiting for any events.
		796	- Queue all outstanding I/O (fd) events.
253	8. Update the "event loop time" and do time jump handling.	797	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
254	9. Queue all outstanding timers.	798	- Queue all expired timers.
255	10. Queue all outstanding periodics.	799	- Queue all expired periodics.
256	11. If no events are pending now, queue all idle watchers.	800	- Queue all idle watchers with priority higher than that of pending events.
257	12. Queue all check watchers.	801	- Queue all check watchers.
258	13. Call all queued watchers in reverse order (i.e. check watchers first).	802	- Call all queued watchers in reverse order (i.e. check watchers first).
259	14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	803	Signals and child watchers are implemented as I/O watchers, and will
260	was used, return, otherwise continue with step #1.	804	be handled here by queueing them when their watcher gets executed.
		805	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
		806	were used, or there are no active watchers, goto FINISH, otherwise
		807	continue with step LOOP.
		808	FINISH:
		809	- Reset the ev_break status iff it was EVBREAK_ONE.
		810	- Decrement the loop depth.
		811	- Return.
261		812
		813	Example: Queue some jobs and then loop until no events are outstanding
		814	anymore.
		815
		816	... queue jobs here, make sure they register event watchers as long
		817	... as they still have work to do (even an idle watcher will do..)
		818	ev_run (my_loop, 0);
		819	... jobs done or somebody called unloop. yeah!
		820
262	=item ev_unloop (loop, how)	821	=item ev_break (loop, how)
263		822
264	Can be used to make a call to C<ev_loop> return early (but only after it	823	Can be used to make a call to C<ev_run> return early (but only after it
265	has processed all outstanding events). The C<how> argument must be either	824	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	825	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
267	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	826	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		827
		828	This "unloop state" will be cleared when entering C<ev_run> again.
		829
		830	It is safe to call C<ev_break> from outside any C<ev_run> calls. ##TODO##
268		831
269	=item ev_ref (loop)	832	=item ev_ref (loop)
270		833
271	=item ev_unref (loop)	834	=item ev_unref (loop)
272		835
273	Ref/unref can be used to add or remove a reference count on the event	836	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	837	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	838	count is nonzero, C<ev_run> will not return on its own.
276	a watcher you never unregister that should not keep C<ev_loop> from	839
277	returning, ev_unref() after starting, and ev_ref() before stopping it. For	840	This is useful when you have a watcher that you never intend to
		841	unregister, but that nevertheless should not keep C<ev_run> from
		842	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		843	before stopping it.
		844
278	example, libev itself uses this for its internal signal pipe: It is not	845	As an example, libev itself uses this for its internal signal pipe: It
279	visible to the libev user and should not keep C<ev_loop> from exiting if	846	is not visible to the libev user and should not keep C<ev_run> from
280	no event watchers registered by it are active. It is also an excellent	847	exiting if no event watchers registered by it are active. It is also an
281	way to do this for generic recurring timers or from within third-party	848	excellent way to do this for generic recurring timers or from within
282	libraries. Just remember to I<unref after start> and I<ref before stop>.	849	third-party libraries. Just remember to I<unref after start> and I<ref
		850	before stop> (but only if the watcher wasn't active before, or was active
		851	before, respectively. Note also that libev might stop watchers itself
		852	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		853	in the callback).
		854
		855	Example: Create a signal watcher, but keep it from keeping C<ev_run>
		856	running when nothing else is active.
		857
		858	ev_signal exitsig;
		859	ev_signal_init (&exitsig, sig_cb, SIGINT);
		860	ev_signal_start (loop, &exitsig);
		861	evf_unref (loop);
		862
		863	Example: For some weird reason, unregister the above signal handler again.
		864
		865	ev_ref (loop);
		866	ev_signal_stop (loop, &exitsig);
		867
		868	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		869
		870	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		871
		872	These advanced functions influence the time that libev will spend waiting
		873	for events. Both time intervals are by default C<0>, meaning that libev
		874	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		875	latency.
		876
		877	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		878	allows libev to delay invocation of I/O and timer/periodic callbacks
		879	to increase efficiency of loop iterations (or to increase power-saving
		880	opportunities).
		881
		882	The idea is that sometimes your program runs just fast enough to handle
		883	one (or very few) event(s) per loop iteration. While this makes the
		884	program responsive, it also wastes a lot of CPU time to poll for new
		885	events, especially with backends like C<select ()> which have a high
		886	overhead for the actual polling but can deliver many events at once.
		887
		888	By setting a higher I<io collect interval> you allow libev to spend more
		889	time collecting I/O events, so you can handle more events per iteration,
		890	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		891	C<ev_timer>) will be not affected. Setting this to a non-null value will
		892	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		893	sleep time ensures that libev will not poll for I/O events more often then
		894	once per this interval, on average.
		895
		896	Likewise, by setting a higher I<timeout collect interval> you allow libev
		897	to spend more time collecting timeouts, at the expense of increased
		898	latency/jitter/inexactness (the watcher callback will be called
		899	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		900	value will not introduce any overhead in libev.
		901
		902	Many (busy) programs can usually benefit by setting the I/O collect
		903	interval to a value near C<0.1> or so, which is often enough for
		904	interactive servers (of course not for games), likewise for timeouts. It
		905	usually doesn't make much sense to set it to a lower value than C<0.01>,
		906	as this approaches the timing granularity of most systems. Note that if
		907	you do transactions with the outside world and you can't increase the
		908	parallelity, then this setting will limit your transaction rate (if you
		909	need to poll once per transaction and the I/O collect interval is 0.01,
		910	then you can't do more than 100 transactions per second).
		911
		912	Setting the I<timeout collect interval> can improve the opportunity for
		913	saving power, as the program will "bundle" timer callback invocations that
		914	are "near" in time together, by delaying some, thus reducing the number of
		915	times the process sleeps and wakes up again. Another useful technique to
		916	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		917	they fire on, say, one-second boundaries only.
		918
		919	Example: we only need 0.1s timeout granularity, and we wish not to poll
		920	more often than 100 times per second:
		921
		922	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		923	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		924
		925	=item ev_invoke_pending (loop)
		926
		927	This call will simply invoke all pending watchers while resetting their
		928	pending state. Normally, C<ev_run> does this automatically when required,
		929	but when overriding the invoke callback this call comes handy. This
		930	function can be invoked from a watcher - this can be useful for example
		931	when you want to do some lengthy calculation and want to pass further
		932	event handling to another thread (you still have to make sure only one
		933	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		934
		935	=item int ev_pending_count (loop)
		936
		937	Returns the number of pending watchers - zero indicates that no watchers
		938	are pending.
		939
		940	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		941
		942	This overrides the invoke pending functionality of the loop: Instead of
		943	invoking all pending watchers when there are any, C<ev_run> will call
		944	this callback instead. This is useful, for example, when you want to
		945	invoke the actual watchers inside another context (another thread etc.).
		946
		947	If you want to reset the callback, use C<ev_invoke_pending> as new
		948	callback.
		949
		950	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		951
		952	Sometimes you want to share the same loop between multiple threads. This
		953	can be done relatively simply by putting mutex_lock/unlock calls around
		954	each call to a libev function.
		955
		956	However, C<ev_run> can run an indefinite time, so it is not feasible
		957	to wait for it to return. One way around this is to wake up the event
		958	loop via C<ev_break> and C<av_async_send>, another way is to set these
		959	I<release> and I<acquire> callbacks on the loop.
		960
		961	When set, then C<release> will be called just before the thread is
		962	suspended waiting for new events, and C<acquire> is called just
		963	afterwards.
		964
		965	Ideally, C<release> will just call your mutex_unlock function, and
		966	C<acquire> will just call the mutex_lock function again.
		967
		968	While event loop modifications are allowed between invocations of
		969	C<release> and C<acquire> (that's their only purpose after all), no
		970	modifications done will affect the event loop, i.e. adding watchers will
		971	have no effect on the set of file descriptors being watched, or the time
		972	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		973	to take note of any changes you made.
		974
		975	In theory, threads executing C<ev_run> will be async-cancel safe between
		976	invocations of C<release> and C<acquire>.
		977
		978	See also the locking example in the C<THREADS> section later in this
		979	document.
		980
		981	=item ev_set_userdata (loop, void *data)
		982
		983	=item ev_userdata (loop)
		984
		985	Set and retrieve a single C<void *> associated with a loop. When
		986	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		987	C<0.>
		988
		989	These two functions can be used to associate arbitrary data with a loop,
		990	and are intended solely for the C<invoke_pending_cb>, C<release> and
		991	C<acquire> callbacks described above, but of course can be (ab-)used for
		992	any other purpose as well.
		993
		994	=item ev_verify (loop)
		995
		996	This function only does something when C<EV_VERIFY> support has been
		997	compiled in, which is the default for non-minimal builds. It tries to go
		998	through all internal structures and checks them for validity. If anything
		999	is found to be inconsistent, it will print an error message to standard
		1000	error and call C<abort ()>.
		1001
		1002	This can be used to catch bugs inside libev itself: under normal
		1003	circumstances, this function will never abort as of course libev keeps its
		1004	data structures consistent.
283		1005
284	=back	1006	=back
285		1007
		1008
286	=head1 ANATOMY OF A WATCHER	1009	=head1 ANATOMY OF A WATCHER
287		1010
		1011	In the following description, uppercase C<TYPE> in names stands for the
		1012	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1013	watchers and C<ev_io_start> for I/O watchers.
		1014
288	A watcher is a structure that you create and register to record your	1015	A watcher is an opaque structure that you allocate and register to record
289	interest in some event. For instance, if you want to wait for STDIN to	1016	your interest in some event. To make a concrete example, imagine you want
290	become readable, you would create an C<ev_io> watcher for that:	1017	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1018	for that:
291		1019
292	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1020	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
293	{	1021	{
294	ev_io_stop (w);	1022	ev_io_stop (w);
295	ev_unloop (loop, EVUNLOOP_ALL);	1023	ev_break (loop, EVBREAK_ALL);
296	}	1024	}
297		1025
298	struct ev_loop *loop = ev_default_loop (0);	1026	struct ev_loop *loop = ev_default_loop (0);
		1027
299	struct ev_io stdin_watcher;	1028	ev_io stdin_watcher;
		1029
300	ev_init (&stdin_watcher, my_cb);	1030	ev_init (&stdin_watcher, my_cb);
301	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1031	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
302	ev_io_start (loop, &stdin_watcher);	1032	ev_io_start (loop, &stdin_watcher);
		1033
303	ev_loop (loop, 0);	1034	ev_run (loop, 0);
304		1035
305	As you can see, you are responsible for allocating the memory for your	1036	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,	1037	watcher structures (and it is I<usually> a bad idea to do this on the
307	although this can sometimes be quite valid).	1038	stack).
308		1039
		1040	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1041	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1042
309	Each watcher structure must be initialised by a call to C<ev_init	1043	Each watcher structure must be initialised by a call to C<ev_init (watcher
310	(watcher *, callback)>, which expects a callback to be provided. This	1044	*, callback)>, which expects a callback to be provided. This callback is
311	callback gets invoked each time the event occurs (or, in the case of io	1045	invoked each time the event occurs (or, in the case of I/O watchers, each
312	watchers, each time the event loop detects that the file descriptor given	1046	time the event loop detects that the file descriptor given is readable
313	is readable and/or writable).	1047	and/or writable).
314		1048
315	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1049	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	1050	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	1051	is also a macro to combine initialisation and setting in one call: C<<
318	(watcher *, callback, ...) >>.	1052	ev_TYPE_init (watcher *, callback, ...) >>.
319		1053
320	To make the watcher actually watch out for events, you have to start it	1054	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	1055	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	1056	*) >>), and you can stop watching for events at any time by calling the
323	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1057	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
324		1058
325	As long as your watcher is active (has been started but not stopped) you	1059	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	1060	must not touch the values stored in it. Most specifically you must never
327	reinitialise it or call its set method.	1061	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		1062
334	Each and every callback receives the event loop pointer as first, the	1063	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	1064	registered watcher structure as second, and a bitset of received events as
336	third argument.	1065	third argument.
337		1066
…		…
346	=item C<EV_WRITE>	1075	=item C<EV_WRITE>
347		1076
348	The file descriptor in the C<ev_io> watcher has become readable and/or	1077	The file descriptor in the C<ev_io> watcher has become readable and/or
349	writable.	1078	writable.
350		1079
351	=item C<EV_TIMEOUT>	1080	=item C<EV_TIMER>
352		1081
353	The C<ev_timer> watcher has timed out.	1082	The C<ev_timer> watcher has timed out.
354		1083
355	=item C<EV_PERIODIC>	1084	=item C<EV_PERIODIC>
356		1085
…		…
362		1091
363	=item C<EV_CHILD>	1092	=item C<EV_CHILD>
364		1093
365	The pid specified in the C<ev_child> watcher has received a status change.	1094	The pid specified in the C<ev_child> watcher has received a status change.
366		1095
		1096	=item C<EV_STAT>
		1097
		1098	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1099
367	=item C<EV_IDLE>	1100	=item C<EV_IDLE>
368		1101
369	The C<ev_idle> watcher has determined that you have nothing better to do.	1102	The C<ev_idle> watcher has determined that you have nothing better to do.
370		1103
371	=item C<EV_PREPARE>	1104	=item C<EV_PREPARE>
372		1105
373	=item C<EV_CHECK>	1106	=item C<EV_CHECK>
374		1107
375	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts	1108	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts
376	to gather new events, and all C<ev_check> watchers are invoked just after	1109	to gather new events, and all C<ev_check> watchers are invoked just after
377	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1110	C<ev_run> has gathered them, but before it invokes any callbacks for any
378	received events. Callbacks of both watcher types can start and stop as	1111	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	1112	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	1113	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
381	C<ev_loop> from blocking).	1114	C<ev_run> from blocking).
		1115
		1116	=item C<EV_EMBED>
		1117
		1118	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1119
		1120	=item C<EV_FORK>
		1121
		1122	The event loop has been resumed in the child process after fork (see
		1123	C<ev_fork>).
		1124
		1125	=item C<EV_CLEANUP>
		1126
		1127	The event loop is about to be destroyed (see C<ev_cleanup>).
		1128
		1129	=item C<EV_ASYNC>
		1130
		1131	The given async watcher has been asynchronously notified (see C<ev_async>).
		1132
		1133	=item C<EV_CUSTOM>
		1134
		1135	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1136	by libev users to signal watchers (e.g. via C<ev_feed_event>).
382		1137
383	=item C<EV_ERROR>	1138	=item C<EV_ERROR>
384		1139
385	An unspecified error has occured, the watcher has been stopped. This might	1140	An unspecified error has occurred, the watcher has been stopped. This might
386	happen because the watcher could not be properly started because libev	1141	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	1142	ran out of memory, a file descriptor was found to be closed or any other
		1143	problem. Libev considers these application bugs.
		1144
388	problem. You best act on it by reporting the problem and somehow coping	1145	You best act on it by reporting the problem and somehow coping with the
389	with the watcher being stopped.	1146	watcher being stopped. Note that well-written programs should not receive
		1147	an error ever, so when your watcher receives it, this usually indicates a
		1148	bug in your program.
390		1149
391	Libev will usually signal a few "dummy" events together with an error,	1150	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	1151	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	1152	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	1153	the error from read() or write(). This will not work in multi-threaded
395	programs, though, so beware.	1154	programs, though, as the fd could already be closed and reused for another
		1155	thing, so beware.
396		1156
397	=back	1157	=back
398		1158
		1159	=head2 WATCHER STATES
		1160
		1161	There are various watcher states mentioned throughout this manual -
		1162	active, pending and so on. In this section these states and the rules to
		1163	transition between them will be described in more detail - and while these
		1164	rules might look complicated, they usually do "the right thing".
		1165
		1166	=over 4
		1167
		1168	=item initialiased
		1169
		1170	Before a watcher can be registered with the event looop it has to be
		1171	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1172	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1173
		1174	In this state it is simply some block of memory that is suitable for use
		1175	in an event loop. It can be moved around, freed, reused etc. at will.
		1176
		1177	=item started/running/active
		1178
		1179	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1180	property of the event loop, and is actively waiting for events. While in
		1181	this state it cannot be accessed (except in a few documented ways), moved,
		1182	freed or anything else - the only legal thing is to keep a pointer to it,
		1183	and call libev functions on it that are documented to work on active watchers.
		1184
		1185	=item pending
		1186
		1187	If a watcher is active and libev determines that an event it is interested
		1188	in has occurred (such as a timer expiring), it will become pending. It will
		1189	stay in this pending state until either it is stopped or its callback is
		1190	about to be invoked, so it is not normally pending inside the watcher
		1191	callback.
		1192
		1193	The watcher might or might not be active while it is pending (for example,
		1194	an expired non-repeating timer can be pending but no longer active). If it
		1195	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1196	but it is still property of the event loop at this time, so cannot be
		1197	moved, freed or reused. And if it is active the rules described in the
		1198	previous item still apply.
		1199
		1200	It is also possible to feed an event on a watcher that is not active (e.g.
		1201	via C<ev_feed_event>), in which case it becomes pending without being
		1202	active.
		1203
		1204	=item stopped
		1205
		1206	A watcher can be stopped implicitly by libev (in which case it might still
		1207	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1208	latter will clear any pending state the watcher might be in, regardless
		1209	of whether it was active or not, so stopping a watcher explicitly before
		1210	freeing it is often a good idea.
		1211
		1212	While stopped (and not pending) the watcher is essentially in the
		1213	initialised state, that is it can be reused, moved, modified in any way
		1214	you wish.
		1215
		1216	=back
		1217
		1218	=head2 GENERIC WATCHER FUNCTIONS
		1219
		1220	=over 4
		1221
		1222	=item C<ev_init> (ev_TYPE *watcher, callback)
		1223
		1224	This macro initialises the generic portion of a watcher. The contents
		1225	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1226	the generic parts of the watcher are initialised, you I<need> to call
		1227	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1228	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1229	which rolls both calls into one.
		1230
		1231	You can reinitialise a watcher at any time as long as it has been stopped
		1232	(or never started) and there are no pending events outstanding.
		1233
		1234	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1235	int revents)>.
		1236
		1237	Example: Initialise an C<ev_io> watcher in two steps.
		1238
		1239	ev_io w;
		1240	ev_init (&w, my_cb);
		1241	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1242
		1243	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1244
		1245	This macro initialises the type-specific parts of a watcher. You need to
		1246	call C<ev_init> at least once before you call this macro, but you can
		1247	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1248	macro on a watcher that is active (it can be pending, however, which is a
		1249	difference to the C<ev_init> macro).
		1250
		1251	Although some watcher types do not have type-specific arguments
		1252	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1253
		1254	See C<ev_init>, above, for an example.
		1255
		1256	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1257
		1258	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1259	calls into a single call. This is the most convenient method to initialise
		1260	a watcher. The same limitations apply, of course.
		1261
		1262	Example: Initialise and set an C<ev_io> watcher in one step.
		1263
		1264	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1265
		1266	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1267
		1268	Starts (activates) the given watcher. Only active watchers will receive
		1269	events. If the watcher is already active nothing will happen.
		1270
		1271	Example: Start the C<ev_io> watcher that is being abused as example in this
		1272	whole section.
		1273
		1274	ev_io_start (EV_DEFAULT_UC, &w);
		1275
		1276	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1277
		1278	Stops the given watcher if active, and clears the pending status (whether
		1279	the watcher was active or not).
		1280
		1281	It is possible that stopped watchers are pending - for example,
		1282	non-repeating timers are being stopped when they become pending - but
		1283	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1284	pending. If you want to free or reuse the memory used by the watcher it is
		1285	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1286
		1287	=item bool ev_is_active (ev_TYPE *watcher)
		1288
		1289	Returns a true value iff the watcher is active (i.e. it has been started
		1290	and not yet been stopped). As long as a watcher is active you must not modify
		1291	it.
		1292
		1293	=item bool ev_is_pending (ev_TYPE *watcher)
		1294
		1295	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1296	events but its callback has not yet been invoked). As long as a watcher
		1297	is pending (but not active) you must not call an init function on it (but
		1298	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1299	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1300	it).
		1301
		1302	=item callback ev_cb (ev_TYPE *watcher)
		1303
		1304	Returns the callback currently set on the watcher.
		1305
		1306	=item ev_cb_set (ev_TYPE *watcher, callback)
		1307
		1308	Change the callback. You can change the callback at virtually any time
		1309	(modulo threads).
		1310
		1311	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1312
		1313	=item int ev_priority (ev_TYPE *watcher)
		1314
		1315	Set and query the priority of the watcher. The priority is a small
		1316	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1317	(default: C<-2>). Pending watchers with higher priority will be invoked
		1318	before watchers with lower priority, but priority will not keep watchers
		1319	from being executed (except for C<ev_idle> watchers).
		1320
		1321	If you need to suppress invocation when higher priority events are pending
		1322	you need to look at C<ev_idle> watchers, which provide this functionality.
		1323
		1324	You I<must not> change the priority of a watcher as long as it is active or
		1325	pending.
		1326
		1327	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1328	fine, as long as you do not mind that the priority value you query might
		1329	or might not have been clamped to the valid range.
		1330
		1331	The default priority used by watchers when no priority has been set is
		1332	always C<0>, which is supposed to not be too high and not be too low :).
		1333
		1334	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1335	priorities.
		1336
		1337	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1338
		1339	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1340	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1341	can deal with that fact, as both are simply passed through to the
		1342	callback.
		1343
		1344	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1345
		1346	If the watcher is pending, this function clears its pending status and
		1347	returns its C<revents> bitset (as if its callback was invoked). If the
		1348	watcher isn't pending it does nothing and returns C<0>.
		1349
		1350	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1351	callback to be invoked, which can be accomplished with this function.
		1352
		1353	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1354
		1355	Feeds the given event set into the event loop, as if the specified event
		1356	had happened for the specified watcher (which must be a pointer to an
		1357	initialised but not necessarily started event watcher). Obviously you must
		1358	not free the watcher as long as it has pending events.
		1359
		1360	Stopping the watcher, letting libev invoke it, or calling
		1361	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1362	not started in the first place.
		1363
		1364	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1365	functions that do not need a watcher.
		1366
		1367	=back
		1368
		1369
399	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1370	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
400		1371
401	Each watcher has, by default, a member C<void *data> that you can change	1372	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	1373	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	1374	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	1375	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	1376	member, you can also "subclass" the watcher type and provide your own
406	data:	1377	data:
407		1378
408	struct my_io	1379	struct my_io
409	{	1380	{
410	struct ev_io io;	1381	ev_io io;
411	int otherfd;	1382	int otherfd;
412	void *somedata;	1383	void *somedata;
413	struct whatever *mostinteresting;	1384	struct whatever *mostinteresting;
414	}	1385	};
		1386
		1387	...
		1388	struct my_io w;
		1389	ev_io_init (&w.io, my_cb, fd, EV_READ);
415		1390
416	And since your callback will be called with a pointer to the watcher, you	1391	And since your callback will be called with a pointer to the watcher, you
417	can cast it back to your own type:	1392	can cast it back to your own type:
418		1393
419	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1394	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
420	{	1395	{
421	struct my_io *w = (struct my_io *)w_;	1396	struct my_io *w = (struct my_io *)w_;
422	...	1397	...
423	}	1398	}
424		1399
425	More interesting and less C-conformant ways of catsing your callback type	1400	More interesting and less C-conformant ways of casting your callback type
426	have been omitted....	1401	instead have been omitted.
		1402
		1403	Another common scenario is to use some data structure with multiple
		1404	embedded watchers:
		1405
		1406	struct my_biggy
		1407	{
		1408	int some_data;
		1409	ev_timer t1;
		1410	ev_timer t2;
		1411	}
		1412
		1413	In this case getting the pointer to C<my_biggy> is a bit more
		1414	complicated: Either you store the address of your C<my_biggy> struct
		1415	in the C<data> member of the watcher (for woozies), or you need to use
		1416	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1417	programmers):
		1418
		1419	#include <stddef.h>
		1420
		1421	static void
		1422	t1_cb (EV_P_ ev_timer *w, int revents)
		1423	{
		1424	struct my_biggy big = (struct my_biggy *)
		1425	(((char *)w) - offsetof (struct my_biggy, t1));
		1426	}
		1427
		1428	static void
		1429	t2_cb (EV_P_ ev_timer *w, int revents)
		1430	{
		1431	struct my_biggy big = (struct my_biggy *)
		1432	(((char *)w) - offsetof (struct my_biggy, t2));
		1433	}
		1434
		1435	=head2 WATCHER PRIORITY MODELS
		1436
		1437	Many event loops support I<watcher priorities>, which are usually small
		1438	integers that influence the ordering of event callback invocation
		1439	between watchers in some way, all else being equal.
		1440
		1441	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1442	description for the more technical details such as the actual priority
		1443	range.
		1444
		1445	There are two common ways how these these priorities are being interpreted
		1446	by event loops:
		1447
		1448	In the more common lock-out model, higher priorities "lock out" invocation
		1449	of lower priority watchers, which means as long as higher priority
		1450	watchers receive events, lower priority watchers are not being invoked.
		1451
		1452	The less common only-for-ordering model uses priorities solely to order
		1453	callback invocation within a single event loop iteration: Higher priority
		1454	watchers are invoked before lower priority ones, but they all get invoked
		1455	before polling for new events.
		1456
		1457	Libev uses the second (only-for-ordering) model for all its watchers
		1458	except for idle watchers (which use the lock-out model).
		1459
		1460	The rationale behind this is that implementing the lock-out model for
		1461	watchers is not well supported by most kernel interfaces, and most event
		1462	libraries will just poll for the same events again and again as long as
		1463	their callbacks have not been executed, which is very inefficient in the
		1464	common case of one high-priority watcher locking out a mass of lower
		1465	priority ones.
		1466
		1467	Static (ordering) priorities are most useful when you have two or more
		1468	watchers handling the same resource: a typical usage example is having an
		1469	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1470	timeouts. Under load, data might be received while the program handles
		1471	other jobs, but since timers normally get invoked first, the timeout
		1472	handler will be executed before checking for data. In that case, giving
		1473	the timer a lower priority than the I/O watcher ensures that I/O will be
		1474	handled first even under adverse conditions (which is usually, but not
		1475	always, what you want).
		1476
		1477	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1478	will only be executed when no same or higher priority watchers have
		1479	received events, they can be used to implement the "lock-out" model when
		1480	required.
		1481
		1482	For example, to emulate how many other event libraries handle priorities,
		1483	you can associate an C<ev_idle> watcher to each such watcher, and in
		1484	the normal watcher callback, you just start the idle watcher. The real
		1485	processing is done in the idle watcher callback. This causes libev to
		1486	continuously poll and process kernel event data for the watcher, but when
		1487	the lock-out case is known to be rare (which in turn is rare :), this is
		1488	workable.
		1489
		1490	Usually, however, the lock-out model implemented that way will perform
		1491	miserably under the type of load it was designed to handle. In that case,
		1492	it might be preferable to stop the real watcher before starting the
		1493	idle watcher, so the kernel will not have to process the event in case
		1494	the actual processing will be delayed for considerable time.
		1495
		1496	Here is an example of an I/O watcher that should run at a strictly lower
		1497	priority than the default, and which should only process data when no
		1498	other events are pending:
		1499
		1500	ev_idle idle; // actual processing watcher
		1501	ev_io io; // actual event watcher
		1502
		1503	static void
		1504	io_cb (EV_P_ ev_io *w, int revents)
		1505	{
		1506	// stop the I/O watcher, we received the event, but
		1507	// are not yet ready to handle it.
		1508	ev_io_stop (EV_A_ w);
		1509
		1510	// start the idle watcher to handle the actual event.
		1511	// it will not be executed as long as other watchers
		1512	// with the default priority are receiving events.
		1513	ev_idle_start (EV_A_ &idle);
		1514	}
		1515
		1516	static void
		1517	idle_cb (EV_P_ ev_idle *w, int revents)
		1518	{
		1519	// actual processing
		1520	read (STDIN_FILENO, ...);
		1521
		1522	// have to start the I/O watcher again, as
		1523	// we have handled the event
		1524	ev_io_start (EV_P_ &io);
		1525	}
		1526
		1527	// initialisation
		1528	ev_idle_init (&idle, idle_cb);
		1529	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1530	ev_io_start (EV_DEFAULT_ &io);
		1531
		1532	In the "real" world, it might also be beneficial to start a timer, so that
		1533	low-priority connections can not be locked out forever under load. This
		1534	enables your program to keep a lower latency for important connections
		1535	during short periods of high load, while not completely locking out less
		1536	important ones.
427		1537
428		1538
429	=head1 WATCHER TYPES	1539	=head1 WATCHER TYPES
430		1540
431	This section describes each watcher in detail, but will not repeat	1541	This section describes each watcher in detail, but will not repeat
432	information given in the last section.	1542	information given in the last section. Any initialisation/set macros,
		1543	functions and members specific to the watcher type are explained.
433		1544
		1545	Members are additionally marked with either I<[read-only]>, meaning that,
		1546	while the watcher is active, you can look at the member and expect some
		1547	sensible content, but you must not modify it (you can modify it while the
		1548	watcher is stopped to your hearts content), or I<[read-write]>, which
		1549	means you can expect it to have some sensible content while the watcher
		1550	is active, but you can also modify it. Modifying it may not do something
		1551	sensible or take immediate effect (or do anything at all), but libev will
		1552	not crash or malfunction in any way.
		1553
		1554
434	=head2 C<ev_io> - is this file descriptor readable or writable	1555	=head2 C<ev_io> - is this file descriptor readable or writable?
435		1556
436	I/O watchers check whether a file descriptor is readable or writable	1557	I/O watchers check whether a file descriptor is readable or writable
437	in each iteration of the event loop (This behaviour is called	1558	in each iteration of the event loop, or, more precisely, when reading
438	level-triggering because you keep receiving events as long as the	1559	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	1560	some data. This behaviour is called level-triggering because you keep
440	act on the event and neither want to receive future events).	1561	receiving events as long as the condition persists. Remember you can stop
		1562	the watcher if you don't want to act on the event and neither want to
		1563	receive future events.
441		1564
442	In general you can register as many read and/or write event watchers per	1565	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	1566	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	1567	descriptors to non-blocking mode is also usually a good idea (but not
445	required if you know what you are doing).	1568	required if you know what you are doing).
446		1569
447	You have to be careful with dup'ed file descriptors, though. Some backends	1570	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	1571	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	1572	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
450	to the same underlying file/socket etc. description (that is, they share	1573	descriptors for which non-blocking operation makes no sense (such as
451	the same underlying "file open").	1574	files) - libev doesn't guarantee any specific behaviour in that case.
452		1575
453	If you must do this, then force the use of a known-to-be-good backend	1576	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	1577	receive "spurious" readiness notifications, that is your callback might
455	EVMETHOD_POLL).	1578	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1579	because there is no data. Not only are some backends known to create a
		1580	lot of those (for example Solaris ports), it is very easy to get into
		1581	this situation even with a relatively standard program structure. Thus
		1582	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1583	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1584
		1585	If you cannot run the fd in non-blocking mode (for example you should
		1586	not play around with an Xlib connection), then you have to separately
		1587	re-test whether a file descriptor is really ready with a known-to-be good
		1588	interface such as poll (fortunately in our Xlib example, Xlib already
		1589	does this on its own, so its quite safe to use). Some people additionally
		1590	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1591	indefinitely.
		1592
		1593	But really, best use non-blocking mode.
		1594
		1595	=head3 The special problem of disappearing file descriptors
		1596
		1597	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1598	descriptor (either due to calling C<close> explicitly or any other means,
		1599	such as C<dup2>). The reason is that you register interest in some file
		1600	descriptor, but when it goes away, the operating system will silently drop
		1601	this interest. If another file descriptor with the same number then is
		1602	registered with libev, there is no efficient way to see that this is, in
		1603	fact, a different file descriptor.
		1604
		1605	To avoid having to explicitly tell libev about such cases, libev follows
		1606	the following policy: Each time C<ev_io_set> is being called, libev
		1607	will assume that this is potentially a new file descriptor, otherwise
		1608	it is assumed that the file descriptor stays the same. That means that
		1609	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1610	descriptor even if the file descriptor number itself did not change.
		1611
		1612	This is how one would do it normally anyway, the important point is that
		1613	the libev application should not optimise around libev but should leave
		1614	optimisations to libev.
		1615
		1616	=head3 The special problem of dup'ed file descriptors
		1617
		1618	Some backends (e.g. epoll), cannot register events for file descriptors,
		1619	but only events for the underlying file descriptions. That means when you
		1620	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1621	events for them, only one file descriptor might actually receive events.
		1622
		1623	There is no workaround possible except not registering events
		1624	for potentially C<dup ()>'ed file descriptors, or to resort to
		1625	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1626
		1627	=head3 The special problem of fork
		1628
		1629	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1630	useless behaviour. Libev fully supports fork, but needs to be told about
		1631	it in the child.
		1632
		1633	To support fork in your programs, you either have to call
		1634	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1635	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1636	C<EVBACKEND_POLL>.
		1637
		1638	=head3 The special problem of SIGPIPE
		1639
		1640	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1641	when writing to a pipe whose other end has been closed, your program gets
		1642	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1643	this is sensible behaviour, for daemons, this is usually undesirable.
		1644
		1645	So when you encounter spurious, unexplained daemon exits, make sure you
		1646	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1647	somewhere, as that would have given you a big clue).
		1648
		1649	=head3 The special problem of accept()ing when you can't
		1650
		1651	Many implementations of the POSIX C<accept> function (for example,
		1652	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1653	connection from the pending queue in all error cases.
		1654
		1655	For example, larger servers often run out of file descriptors (because
		1656	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1657	rejecting the connection, leading to libev signalling readiness on
		1658	the next iteration again (the connection still exists after all), and
		1659	typically causing the program to loop at 100% CPU usage.
		1660
		1661	Unfortunately, the set of errors that cause this issue differs between
		1662	operating systems, there is usually little the app can do to remedy the
		1663	situation, and no known thread-safe method of removing the connection to
		1664	cope with overload is known (to me).
		1665
		1666	One of the easiest ways to handle this situation is to just ignore it
		1667	- when the program encounters an overload, it will just loop until the
		1668	situation is over. While this is a form of busy waiting, no OS offers an
		1669	event-based way to handle this situation, so it's the best one can do.
		1670
		1671	A better way to handle the situation is to log any errors other than
		1672	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1673	messages, and continue as usual, which at least gives the user an idea of
		1674	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1675	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1676	usage.
		1677
		1678	If your program is single-threaded, then you could also keep a dummy file
		1679	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1680	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1681	close that fd, and create a new dummy fd. This will gracefully refuse
		1682	clients under typical overload conditions.
		1683
		1684	The last way to handle it is to simply log the error and C<exit>, as
		1685	is often done with C<malloc> failures, but this results in an easy
		1686	opportunity for a DoS attack.
		1687
		1688	=head3 Watcher-Specific Functions
456		1689
457	=over 4	1690	=over 4
458		1691
459	=item ev_io_init (ev_io *, callback, int fd, int events)	1692	=item ev_io_init (ev_io *, callback, int fd, int events)
460		1693
461	=item ev_io_set (ev_io *, int fd, int events)	1694	=item ev_io_set (ev_io *, int fd, int events)
462		1695
463	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1696	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 |	1697	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
465	EV_WRITE> to receive the given events.	1698	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1699
		1700	=item int fd [read-only]
		1701
		1702	The file descriptor being watched.
		1703
		1704	=item int events [read-only]
		1705
		1706	The events being watched.
466		1707
467	=back	1708	=back
468		1709
		1710	=head3 Examples
		1711
		1712	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1713	readable, but only once. Since it is likely line-buffered, you could
		1714	attempt to read a whole line in the callback.
		1715
		1716	static void
		1717	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1718	{
		1719	ev_io_stop (loop, w);
		1720	.. read from stdin here (or from w->fd) and handle any I/O errors
		1721	}
		1722
		1723	...
		1724	struct ev_loop *loop = ev_default_init (0);
		1725	ev_io stdin_readable;
		1726	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1727	ev_io_start (loop, &stdin_readable);
		1728	ev_run (loop, 0);
		1729
		1730
469	=head2 C<ev_timer> - relative and optionally recurring timeouts	1731	=head2 C<ev_timer> - relative and optionally repeating timeouts
470		1732
471	Timer watchers are simple relative timers that generate an event after a	1733	Timer watchers are simple relative timers that generate an event after a
472	given time, and optionally repeating in regular intervals after that.	1734	given time, and optionally repeating in regular intervals after that.
473		1735
474	The timers are based on real time, that is, if you register an event that	1736	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	1737	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	1738	year, it will still time out after (roughly) one hour. "Roughly" because
477	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1739	detecting time jumps is hard, and some inaccuracies are unavoidable (the
478	monotonic clock option helps a lot here).	1740	monotonic clock option helps a lot here).
		1741
		1742	The callback is guaranteed to be invoked only I<after> its timeout has
		1743	passed (not I<at>, so on systems with very low-resolution clocks this
		1744	might introduce a small delay). If multiple timers become ready during the
		1745	same loop iteration then the ones with earlier time-out values are invoked
		1746	before ones of the same priority with later time-out values (but this is
		1747	no longer true when a callback calls C<ev_run> recursively).
		1748
		1749	=head3 Be smart about timeouts
		1750
		1751	Many real-world problems involve some kind of timeout, usually for error
		1752	recovery. A typical example is an HTTP request - if the other side hangs,
		1753	you want to raise some error after a while.
		1754
		1755	What follows are some ways to handle this problem, from obvious and
		1756	inefficient to smart and efficient.
		1757
		1758	In the following, a 60 second activity timeout is assumed - a timeout that
		1759	gets reset to 60 seconds each time there is activity (e.g. each time some
		1760	data or other life sign was received).
		1761
		1762	=over 4
		1763
		1764	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1765
		1766	This is the most obvious, but not the most simple way: In the beginning,
		1767	start the watcher:
		1768
		1769	ev_timer_init (timer, callback, 60., 0.);
		1770	ev_timer_start (loop, timer);
		1771
		1772	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1773	and start it again:
		1774
		1775	ev_timer_stop (loop, timer);
		1776	ev_timer_set (timer, 60., 0.);
		1777	ev_timer_start (loop, timer);
		1778
		1779	This is relatively simple to implement, but means that each time there is
		1780	some activity, libev will first have to remove the timer from its internal
		1781	data structure and then add it again. Libev tries to be fast, but it's
		1782	still not a constant-time operation.
		1783
		1784	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1785
		1786	This is the easiest way, and involves using C<ev_timer_again> instead of
		1787	C<ev_timer_start>.
		1788
		1789	To implement this, configure an C<ev_timer> with a C<repeat> value
		1790	of C<60> and then call C<ev_timer_again> at start and each time you
		1791	successfully read or write some data. If you go into an idle state where
		1792	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1793	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1794
		1795	That means you can ignore both the C<ev_timer_start> function and the
		1796	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1797	member and C<ev_timer_again>.
		1798
		1799	At start:
		1800
		1801	ev_init (timer, callback);
		1802	timer->repeat = 60.;
		1803	ev_timer_again (loop, timer);
		1804
		1805	Each time there is some activity:
		1806
		1807	ev_timer_again (loop, timer);
		1808
		1809	It is even possible to change the time-out on the fly, regardless of
		1810	whether the watcher is active or not:
		1811
		1812	timer->repeat = 30.;
		1813	ev_timer_again (loop, timer);
		1814
		1815	This is slightly more efficient then stopping/starting the timer each time
		1816	you want to modify its timeout value, as libev does not have to completely
		1817	remove and re-insert the timer from/into its internal data structure.
		1818
		1819	It is, however, even simpler than the "obvious" way to do it.
		1820
		1821	=item 3. Let the timer time out, but then re-arm it as required.
		1822
		1823	This method is more tricky, but usually most efficient: Most timeouts are
		1824	relatively long compared to the intervals between other activity - in
		1825	our example, within 60 seconds, there are usually many I/O events with
		1826	associated activity resets.
		1827
		1828	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1829	but remember the time of last activity, and check for a real timeout only
		1830	within the callback:
		1831
		1832	ev_tstamp last_activity; // time of last activity
		1833
		1834	static void
		1835	callback (EV_P_ ev_timer *w, int revents)
		1836	{
		1837	ev_tstamp now = ev_now (EV_A);
		1838	ev_tstamp timeout = last_activity + 60.;
		1839
		1840	// if last_activity + 60. is older than now, we did time out
		1841	if (timeout < now)
		1842	{
		1843	// timeout occurred, take action
		1844	}
		1845	else
		1846	{
		1847	// callback was invoked, but there was some activity, re-arm
		1848	// the watcher to fire in last_activity + 60, which is
		1849	// guaranteed to be in the future, so "again" is positive:
		1850	w->repeat = timeout - now;
		1851	ev_timer_again (EV_A_ w);
		1852	}
		1853	}
		1854
		1855	To summarise the callback: first calculate the real timeout (defined
		1856	as "60 seconds after the last activity"), then check if that time has
		1857	been reached, which means something I<did>, in fact, time out. Otherwise
		1858	the callback was invoked too early (C<timeout> is in the future), so
		1859	re-schedule the timer to fire at that future time, to see if maybe we have
		1860	a timeout then.
		1861
		1862	Note how C<ev_timer_again> is used, taking advantage of the
		1863	C<ev_timer_again> optimisation when the timer is already running.
		1864
		1865	This scheme causes more callback invocations (about one every 60 seconds
		1866	minus half the average time between activity), but virtually no calls to
		1867	libev to change the timeout.
		1868
		1869	To start the timer, simply initialise the watcher and set C<last_activity>
		1870	to the current time (meaning we just have some activity :), then call the
		1871	callback, which will "do the right thing" and start the timer:
		1872
		1873	ev_init (timer, callback);
		1874	last_activity = ev_now (loop);
		1875	callback (loop, timer, EV_TIMER);
		1876
		1877	And when there is some activity, simply store the current time in
		1878	C<last_activity>, no libev calls at all:
		1879
		1880	last_activity = ev_now (loop);
		1881
		1882	This technique is slightly more complex, but in most cases where the
		1883	time-out is unlikely to be triggered, much more efficient.
		1884
		1885	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1886	callback :) - just change the timeout and invoke the callback, which will
		1887	fix things for you.
		1888
		1889	=item 4. Wee, just use a double-linked list for your timeouts.
		1890
		1891	If there is not one request, but many thousands (millions...), all
		1892	employing some kind of timeout with the same timeout value, then one can
		1893	do even better:
		1894
		1895	When starting the timeout, calculate the timeout value and put the timeout
		1896	at the I<end> of the list.
		1897
		1898	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1899	the list is expected to fire (for example, using the technique #3).
		1900
		1901	When there is some activity, remove the timer from the list, recalculate
		1902	the timeout, append it to the end of the list again, and make sure to
		1903	update the C<ev_timer> if it was taken from the beginning of the list.
		1904
		1905	This way, one can manage an unlimited number of timeouts in O(1) time for
		1906	starting, stopping and updating the timers, at the expense of a major
		1907	complication, and having to use a constant timeout. The constant timeout
		1908	ensures that the list stays sorted.
		1909
		1910	=back
		1911
		1912	So which method the best?
		1913
		1914	Method #2 is a simple no-brain-required solution that is adequate in most
		1915	situations. Method #3 requires a bit more thinking, but handles many cases
		1916	better, and isn't very complicated either. In most case, choosing either
		1917	one is fine, with #3 being better in typical situations.
		1918
		1919	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1920	rather complicated, but extremely efficient, something that really pays
		1921	off after the first million or so of active timers, i.e. it's usually
		1922	overkill :)
		1923
		1924	=head3 The special problem of time updates
		1925
		1926	Establishing the current time is a costly operation (it usually takes at
		1927	least two system calls): EV therefore updates its idea of the current
		1928	time only before and after C<ev_run> collects new events, which causes a
		1929	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1930	lots of events in one iteration.
479		1931
480	The relative timeouts are calculated relative to the C<ev_now ()>	1932	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	1933	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	1934	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	1935	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:	1936	timeout on the current time, use something like this to adjust for this:
485		1937
486	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1938	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
487		1939
		1940	If the event loop is suspended for a long time, you can also force an
		1941	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1942	()>.
		1943
		1944	=head3 The special problems of suspended animation
		1945
		1946	When you leave the server world it is quite customary to hit machines that
		1947	can suspend/hibernate - what happens to the clocks during such a suspend?
		1948
		1949	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1950	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1951	to run until the system is suspended, but they will not advance while the
		1952	system is suspended. That means, on resume, it will be as if the program
		1953	was frozen for a few seconds, but the suspend time will not be counted
		1954	towards C<ev_timer> when a monotonic clock source is used. The real time
		1955	clock advanced as expected, but if it is used as sole clocksource, then a
		1956	long suspend would be detected as a time jump by libev, and timers would
		1957	be adjusted accordingly.
		1958
		1959	I would not be surprised to see different behaviour in different between
		1960	operating systems, OS versions or even different hardware.
		1961
		1962	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1963	time jump in the monotonic clocks and the realtime clock. If the program
		1964	is suspended for a very long time, and monotonic clock sources are in use,
		1965	then you can expect C<ev_timer>s to expire as the full suspension time
		1966	will be counted towards the timers. When no monotonic clock source is in
		1967	use, then libev will again assume a timejump and adjust accordingly.
		1968
		1969	It might be beneficial for this latter case to call C<ev_suspend>
		1970	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1971	deterministic behaviour in this case (you can do nothing against
		1972	C<SIGSTOP>).
		1973
		1974	=head3 Watcher-Specific Functions and Data Members
		1975
488	=over 4	1976	=over 4
489		1977
490	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1978	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
491		1979
492	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1980	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
493		1981
494	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1982	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	1983	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	1984	reached. If it is positive, then the timer will automatically be
497	later, again, and again, until stopped manually.	1985	configured to trigger again C<repeat> seconds later, again, and again,
		1986	until stopped manually.
498		1987
499	The timer itself will do a best-effort at avoiding drift, that is, if you	1988	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	1989	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	1990	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	1991	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.	1992	do stuff) the timer will not fire more than once per event loop iteration.
504		1993
505	=item ev_timer_again (loop)	1994	=item ev_timer_again (loop, ev_timer *)
506		1995
507	This will act as if the timer timed out and restart it again if it is	1996	This will act as if the timer timed out and restart it again if it is
508	repeating. The exact semantics are:	1997	repeating. The exact semantics are:
509		1998
		1999	If the timer is pending, its pending status is cleared.
		2000
510	If the timer is started but nonrepeating, stop it.	2001	If the timer is started but non-repeating, stop it (as if it timed out).
511		2002
512	If the timer is repeating, either start it if necessary (with the repeat	2003	If the timer is repeating, either start it if necessary (with the
513	value), or reset the running timer to the repeat value.	2004	C<repeat> value), or reset the running timer to the C<repeat> value.
514		2005
515	This sounds a bit complicated, but here is a useful and typical	2006	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
516	example: Imagine you have a tcp connection and you want a so-called idle	2007	usage example.
517	timeout, that is, you want to be called when there have been, say, 60	2008
518	seconds of inactivity on the socket. The easiest way to do this is to	2009	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
519	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	2010
520	time you successfully read or write some data. If you go into an idle	2011	Returns the remaining time until a timer fires. If the timer is active,
521	state where you do not expect data to travel on the socket, you can stop	2012	then this time is relative to the current event loop time, otherwise it's
522	the timer, and again will automatically restart it if need be.	2013	the timeout value currently configured.
		2014
		2015	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		2016	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2017	will return C<4>. When the timer expires and is restarted, it will return
		2018	roughly C<7> (likely slightly less as callback invocation takes some time,
		2019	too), and so on.
		2020
		2021	=item ev_tstamp repeat [read-write]
		2022
		2023	The current C<repeat> value. Will be used each time the watcher times out
		2024	or C<ev_timer_again> is called, and determines the next timeout (if any),
		2025	which is also when any modifications are taken into account.
523		2026
524	=back	2027	=back
525		2028
		2029	=head3 Examples
		2030
		2031	Example: Create a timer that fires after 60 seconds.
		2032
		2033	static void
		2034	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2035	{
		2036	.. one minute over, w is actually stopped right here
		2037	}
		2038
		2039	ev_timer mytimer;
		2040	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		2041	ev_timer_start (loop, &mytimer);
		2042
		2043	Example: Create a timeout timer that times out after 10 seconds of
		2044	inactivity.
		2045
		2046	static void
		2047	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2048	{
		2049	.. ten seconds without any activity
		2050	}
		2051
		2052	ev_timer mytimer;
		2053	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		2054	ev_timer_again (&mytimer); /* start timer */
		2055	ev_run (loop, 0);
		2056
		2057	// and in some piece of code that gets executed on any "activity":
		2058	// reset the timeout to start ticking again at 10 seconds
		2059	ev_timer_again (&mytimer);
		2060
		2061
526	=head2 C<ev_periodic> - to cron or not to cron	2062	=head2 C<ev_periodic> - to cron or not to cron?
527		2063
528	Periodic watchers are also timers of a kind, but they are very versatile	2064	Periodic watchers are also timers of a kind, but they are very versatile
529	(and unfortunately a bit complex).	2065	(and unfortunately a bit complex).
530		2066
531	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	2067	Unlike C<ev_timer>, periodic watchers are not based on real time (or
532	but on wallclock time (absolute time). You can tell a periodic watcher	2068	relative time, the physical time that passes) but on wall clock time
533	to trigger "at" some specific point in time. For example, if you tell a	2069	(absolute time, the thing you can read on your calender or clock). The
534	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	2070	difference is that wall clock time can run faster or slower than real
535	+ 10.>) and then reset your system clock to the last year, then it will	2071	time, and time jumps are not uncommon (e.g. when you adjust your
536	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	2072	wrist-watch).
537	roughly 10 seconds later and of course not if you reset your system time
538	again).
539		2073
540	They can also be used to implement vastly more complex timers, such as	2074	You can tell a periodic watcher to trigger after some specific point
		2075	in time: for example, if you tell a periodic watcher to trigger "in 10
		2076	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2077	not a delay) and then reset your system clock to January of the previous
		2078	year, then it will take a year or more to trigger the event (unlike an
		2079	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2080	it, as it uses a relative timeout).
		2081
		2082	C<ev_periodic> watchers can also be used to implement vastly more complex
541	triggering an event on eahc midnight, local time.	2083	timers, such as triggering an event on each "midnight, local time", or
		2084	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		2085	those cannot react to time jumps.
		2086
		2087	As with timers, the callback is guaranteed to be invoked only when the
		2088	point in time where it is supposed to trigger has passed. If multiple
		2089	timers become ready during the same loop iteration then the ones with
		2090	earlier time-out values are invoked before ones with later time-out values
		2091	(but this is no longer true when a callback calls C<ev_run> recursively).
		2092
		2093	=head3 Watcher-Specific Functions and Data Members
542		2094
543	=over 4	2095	=over 4
544		2096
545	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2097	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
546		2098
547	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2099	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
548		2100
549	Lots of arguments, lets sort it out... There are basically three modes of	2101	Lots of arguments, let's sort it out... There are basically three modes of
550	operation, and we will explain them from simplest to complex:	2102	operation, and we will explain them from simplest to most complex:
551
552		2103
553	=over 4	2104	=over 4
554		2105
555	=item * absolute timer (interval = reschedule_cb = 0)	2106	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
556		2107
557	In this configuration the watcher triggers an event at the wallclock time	2108	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,	2109	time C<offset> has passed. It will not repeat and will not adjust when a
559	that is, if it is to be run at January 1st 2011 then it will run when the	2110	time jump occurs, that is, if it is to be run at January 1st 2011 then it
560	system time reaches or surpasses this time.	2111	will be stopped and invoked when the system clock reaches or surpasses
		2112	this point in time.
561		2113
562	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2114	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
563		2115
564	In this mode the watcher will always be scheduled to time out at the next	2116	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	2117	C<offset + N * interval> time (for some integer N, which can also be
566	of any time jumps.	2118	negative) and then repeat, regardless of any time jumps. The C<offset>
		2119	argument is merely an offset into the C<interval> periods.
567		2120
568	This can be used to create timers that do not drift with respect to system	2121	This can be used to create timers that do not drift with respect to the
569	time:	2122	system clock, for example, here is an C<ev_periodic> that triggers each
		2123	hour, on the hour (with respect to UTC):
570		2124
571	ev_periodic_set (&periodic, 0., 3600., 0);	2125	ev_periodic_set (&periodic, 0., 3600., 0);
572		2126
573	This doesn't mean there will always be 3600 seconds in between triggers,	2127	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	2128	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	2129	full hour (UTC), or more correctly, when the system time is evenly divisible
576	by 3600.	2130	by 3600.
577		2131
578	Another way to think about it (for the mathematically inclined) is that	2132	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	2133	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.	2134	time where C<time = offset (mod interval)>, regardless of any time jumps.
581		2135
		2136	For numerical stability it is preferable that the C<offset> value is near
		2137	C<ev_now ()> (the current time), but there is no range requirement for
		2138	this value, and in fact is often specified as zero.
		2139
		2140	Note also that there is an upper limit to how often a timer can fire (CPU
		2141	speed for example), so if C<interval> is very small then timing stability
		2142	will of course deteriorate. Libev itself tries to be exact to be about one
		2143	millisecond (if the OS supports it and the machine is fast enough).
		2144
582	=item * manual reschedule mode (reschedule_cb = callback)	2145	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
583		2146
584	In this mode the values for C<interval> and C<at> are both being	2147	In this mode the values for C<interval> and C<offset> are both being
585	ignored. Instead, each time the periodic watcher gets scheduled, the	2148	ignored. Instead, each time the periodic watcher gets scheduled, the
586	reschedule callback will be called with the watcher as first, and the	2149	reschedule callback will be called with the watcher as first, and the
587	current time as second argument.	2150	current time as second argument.
588		2151
589	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2152	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
590	ever, or make any event loop modifications>. If you need to stop it,	2153	or make ANY other event loop modifications whatsoever, unless explicitly
591	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by	2154	allowed by documentation here>.
592	starting a prepare watcher).
593		2155
		2156	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2157	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2158	only event loop modification you are allowed to do).
		2159
594	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2160	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
595	ev_tstamp now)>, e.g.:	2161	*w, ev_tstamp now)>, e.g.:
596		2162
		2163	static ev_tstamp
597	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2164	my_rescheduler (ev_periodic *w, ev_tstamp now)
598	{	2165	{
599	return now + 60.;	2166	return now + 60.;
600	}	2167	}
601		2168
602	It must return the next time to trigger, based on the passed time value	2169	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	2170	(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	2171	will usually be called just before the callback will be triggered, but
605	might be called at other times, too.	2172	might be called at other times, too.
606		2173
607	NOTE: I<< This callback must always return a time that is later than the	2174	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.	2175	equal to the passed C<now> value >>.
609		2176
610	This can be used to create very complex timers, such as a timer that	2177	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	2178	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	2179	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	2180	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).	2181	reason I omitted it as an example).
615		2182
616	=back	2183	=back
…		…
620	Simply stops and restarts the periodic watcher again. This is only useful	2187	Simply stops and restarts the periodic watcher again. This is only useful
621	when you changed some parameters or the reschedule callback would return	2188	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	2189	a different time than the last time it was called (e.g. in a crond like
623	program when the crontabs have changed).	2190	program when the crontabs have changed).
624		2191
		2192	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2193
		2194	When active, returns the absolute time that the watcher is supposed
		2195	to trigger next. This is not the same as the C<offset> argument to
		2196	C<ev_periodic_set>, but indeed works even in interval and manual
		2197	rescheduling modes.
		2198
		2199	=item ev_tstamp offset [read-write]
		2200
		2201	When repeating, this contains the offset value, otherwise this is the
		2202	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2203	although libev might modify this value for better numerical stability).
		2204
		2205	Can be modified any time, but changes only take effect when the periodic
		2206	timer fires or C<ev_periodic_again> is being called.
		2207
		2208	=item ev_tstamp interval [read-write]
		2209
		2210	The current interval value. Can be modified any time, but changes only
		2211	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2212	called.
		2213
		2214	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2215
		2216	The current reschedule callback, or C<0>, if this functionality is
		2217	switched off. Can be changed any time, but changes only take effect when
		2218	the periodic timer fires or C<ev_periodic_again> is being called.
		2219
625	=back	2220	=back
626		2221
		2222	=head3 Examples
		2223
		2224	Example: Call a callback every hour, or, more precisely, whenever the
		2225	system time is divisible by 3600. The callback invocation times have
		2226	potentially a lot of jitter, but good long-term stability.
		2227
		2228	static void
		2229	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
		2230	{
		2231	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2232	}
		2233
		2234	ev_periodic hourly_tick;
		2235	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2236	ev_periodic_start (loop, &hourly_tick);
		2237
		2238	Example: The same as above, but use a reschedule callback to do it:
		2239
		2240	#include <math.h>
		2241
		2242	static ev_tstamp
		2243	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2244	{
		2245	return now + (3600. - fmod (now, 3600.));
		2246	}
		2247
		2248	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2249
		2250	Example: Call a callback every hour, starting now:
		2251
		2252	ev_periodic hourly_tick;
		2253	ev_periodic_init (&hourly_tick, clock_cb,
		2254	fmod (ev_now (loop), 3600.), 3600., 0);
		2255	ev_periodic_start (loop, &hourly_tick);
		2256
		2257
627	=head2 C<ev_signal> - signal me when a signal gets signalled	2258	=head2 C<ev_signal> - signal me when a signal gets signalled!
628		2259
629	Signal watchers will trigger an event when the process receives a specific	2260	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	2261	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	2262	will try it's best to deliver signals synchronously, i.e. as part of the
632	normal event processing, like any other event.	2263	normal event processing, like any other event.
633		2264
		2265	If you want signals to be delivered truly asynchronously, just use
		2266	C<sigaction> as you would do without libev and forget about sharing
		2267	the signal. You can even use C<ev_async> from a signal handler to
		2268	synchronously wake up an event loop.
		2269
634	You can configure as many watchers as you like per signal. Only when the	2270	You can configure as many watchers as you like for the same signal, but
		2271	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2272	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2273	C<SIGINT> in both the default loop and another loop at the same time. At
		2274	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2275
635	first watcher gets started will libev actually register a signal watcher	2276	When the first watcher gets started will libev actually register something
636	with the kernel (thus it coexists with your own signal handlers as long	2277	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	2278	you don't register any with libev for the same signal).
638	watcher for a signal is stopped libev will reset the signal handler to	2279
639	SIG_DFL (regardless of what it was set to before).	2280	If possible and supported, libev will install its handlers with
		2281	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2282	not be unduly interrupted. If you have a problem with system calls getting
		2283	interrupted by signals you can block all signals in an C<ev_check> watcher
		2284	and unblock them in an C<ev_prepare> watcher.
		2285
		2286	=head3 The special problem of inheritance over fork/execve/pthread_create
		2287
		2288	Both the signal mask (C<sigprocmask>) and the signal disposition
		2289	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2290	stopping it again), that is, libev might or might not block the signal,
		2291	and might or might not set or restore the installed signal handler.
		2292
		2293	While this does not matter for the signal disposition (libev never
		2294	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2295	C<execve>), this matters for the signal mask: many programs do not expect
		2296	certain signals to be blocked.
		2297
		2298	This means that before calling C<exec> (from the child) you should reset
		2299	the signal mask to whatever "default" you expect (all clear is a good
		2300	choice usually).
		2301
		2302	The simplest way to ensure that the signal mask is reset in the child is
		2303	to install a fork handler with C<pthread_atfork> that resets it. That will
		2304	catch fork calls done by libraries (such as the libc) as well.
		2305
		2306	In current versions of libev, the signal will not be blocked indefinitely
		2307	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2308	the window of opportunity for problems, it will not go away, as libev
		2309	I<has> to modify the signal mask, at least temporarily.
		2310
		2311	So I can't stress this enough: I<If you do not reset your signal mask when
		2312	you expect it to be empty, you have a race condition in your code>. This
		2313	is not a libev-specific thing, this is true for most event libraries.
		2314
		2315	=head3 Watcher-Specific Functions and Data Members
640		2316
641	=over 4	2317	=over 4
642		2318
643	=item ev_signal_init (ev_signal *, callback, int signum)	2319	=item ev_signal_init (ev_signal *, callback, int signum)
644		2320
645	=item ev_signal_set (ev_signal *, int signum)	2321	=item ev_signal_set (ev_signal *, int signum)
646		2322
647	Configures the watcher to trigger on the given signal number (usually one	2323	Configures the watcher to trigger on the given signal number (usually one
648	of the C<SIGxxx> constants).	2324	of the C<SIGxxx> constants).
649		2325
		2326	=item int signum [read-only]
		2327
		2328	The signal the watcher watches out for.
		2329
650	=back	2330	=back
651		2331
		2332	=head3 Examples
		2333
		2334	Example: Try to exit cleanly on SIGINT.
		2335
		2336	static void
		2337	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2338	{
		2339	ev_break (loop, EVBREAK_ALL);
		2340	}
		2341
		2342	ev_signal signal_watcher;
		2343	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2344	ev_signal_start (loop, &signal_watcher);
		2345
		2346
652	=head2 C<ev_child> - wait for pid status changes	2347	=head2 C<ev_child> - watch out for process status changes
653		2348
654	Child watchers trigger when your process receives a SIGCHLD in response to	2349	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).	2350	some child status changes (most typically when a child of yours dies or
		2351	exits). It is permissible to install a child watcher I<after> the child
		2352	has been forked (which implies it might have already exited), as long
		2353	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2354	forking and then immediately registering a watcher for the child is fine,
		2355	but forking and registering a watcher a few event loop iterations later or
		2356	in the next callback invocation is not.
		2357
		2358	Only the default event loop is capable of handling signals, and therefore
		2359	you can only register child watchers in the default event loop.
		2360
		2361	Due to some design glitches inside libev, child watchers will always be
		2362	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2363	libev)
		2364
		2365	=head3 Process Interaction
		2366
		2367	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2368	initialised. This is necessary to guarantee proper behaviour even if the
		2369	first child watcher is started after the child exits. The occurrence
		2370	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2371	synchronously as part of the event loop processing. Libev always reaps all
		2372	children, even ones not watched.
		2373
		2374	=head3 Overriding the Built-In Processing
		2375
		2376	Libev offers no special support for overriding the built-in child
		2377	processing, but if your application collides with libev's default child
		2378	handler, you can override it easily by installing your own handler for
		2379	C<SIGCHLD> after initialising the default loop, and making sure the
		2380	default loop never gets destroyed. You are encouraged, however, to use an
		2381	event-based approach to child reaping and thus use libev's support for
		2382	that, so other libev users can use C<ev_child> watchers freely.
		2383
		2384	=head3 Stopping the Child Watcher
		2385
		2386	Currently, the child watcher never gets stopped, even when the
		2387	child terminates, so normally one needs to stop the watcher in the
		2388	callback. Future versions of libev might stop the watcher automatically
		2389	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2390	problem).
		2391
		2392	=head3 Watcher-Specific Functions and Data Members
656		2393
657	=over 4	2394	=over 4
658		2395
659	=item ev_child_init (ev_child *, callback, int pid)	2396	=item ev_child_init (ev_child *, callback, int pid, int trace)
660		2397
661	=item ev_child_set (ev_child *, int pid)	2398	=item ev_child_set (ev_child *, int pid, int trace)
662		2399
663	Configures the watcher to wait for status changes of process C<pid> (or	2400	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	2401	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	2402	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	2403	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	2404	C<waitpid> documentation). The C<rpid> member contains the pid of the
668	process causing the status change.	2405	process causing the status change. C<trace> must be either C<0> (only
		2406	activate the watcher when the process terminates) or C<1> (additionally
		2407	activate the watcher when the process is stopped or continued).
		2408
		2409	=item int pid [read-only]
		2410
		2411	The process id this watcher watches out for, or C<0>, meaning any process id.
		2412
		2413	=item int rpid [read-write]
		2414
		2415	The process id that detected a status change.
		2416
		2417	=item int rstatus [read-write]
		2418
		2419	The process exit/trace status caused by C<rpid> (see your systems
		2420	C<waitpid> and C<sys/wait.h> documentation for details).
669		2421
670	=back	2422	=back
671		2423
		2424	=head3 Examples
		2425
		2426	Example: C<fork()> a new process and install a child handler to wait for
		2427	its completion.
		2428
		2429	ev_child cw;
		2430
		2431	static void
		2432	child_cb (EV_P_ ev_child *w, int revents)
		2433	{
		2434	ev_child_stop (EV_A_ w);
		2435	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2436	}
		2437
		2438	pid_t pid = fork ();
		2439
		2440	if (pid < 0)
		2441	// error
		2442	else if (pid == 0)
		2443	{
		2444	// the forked child executes here
		2445	exit (1);
		2446	}
		2447	else
		2448	{
		2449	ev_child_init (&cw, child_cb, pid, 0);
		2450	ev_child_start (EV_DEFAULT_ &cw);
		2451	}
		2452
		2453
		2454	=head2 C<ev_stat> - did the file attributes just change?
		2455
		2456	This watches a file system path for attribute changes. That is, it calls
		2457	C<stat> on that path in regular intervals (or when the OS says it changed)
		2458	and sees if it changed compared to the last time, invoking the callback if
		2459	it did.
		2460
		2461	The path does not need to exist: changing from "path exists" to "path does
		2462	not exist" is a status change like any other. The condition "path does not
		2463	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2464	C<st_nlink> field being zero (which is otherwise always forced to be at
		2465	least one) and all the other fields of the stat buffer having unspecified
		2466	contents.
		2467
		2468	The path I<must not> end in a slash or contain special components such as
		2469	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2470	your working directory changes, then the behaviour is undefined.
		2471
		2472	Since there is no portable change notification interface available, the
		2473	portable implementation simply calls C<stat(2)> regularly on the path
		2474	to see if it changed somehow. You can specify a recommended polling
		2475	interval for this case. If you specify a polling interval of C<0> (highly
		2476	recommended!) then a I<suitable, unspecified default> value will be used
		2477	(which you can expect to be around five seconds, although this might
		2478	change dynamically). Libev will also impose a minimum interval which is
		2479	currently around C<0.1>, but that's usually overkill.
		2480
		2481	This watcher type is not meant for massive numbers of stat watchers,
		2482	as even with OS-supported change notifications, this can be
		2483	resource-intensive.
		2484
		2485	At the time of this writing, the only OS-specific interface implemented
		2486	is the Linux inotify interface (implementing kqueue support is left as an
		2487	exercise for the reader. Note, however, that the author sees no way of
		2488	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2489
		2490	=head3 ABI Issues (Largefile Support)
		2491
		2492	Libev by default (unless the user overrides this) uses the default
		2493	compilation environment, which means that on systems with large file
		2494	support disabled by default, you get the 32 bit version of the stat
		2495	structure. When using the library from programs that change the ABI to
		2496	use 64 bit file offsets the programs will fail. In that case you have to
		2497	compile libev with the same flags to get binary compatibility. This is
		2498	obviously the case with any flags that change the ABI, but the problem is
		2499	most noticeably displayed with ev_stat and large file support.
		2500
		2501	The solution for this is to lobby your distribution maker to make large
		2502	file interfaces available by default (as e.g. FreeBSD does) and not
		2503	optional. Libev cannot simply switch on large file support because it has
		2504	to exchange stat structures with application programs compiled using the
		2505	default compilation environment.
		2506
		2507	=head3 Inotify and Kqueue
		2508
		2509	When C<inotify (7)> support has been compiled into libev and present at
		2510	runtime, it will be used to speed up change detection where possible. The
		2511	inotify descriptor will be created lazily when the first C<ev_stat>
		2512	watcher is being started.
		2513
		2514	Inotify presence does not change the semantics of C<ev_stat> watchers
		2515	except that changes might be detected earlier, and in some cases, to avoid
		2516	making regular C<stat> calls. Even in the presence of inotify support
		2517	there are many cases where libev has to resort to regular C<stat> polling,
		2518	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2519	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2520	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2521	xfs are fully working) libev usually gets away without polling.
		2522
		2523	There is no support for kqueue, as apparently it cannot be used to
		2524	implement this functionality, due to the requirement of having a file
		2525	descriptor open on the object at all times, and detecting renames, unlinks
		2526	etc. is difficult.
		2527
		2528	=head3 C<stat ()> is a synchronous operation
		2529
		2530	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2531	the process. The exception are C<ev_stat> watchers - those call C<stat
		2532	()>, which is a synchronous operation.
		2533
		2534	For local paths, this usually doesn't matter: unless the system is very
		2535	busy or the intervals between stat's are large, a stat call will be fast,
		2536	as the path data is usually in memory already (except when starting the
		2537	watcher).
		2538
		2539	For networked file systems, calling C<stat ()> can block an indefinite
		2540	time due to network issues, and even under good conditions, a stat call
		2541	often takes multiple milliseconds.
		2542
		2543	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2544	paths, although this is fully supported by libev.
		2545
		2546	=head3 The special problem of stat time resolution
		2547
		2548	The C<stat ()> system call only supports full-second resolution portably,
		2549	and even on systems where the resolution is higher, most file systems
		2550	still only support whole seconds.
		2551
		2552	That means that, if the time is the only thing that changes, you can
		2553	easily miss updates: on the first update, C<ev_stat> detects a change and
		2554	calls your callback, which does something. When there is another update
		2555	within the same second, C<ev_stat> will be unable to detect unless the
		2556	stat data does change in other ways (e.g. file size).
		2557
		2558	The solution to this is to delay acting on a change for slightly more
		2559	than a second (or till slightly after the next full second boundary), using
		2560	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2561	ev_timer_again (loop, w)>).
		2562
		2563	The C<.02> offset is added to work around small timing inconsistencies
		2564	of some operating systems (where the second counter of the current time
		2565	might be be delayed. One such system is the Linux kernel, where a call to
		2566	C<gettimeofday> might return a timestamp with a full second later than
		2567	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2568	update file times then there will be a small window where the kernel uses
		2569	the previous second to update file times but libev might already execute
		2570	the timer callback).
		2571
		2572	=head3 Watcher-Specific Functions and Data Members
		2573
		2574	=over 4
		2575
		2576	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2577
		2578	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2579
		2580	Configures the watcher to wait for status changes of the given
		2581	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2582	be detected and should normally be specified as C<0> to let libev choose
		2583	a suitable value. The memory pointed to by C<path> must point to the same
		2584	path for as long as the watcher is active.
		2585
		2586	The callback will receive an C<EV_STAT> event when a change was detected,
		2587	relative to the attributes at the time the watcher was started (or the
		2588	last change was detected).
		2589
		2590	=item ev_stat_stat (loop, ev_stat *)
		2591
		2592	Updates the stat buffer immediately with new values. If you change the
		2593	watched path in your callback, you could call this function to avoid
		2594	detecting this change (while introducing a race condition if you are not
		2595	the only one changing the path). Can also be useful simply to find out the
		2596	new values.
		2597
		2598	=item ev_statdata attr [read-only]
		2599
		2600	The most-recently detected attributes of the file. Although the type is
		2601	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2602	suitable for your system, but you can only rely on the POSIX-standardised
		2603	members to be present. If the C<st_nlink> member is C<0>, then there was
		2604	some error while C<stat>ing the file.
		2605
		2606	=item ev_statdata prev [read-only]
		2607
		2608	The previous attributes of the file. The callback gets invoked whenever
		2609	C<prev> != C<attr>, or, more precisely, one or more of these members
		2610	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2611	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2612
		2613	=item ev_tstamp interval [read-only]
		2614
		2615	The specified interval.
		2616
		2617	=item const char *path [read-only]
		2618
		2619	The file system path that is being watched.
		2620
		2621	=back
		2622
		2623	=head3 Examples
		2624
		2625	Example: Watch C</etc/passwd> for attribute changes.
		2626
		2627	static void
		2628	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2629	{
		2630	/* /etc/passwd changed in some way */
		2631	if (w->attr.st_nlink)
		2632	{
		2633	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2634	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2635	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2636	}
		2637	else
		2638	/* you shalt not abuse printf for puts */
		2639	puts ("wow, /etc/passwd is not there, expect problems. "
		2640	"if this is windows, they already arrived\n");
		2641	}
		2642
		2643	...
		2644	ev_stat passwd;
		2645
		2646	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2647	ev_stat_start (loop, &passwd);
		2648
		2649	Example: Like above, but additionally use a one-second delay so we do not
		2650	miss updates (however, frequent updates will delay processing, too, so
		2651	one might do the work both on C<ev_stat> callback invocation I<and> on
		2652	C<ev_timer> callback invocation).
		2653
		2654	static ev_stat passwd;
		2655	static ev_timer timer;
		2656
		2657	static void
		2658	timer_cb (EV_P_ ev_timer *w, int revents)
		2659	{
		2660	ev_timer_stop (EV_A_ w);
		2661
		2662	/* now it's one second after the most recent passwd change */
		2663	}
		2664
		2665	static void
		2666	stat_cb (EV_P_ ev_stat *w, int revents)
		2667	{
		2668	/* reset the one-second timer */
		2669	ev_timer_again (EV_A_ &timer);
		2670	}
		2671
		2672	...
		2673	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2674	ev_stat_start (loop, &passwd);
		2675	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2676
		2677
672	=head2 C<ev_idle> - when you've got nothing better to do	2678	=head2 C<ev_idle> - when you've got nothing better to do...
673		2679
674	Idle watchers trigger events when there are no other events are pending	2680	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	2681	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,	2682	as receiving "events").
677	imagine) it will not be triggered. But when your process is idle all idle	2683
678	watchers are being called again and again, once per event loop iteration -	2684	That is, as long as your process is busy handling sockets or timeouts
		2685	(or even signals, imagine) of the same or higher priority it will not be
		2686	triggered. But when your process is idle (or only lower-priority watchers
		2687	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	2688	iteration - until stopped, that is, or your process receives more events
680	busy.	2689	and becomes busy again with higher priority stuff.
681		2690
682	The most noteworthy effect is that as long as any idle watchers are	2691	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.	2692	active, the process will not block when waiting for new events.
684		2693
685	Apart from keeping your process non-blocking (which is a useful	2694	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	2695	effect on its own sometimes), idle watchers are a good place to do
687	"pseudo-background processing", or delay processing stuff to after the	2696	"pseudo-background processing", or delay processing stuff to after the
688	event loop has handled all outstanding events.	2697	event loop has handled all outstanding events.
689		2698
		2699	=head3 Watcher-Specific Functions and Data Members
		2700
690	=over 4	2701	=over 4
691		2702
692	=item ev_idle_init (ev_signal *, callback)	2703	=item ev_idle_init (ev_idle *, callback)
693		2704
694	Initialises and configures the idle watcher - it has no parameters of any	2705	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,	2706	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
696	believe me.	2707	believe me.
697		2708
698	=back	2709	=back
699		2710
		2711	=head3 Examples
		2712
		2713	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2714	callback, free it. Also, use no error checking, as usual.
		2715
		2716	static void
		2717	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2718	{
		2719	free (w);
		2720	// now do something you wanted to do when the program has
		2721	// no longer anything immediate to do.
		2722	}
		2723
		2724	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2725	ev_idle_init (idle_watcher, idle_cb);
		2726	ev_idle_start (loop, idle_watcher);
		2727
		2728
700	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2729	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
701		2730
702	Prepare and check watchers are usually (but not always) used in tandem:	2731	Prepare and check watchers are usually (but not always) used in pairs:
703	prepare watchers get invoked before the process blocks and check watchers	2732	prepare watchers get invoked before the process blocks and check watchers
704	afterwards.	2733	afterwards.
705		2734
		2735	You I<must not> call C<ev_run> or similar functions that enter
		2736	the current event loop from either C<ev_prepare> or C<ev_check>
		2737	watchers. Other loops than the current one are fine, however. The
		2738	rationale behind this is that you do not need to check for recursion in
		2739	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2740	C<ev_check> so if you have one watcher of each kind they will always be
		2741	called in pairs bracketing the blocking call.
		2742
706	Their main purpose is to integrate other event mechanisms into libev. This	2743	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	2744	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.	2745	variable changes, implement your own watchers, integrate net-snmp or a
		2746	coroutine library and lots more. They are also occasionally useful if
		2747	you cache some data and want to flush it before blocking (for example,
		2748	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2749	watcher).
709		2750
710	This is done by examining in each prepare call which file descriptors need	2751	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	2752	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	2753	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	2754	libraries provide exactly this functionality). Then, in the check watcher,
714	any events that occured (by checking the pending status of all watchers	2755	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	2756	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,	2757	I/O and timer callbacks will never actually be called (but must be valid
717	because you never know, you know?).	2758	nevertheless, because you never know, you know?).
718		2759
719	As another example, the Perl Coro module uses these hooks to integrate	2760	As another example, the Perl Coro module uses these hooks to integrate
720	coroutines into libev programs, by yielding to other active coroutines	2761	coroutines into libev programs, by yielding to other active coroutines
721	during each prepare and only letting the process block if no coroutines	2762	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	2763	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	2764	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	2765	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	2766	loop from blocking if lower-priority coroutines are active, thus mapping
726	low-priority coroutines to idle/background tasks).	2767	low-priority coroutines to idle/background tasks).
727		2768
		2769	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2770	priority, to ensure that they are being run before any other watchers
		2771	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2772
		2773	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2774	activate ("feed") events into libev. While libev fully supports this, they
		2775	might get executed before other C<ev_check> watchers did their job. As
		2776	C<ev_check> watchers are often used to embed other (non-libev) event
		2777	loops those other event loops might be in an unusable state until their
		2778	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2779	others).
		2780
		2781	=head3 Watcher-Specific Functions and Data Members
		2782
728	=over 4	2783	=over 4
729		2784
730	=item ev_prepare_init (ev_prepare *, callback)	2785	=item ev_prepare_init (ev_prepare *, callback)
731		2786
732	=item ev_check_init (ev_check *, callback)	2787	=item ev_check_init (ev_check *, callback)
733		2788
734	Initialises and configures the prepare or check watcher - they have no	2789	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>	2790	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.	2791	macros, but using them is utterly, utterly, utterly and completely
		2792	pointless.
737		2793
738	=back	2794	=back
739		2795
		2796	=head3 Examples
		2797
		2798	There are a number of principal ways to embed other event loops or modules
		2799	into libev. Here are some ideas on how to include libadns into libev
		2800	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2801	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2802	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2803	Glib event loop).
		2804
		2805	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2806	and in a check watcher, destroy them and call into libadns. What follows
		2807	is pseudo-code only of course. This requires you to either use a low
		2808	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2809	the callbacks for the IO/timeout watchers might not have been called yet.
		2810
		2811	static ev_io iow [nfd];
		2812	static ev_timer tw;
		2813
		2814	static void
		2815	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2816	{
		2817	}
		2818
		2819	// create io watchers for each fd and a timer before blocking
		2820	static void
		2821	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2822	{
		2823	int timeout = 3600000;
		2824	struct pollfd fds [nfd];
		2825	// actual code will need to loop here and realloc etc.
		2826	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2827
		2828	/* the callback is illegal, but won't be called as we stop during check */
		2829	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		2830	ev_timer_start (loop, &tw);
		2831
		2832	// create one ev_io per pollfd
		2833	for (int i = 0; i < nfd; ++i)
		2834	{
		2835	ev_io_init (iow + i, io_cb, fds [i].fd,
		2836	((fds [i].events & POLLIN ? EV_READ : 0)
		2837	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2838
		2839	fds [i].revents = 0;
		2840	ev_io_start (loop, iow + i);
		2841	}
		2842	}
		2843
		2844	// stop all watchers after blocking
		2845	static void
		2846	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2847	{
		2848	ev_timer_stop (loop, &tw);
		2849
		2850	for (int i = 0; i < nfd; ++i)
		2851	{
		2852	// set the relevant poll flags
		2853	// could also call adns_processreadable etc. here
		2854	struct pollfd *fd = fds + i;
		2855	int revents = ev_clear_pending (iow + i);
		2856	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2857	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2858
		2859	// now stop the watcher
		2860	ev_io_stop (loop, iow + i);
		2861	}
		2862
		2863	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2864	}
		2865
		2866	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2867	in the prepare watcher and would dispose of the check watcher.
		2868
		2869	Method 3: If the module to be embedded supports explicit event
		2870	notification (libadns does), you can also make use of the actual watcher
		2871	callbacks, and only destroy/create the watchers in the prepare watcher.
		2872
		2873	static void
		2874	timer_cb (EV_P_ ev_timer *w, int revents)
		2875	{
		2876	adns_state ads = (adns_state)w->data;
		2877	update_now (EV_A);
		2878
		2879	adns_processtimeouts (ads, &tv_now);
		2880	}
		2881
		2882	static void
		2883	io_cb (EV_P_ ev_io *w, int revents)
		2884	{
		2885	adns_state ads = (adns_state)w->data;
		2886	update_now (EV_A);
		2887
		2888	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2889	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2890	}
		2891
		2892	// do not ever call adns_afterpoll
		2893
		2894	Method 4: Do not use a prepare or check watcher because the module you
		2895	want to embed is not flexible enough to support it. Instead, you can
		2896	override their poll function. The drawback with this solution is that the
		2897	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2898	this approach, effectively embedding EV as a client into the horrible
		2899	libglib event loop.
		2900
		2901	static gint
		2902	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2903	{
		2904	int got_events = 0;
		2905
		2906	for (n = 0; n < nfds; ++n)
		2907	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2908
		2909	if (timeout >= 0)
		2910	// create/start timer
		2911
		2912	// poll
		2913	ev_run (EV_A_ 0);
		2914
		2915	// stop timer again
		2916	if (timeout >= 0)
		2917	ev_timer_stop (EV_A_ &to);
		2918
		2919	// stop io watchers again - their callbacks should have set
		2920	for (n = 0; n < nfds; ++n)
		2921	ev_io_stop (EV_A_ iow [n]);
		2922
		2923	return got_events;
		2924	}
		2925
		2926
		2927	=head2 C<ev_embed> - when one backend isn't enough...
		2928
		2929	This is a rather advanced watcher type that lets you embed one event loop
		2930	into another (currently only C<ev_io> events are supported in the embedded
		2931	loop, other types of watchers might be handled in a delayed or incorrect
		2932	fashion and must not be used).
		2933
		2934	There are primarily two reasons you would want that: work around bugs and
		2935	prioritise I/O.
		2936
		2937	As an example for a bug workaround, the kqueue backend might only support
		2938	sockets on some platform, so it is unusable as generic backend, but you
		2939	still want to make use of it because you have many sockets and it scales
		2940	so nicely. In this case, you would create a kqueue-based loop and embed
		2941	it into your default loop (which might use e.g. poll). Overall operation
		2942	will be a bit slower because first libev has to call C<poll> and then
		2943	C<kevent>, but at least you can use both mechanisms for what they are
		2944	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2945
		2946	As for prioritising I/O: under rare circumstances you have the case where
		2947	some fds have to be watched and handled very quickly (with low latency),
		2948	and even priorities and idle watchers might have too much overhead. In
		2949	this case you would put all the high priority stuff in one loop and all
		2950	the rest in a second one, and embed the second one in the first.
		2951
		2952	As long as the watcher is active, the callback will be invoked every
		2953	time there might be events pending in the embedded loop. The callback
		2954	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		2955	sweep and invoke their callbacks (the callback doesn't need to invoke the
		2956	C<ev_embed_sweep> function directly, it could also start an idle watcher
		2957	to give the embedded loop strictly lower priority for example).
		2958
		2959	You can also set the callback to C<0>, in which case the embed watcher
		2960	will automatically execute the embedded loop sweep whenever necessary.
		2961
		2962	Fork detection will be handled transparently while the C<ev_embed> watcher
		2963	is active, i.e., the embedded loop will automatically be forked when the
		2964	embedding loop forks. In other cases, the user is responsible for calling
		2965	C<ev_loop_fork> on the embedded loop.
		2966
		2967	Unfortunately, not all backends are embeddable: only the ones returned by
		2968	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2969	portable one.
		2970
		2971	So when you want to use this feature you will always have to be prepared
		2972	that you cannot get an embeddable loop. The recommended way to get around
		2973	this is to have a separate variables for your embeddable loop, try to
		2974	create it, and if that fails, use the normal loop for everything.
		2975
		2976	=head3 C<ev_embed> and fork
		2977
		2978	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2979	automatically be applied to the embedded loop as well, so no special
		2980	fork handling is required in that case. When the watcher is not running,
		2981	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2982	as applicable.
		2983
		2984	=head3 Watcher-Specific Functions and Data Members
		2985
		2986	=over 4
		2987
		2988	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2989
		2990	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2991
		2992	Configures the watcher to embed the given loop, which must be
		2993	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2994	invoked automatically, otherwise it is the responsibility of the callback
		2995	to invoke it (it will continue to be called until the sweep has been done,
		2996	if you do not want that, you need to temporarily stop the embed watcher).
		2997
		2998	=item ev_embed_sweep (loop, ev_embed *)
		2999
		3000	Make a single, non-blocking sweep over the embedded loop. This works
		3001	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
		3002	appropriate way for embedded loops.
		3003
		3004	=item struct ev_loop *other [read-only]
		3005
		3006	The embedded event loop.
		3007
		3008	=back
		3009
		3010	=head3 Examples
		3011
		3012	Example: Try to get an embeddable event loop and embed it into the default
		3013	event loop. If that is not possible, use the default loop. The default
		3014	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3015	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3016	used).
		3017
		3018	struct ev_loop *loop_hi = ev_default_init (0);
		3019	struct ev_loop *loop_lo = 0;
		3020	ev_embed embed;
		3021
		3022	// see if there is a chance of getting one that works
		3023	// (remember that a flags value of 0 means autodetection)
		3024	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3025	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3026	: 0;
		3027
		3028	// if we got one, then embed it, otherwise default to loop_hi
		3029	if (loop_lo)
		3030	{
		3031	ev_embed_init (&embed, 0, loop_lo);
		3032	ev_embed_start (loop_hi, &embed);
		3033	}
		3034	else
		3035	loop_lo = loop_hi;
		3036
		3037	Example: Check if kqueue is available but not recommended and create
		3038	a kqueue backend for use with sockets (which usually work with any
		3039	kqueue implementation). Store the kqueue/socket-only event loop in
		3040	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3041
		3042	struct ev_loop *loop = ev_default_init (0);
		3043	struct ev_loop *loop_socket = 0;
		3044	ev_embed embed;
		3045
		3046	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3047	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3048	{
		3049	ev_embed_init (&embed, 0, loop_socket);
		3050	ev_embed_start (loop, &embed);
		3051	}
		3052
		3053	if (!loop_socket)
		3054	loop_socket = loop;
		3055
		3056	// now use loop_socket for all sockets, and loop for everything else
		3057
		3058
		3059	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		3060
		3061	Fork watchers are called when a C<fork ()> was detected (usually because
		3062	whoever is a good citizen cared to tell libev about it by calling
		3063	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		3064	event loop blocks next and before C<ev_check> watchers are being called,
		3065	and only in the child after the fork. If whoever good citizen calling
		3066	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		3067	handlers will be invoked, too, of course.
		3068
		3069	=head3 The special problem of life after fork - how is it possible?
		3070
		3071	Most uses of C<fork()> consist of forking, then some simple calls to set
		3072	up/change the process environment, followed by a call to C<exec()>. This
		3073	sequence should be handled by libev without any problems.
		3074
		3075	This changes when the application actually wants to do event handling
		3076	in the child, or both parent in child, in effect "continuing" after the
		3077	fork.
		3078
		3079	The default mode of operation (for libev, with application help to detect
		3080	forks) is to duplicate all the state in the child, as would be expected
		3081	when I<either> the parent I<or> the child process continues.
		3082
		3083	When both processes want to continue using libev, then this is usually the
		3084	wrong result. In that case, usually one process (typically the parent) is
		3085	supposed to continue with all watchers in place as before, while the other
		3086	process typically wants to start fresh, i.e. without any active watchers.
		3087
		3088	The cleanest and most efficient way to achieve that with libev is to
		3089	simply create a new event loop, which of course will be "empty", and
		3090	use that for new watchers. This has the advantage of not touching more
		3091	memory than necessary, and thus avoiding the copy-on-write, and the
		3092	disadvantage of having to use multiple event loops (which do not support
		3093	signal watchers).
		3094
		3095	When this is not possible, or you want to use the default loop for
		3096	other reasons, then in the process that wants to start "fresh", call
		3097	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3098	Destroying the default loop will "orphan" (not stop) all registered
		3099	watchers, so you have to be careful not to execute code that modifies
		3100	those watchers. Note also that in that case, you have to re-register any
		3101	signal watchers.
		3102
		3103	=head3 Watcher-Specific Functions and Data Members
		3104
		3105	=over 4
		3106
		3107	=item ev_fork_init (ev_fork *, callback)
		3108
		3109	Initialises and configures the fork watcher - it has no parameters of any
		3110	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3111	really.
		3112
		3113	=back
		3114
		3115
		3116	=head2 C<ev_cleanup> - even the best things end
		3117
		3118	Cleanup watchers are called just before the event loop is being destroyed
		3119	by a call to C<ev_loop_destroy>.
		3120
		3121	While there is no guarantee that the event loop gets destroyed, cleanup
		3122	watchers provide a convenient method to install cleanup hooks for your
		3123	program, worker threads and so on - you just to make sure to destroy the
		3124	loop when you want them to be invoked.
		3125
		3126	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3127	all other watchers, they do not keep a reference to the event loop (which
		3128	makes a lot of sense if you think about it). Like all other watchers, you
		3129	can call libev functions in the callback, except C<ev_cleanup_start>.
		3130
		3131	=head3 Watcher-Specific Functions and Data Members
		3132
		3133	=over 4
		3134
		3135	=item ev_cleanup_init (ev_cleanup *, callback)
		3136
		3137	Initialises and configures the cleanup watcher - it has no parameters of
		3138	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3139	pointless, I assure you.
		3140
		3141	=back
		3142
		3143	Example: Register an atexit handler to destroy the default loop, so any
		3144	cleanup functions are called.
		3145
		3146	static void
		3147	program_exits (void)
		3148	{
		3149	ev_loop_destroy (EV_DEFAULT_UC);
		3150	}
		3151
		3152	...
		3153	atexit (program_exits);
		3154
		3155
		3156	=head2 C<ev_async> - how to wake up an event loop
		3157
		3158	In general, you cannot use an C<ev_run> from multiple threads or other
		3159	asynchronous sources such as signal handlers (as opposed to multiple event
		3160	loops - those are of course safe to use in different threads).
		3161
		3162	Sometimes, however, you need to wake up an event loop you do not control,
		3163	for example because it belongs to another thread. This is what C<ev_async>
		3164	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3165	it by calling C<ev_async_send>, which is thread- and signal safe.
		3166
		3167	This functionality is very similar to C<ev_signal> watchers, as signals,
		3168	too, are asynchronous in nature, and signals, too, will be compressed
		3169	(i.e. the number of callback invocations may be less than the number of
		3170	C<ev_async_sent> calls).
		3171
		3172	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		3173	just the default loop.
		3174
		3175	=head3 Queueing
		3176
		3177	C<ev_async> does not support queueing of data in any way. The reason
		3178	is that the author does not know of a simple (or any) algorithm for a
		3179	multiple-writer-single-reader queue that works in all cases and doesn't
		3180	need elaborate support such as pthreads or unportable memory access
		3181	semantics.
		3182
		3183	That means that if you want to queue data, you have to provide your own
		3184	queue. But at least I can tell you how to implement locking around your
		3185	queue:
		3186
		3187	=over 4
		3188
		3189	=item queueing from a signal handler context
		3190
		3191	To implement race-free queueing, you simply add to the queue in the signal
		3192	handler but you block the signal handler in the watcher callback. Here is
		3193	an example that does that for some fictitious SIGUSR1 handler:
		3194
		3195	static ev_async mysig;
		3196
		3197	static void
		3198	sigusr1_handler (void)
		3199	{
		3200	sometype data;
		3201
		3202	// no locking etc.
		3203	queue_put (data);
		3204	ev_async_send (EV_DEFAULT_ &mysig);
		3205	}
		3206
		3207	static void
		3208	mysig_cb (EV_P_ ev_async *w, int revents)
		3209	{
		3210	sometype data;
		3211	sigset_t block, prev;
		3212
		3213	sigemptyset (&block);
		3214	sigaddset (&block, SIGUSR1);
		3215	sigprocmask (SIG_BLOCK, &block, &prev);
		3216
		3217	while (queue_get (&data))
		3218	process (data);
		3219
		3220	if (sigismember (&prev, SIGUSR1)
		3221	sigprocmask (SIG_UNBLOCK, &block, 0);
		3222	}
		3223
		3224	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3225	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3226	either...).
		3227
		3228	=item queueing from a thread context
		3229
		3230	The strategy for threads is different, as you cannot (easily) block
		3231	threads but you can easily preempt them, so to queue safely you need to
		3232	employ a traditional mutex lock, such as in this pthread example:
		3233
		3234	static ev_async mysig;
		3235	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3236
		3237	static void
		3238	otherthread (void)
		3239	{
		3240	// only need to lock the actual queueing operation
		3241	pthread_mutex_lock (&mymutex);
		3242	queue_put (data);
		3243	pthread_mutex_unlock (&mymutex);
		3244
		3245	ev_async_send (EV_DEFAULT_ &mysig);
		3246	}
		3247
		3248	static void
		3249	mysig_cb (EV_P_ ev_async *w, int revents)
		3250	{
		3251	pthread_mutex_lock (&mymutex);
		3252
		3253	while (queue_get (&data))
		3254	process (data);
		3255
		3256	pthread_mutex_unlock (&mymutex);
		3257	}
		3258
		3259	=back
		3260
		3261
		3262	=head3 Watcher-Specific Functions and Data Members
		3263
		3264	=over 4
		3265
		3266	=item ev_async_init (ev_async *, callback)
		3267
		3268	Initialises and configures the async watcher - it has no parameters of any
		3269	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3270	trust me.
		3271
		3272	=item ev_async_send (loop, ev_async *)
		3273
		3274	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3275	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3276	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3277	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3278	section below on what exactly this means).
		3279
		3280	Note that, as with other watchers in libev, multiple events might get
		3281	compressed into a single callback invocation (another way to look at this
		3282	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3283	reset when the event loop detects that).
		3284
		3285	This call incurs the overhead of a system call only once per event loop
		3286	iteration, so while the overhead might be noticeable, it doesn't apply to
		3287	repeated calls to C<ev_async_send> for the same event loop.
		3288
		3289	=item bool = ev_async_pending (ev_async *)
		3290
		3291	Returns a non-zero value when C<ev_async_send> has been called on the
		3292	watcher but the event has not yet been processed (or even noted) by the
		3293	event loop.
		3294
		3295	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3296	the loop iterates next and checks for the watcher to have become active,
		3297	it will reset the flag again. C<ev_async_pending> can be used to very
		3298	quickly check whether invoking the loop might be a good idea.
		3299
		3300	Not that this does I<not> check whether the watcher itself is pending,
		3301	only whether it has been requested to make this watcher pending: there
		3302	is a time window between the event loop checking and resetting the async
		3303	notification, and the callback being invoked.
		3304
		3305	=back
		3306
		3307
740	=head1 OTHER FUNCTIONS	3308	=head1 OTHER FUNCTIONS
741		3309
742	There are some other functions of possible interest. Described. Here. Now.	3310	There are some other functions of possible interest. Described. Here. Now.
743		3311
744	=over 4	3312	=over 4
745		3313
746	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3314	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
747		3315
748	This function combines a simple timer and an I/O watcher, calls your	3316	This function combines a simple timer and an I/O watcher, calls your
749	callback on whichever event happens first and automatically stop both	3317	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	3318	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	3319	or timeout without having to allocate/configure/start/stop/free one or
752	more watchers yourself.	3320	more watchers yourself.
753		3321
754	If C<fd> is less than 0, then no I/O watcher will be started and events	3322	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	3323	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
756	C<events> set will be craeted and started.	3324	the given C<fd> and C<events> set will be created and started.
757		3325
758	If C<timeout> is less than 0, then no timeout watcher will be	3326	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	3327	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	3328	repeat = 0) will be started. C<0> is a valid timeout.
761	dubious value.
762		3329
763	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3330	The callback has the type C<void (*cb)(int revents, void *arg)> and is
764	passed an C<revents> set like normal event callbacks (a combination of	3331	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>	3332	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
766	value passed to C<ev_once>:	3333	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3334	a timeout and an io event at the same time - you probably should give io
		3335	events precedence.
767		3336
		3337	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3338
768	static void stdin_ready (int revents, void *arg)	3339	static void stdin_ready (int revents, void *arg)
769	{	3340	{
770	if (revents & EV_TIMEOUT)
771	/* doh, nothing entered */;
772	else if (revents & EV_READ)	3341	if (revents & EV_READ)
773	/* stdin might have data for us, joy! */;	3342	/* stdin might have data for us, joy! */;
		3343	else if (revents & EV_TIMER)
		3344	/* doh, nothing entered */;
774	}	3345	}
775		3346
776	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3347	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
777
778	=item ev_feed_event (loop, watcher, int events)
779
780	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
782	initialised but not necessarily started event watcher).
783		3348
784	=item ev_feed_fd_event (loop, int fd, int revents)	3349	=item ev_feed_fd_event (loop, int fd, int revents)
785		3350
786	Feed an event on the given fd, as if a file descriptor backend detected	3351	Feed an event on the given fd, as if a file descriptor backend detected
787	the given events it.	3352	the given events it.
788		3353
789	=item ev_feed_signal_event (loop, int signum)	3354	=item ev_feed_signal_event (loop, int signum)
790		3355
791	Feed an event as if the given signal occured (loop must be the default loop!).	3356	Feed an event as if the given signal occurred (C<loop> must be the default
		3357	loop!).
792		3358
793	=back	3359	=back
		3360
794		3361
795	=head1 LIBEVENT EMULATION	3362	=head1 LIBEVENT EMULATION
796		3363
797	Libev offers a compatibility emulation layer for libevent. It cannot	3364	Libev offers a compatibility emulation layer for libevent. It cannot
798	emulate the internals of libevent, so here are some usage hints:	3365	emulate the internals of libevent, so here are some usage hints:
…		…
810		3377
811	=item * Priorities are not currently supported. Initialising priorities	3378	=item * Priorities are not currently supported. Initialising priorities
812	will fail and all watchers will have the same priority, even though there	3379	will fail and all watchers will have the same priority, even though there
813	is an ev_pri field.	3380	is an ev_pri field.
814		3381
		3382	=item * In libevent, the last base created gets the signals, in libev, the
		3383	first base created (== the default loop) gets the signals.
		3384
815	=item * Other members are not supported.	3385	=item * Other members are not supported.
816		3386
817	=item * The libev emulation is I<not> ABI compatible to libevent, you need	3387	=item * The libev emulation is I<not> ABI compatible to libevent, you need
818	to use the libev header file and library.	3388	to use the libev header file and library.
819		3389
820	=back	3390	=back
821		3391
822	=head1 C++ SUPPORT	3392	=head1 C++ SUPPORT
823		3393
824	TBD.	3394	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		3395	you to use some convenience methods to start/stop watchers and also change
		3396	the callback model to a model using method callbacks on objects.
		3397
		3398	To use it,
		3399
		3400	#include <ev++.h>
		3401
		3402	This automatically includes F<ev.h> and puts all of its definitions (many
		3403	of them macros) into the global namespace. All C++ specific things are
		3404	put into the C<ev> namespace. It should support all the same embedding
		3405	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		3406
		3407	Care has been taken to keep the overhead low. The only data member the C++
		3408	classes add (compared to plain C-style watchers) is the event loop pointer
		3409	that the watcher is associated with (or no additional members at all if
		3410	you disable C<EV_MULTIPLICITY> when embedding libev).
		3411
		3412	Currently, functions, and static and non-static member functions can be
		3413	used as callbacks. Other types should be easy to add as long as they only
		3414	need one additional pointer for context. If you need support for other
		3415	types of functors please contact the author (preferably after implementing
		3416	it).
		3417
		3418	Here is a list of things available in the C<ev> namespace:
		3419
		3420	=over 4
		3421
		3422	=item C<ev::READ>, C<ev::WRITE> etc.
		3423
		3424	These are just enum values with the same values as the C<EV_READ> etc.
		3425	macros from F<ev.h>.
		3426
		3427	=item C<ev::tstamp>, C<ev::now>
		3428
		3429	Aliases to the same types/functions as with the C<ev_> prefix.
		3430
		3431	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		3432
		3433	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		3434	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		3435	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		3436	defines by many implementations.
		3437
		3438	All of those classes have these methods:
		3439
		3440	=over 4
		3441
		3442	=item ev::TYPE::TYPE ()
		3443
		3444	=item ev::TYPE::TYPE (loop)
		3445
		3446	=item ev::TYPE::~TYPE
		3447
		3448	The constructor (optionally) takes an event loop to associate the watcher
		3449	with. If it is omitted, it will use C<EV_DEFAULT>.
		3450
		3451	The constructor calls C<ev_init> for you, which means you have to call the
		3452	C<set> method before starting it.
		3453
		3454	It will not set a callback, however: You have to call the templated C<set>
		3455	method to set a callback before you can start the watcher.
		3456
		3457	(The reason why you have to use a method is a limitation in C++ which does
		3458	not allow explicit template arguments for constructors).
		3459
		3460	The destructor automatically stops the watcher if it is active.
		3461
		3462	=item w->set<class, &class::method> (object *)
		3463
		3464	This method sets the callback method to call. The method has to have a
		3465	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		3466	first argument and the C<revents> as second. The object must be given as
		3467	parameter and is stored in the C<data> member of the watcher.
		3468
		3469	This method synthesizes efficient thunking code to call your method from
		3470	the C callback that libev requires. If your compiler can inline your
		3471	callback (i.e. it is visible to it at the place of the C<set> call and
		3472	your compiler is good :), then the method will be fully inlined into the
		3473	thunking function, making it as fast as a direct C callback.
		3474
		3475	Example: simple class declaration and watcher initialisation
		3476
		3477	struct myclass
		3478	{
		3479	void io_cb (ev::io &w, int revents) { }
		3480	}
		3481
		3482	myclass obj;
		3483	ev::io iow;
		3484	iow.set <myclass, &myclass::io_cb> (&obj);
		3485
		3486	=item w->set (object *)
		3487
		3488	This is a variation of a method callback - leaving out the method to call
		3489	will default the method to C<operator ()>, which makes it possible to use
		3490	functor objects without having to manually specify the C<operator ()> all
		3491	the time. Incidentally, you can then also leave out the template argument
		3492	list.
		3493
		3494	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3495	int revents)>.
		3496
		3497	See the method-C<set> above for more details.
		3498
		3499	Example: use a functor object as callback.
		3500
		3501	struct myfunctor
		3502	{
		3503	void operator() (ev::io &w, int revents)
		3504	{
		3505	...
		3506	}
		3507	}
		3508
		3509	myfunctor f;
		3510
		3511	ev::io w;
		3512	w.set (&f);
		3513
		3514	=item w->set<function> (void *data = 0)
		3515
		3516	Also sets a callback, but uses a static method or plain function as
		3517	callback. The optional C<data> argument will be stored in the watcher's
		3518	C<data> member and is free for you to use.
		3519
		3520	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		3521
		3522	See the method-C<set> above for more details.
		3523
		3524	Example: Use a plain function as callback.
		3525
		3526	static void io_cb (ev::io &w, int revents) { }
		3527	iow.set <io_cb> ();
		3528
		3529	=item w->set (loop)
		3530
		3531	Associates a different C<struct ev_loop> with this watcher. You can only
		3532	do this when the watcher is inactive (and not pending either).
		3533
		3534	=item w->set ([arguments])
		3535
		3536	Basically the same as C<ev_TYPE_set>, with the same arguments. Either this
		3537	method or a suitable start method must be called at least once. Unlike the
		3538	C counterpart, an active watcher gets automatically stopped and restarted
		3539	when reconfiguring it with this method.
		3540
		3541	=item w->start ()
		3542
		3543	Starts the watcher. Note that there is no C<loop> argument, as the
		3544	constructor already stores the event loop.
		3545
		3546	=item w->start ([arguments])
		3547
		3548	Instead of calling C<set> and C<start> methods separately, it is often
		3549	convenient to wrap them in one call. Uses the same type of arguments as
		3550	the configure C<set> method of the watcher.
		3551
		3552	=item w->stop ()
		3553
		3554	Stops the watcher if it is active. Again, no C<loop> argument.
		3555
		3556	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		3557
		3558	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		3559	C<ev_TYPE_again> function.
		3560
		3561	=item w->sweep () (C<ev::embed> only)
		3562
		3563	Invokes C<ev_embed_sweep>.
		3564
		3565	=item w->update () (C<ev::stat> only)
		3566
		3567	Invokes C<ev_stat_stat>.
		3568
		3569	=back
		3570
		3571	=back
		3572
		3573	Example: Define a class with two I/O and idle watchers, start the I/O
		3574	watchers in the constructor.
		3575
		3576	class myclass
		3577	{
		3578	ev::io io ; void io_cb (ev::io &w, int revents);
		3579	ev::io2 io2 ; void io2_cb (ev::io &w, int revents);
		3580	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		3581
		3582	myclass (int fd)
		3583	{
		3584	io .set <myclass, &myclass::io_cb > (this);
		3585	io2 .set <myclass, &myclass::io2_cb > (this);
		3586	idle.set <myclass, &myclass::idle_cb> (this);
		3587
		3588	io.set (fd, ev::WRITE); // configure the watcher
		3589	io.start (); // start it whenever convenient
		3590
		3591	io2.start (fd, ev::READ); // set + start in one call
		3592	}
		3593	};
		3594
		3595
		3596	=head1 OTHER LANGUAGE BINDINGS
		3597
		3598	Libev does not offer other language bindings itself, but bindings for a
		3599	number of languages exist in the form of third-party packages. If you know
		3600	any interesting language binding in addition to the ones listed here, drop
		3601	me a note.
		3602
		3603	=over 4
		3604
		3605	=item Perl
		3606
		3607	The EV module implements the full libev API and is actually used to test
		3608	libev. EV is developed together with libev. Apart from the EV core module,
		3609	there are additional modules that implement libev-compatible interfaces
		3610	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3611	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3612	and C<EV::Glib>).
		3613
		3614	It can be found and installed via CPAN, its homepage is at
		3615	L<http://software.schmorp.de/pkg/EV>.
		3616
		3617	=item Python
		3618
		3619	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3620	seems to be quite complete and well-documented.
		3621
		3622	=item Ruby
		3623
		3624	Tony Arcieri has written a ruby extension that offers access to a subset
		3625	of the libev API and adds file handle abstractions, asynchronous DNS and
		3626	more on top of it. It can be found via gem servers. Its homepage is at
		3627	L<http://rev.rubyforge.org/>.
		3628
		3629	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3630	makes rev work even on mingw.
		3631
		3632	=item Haskell
		3633
		3634	A haskell binding to libev is available at
		3635	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3636
		3637	=item D
		3638
		3639	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3640	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3641
		3642	=item Ocaml
		3643
		3644	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3645	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3646
		3647	=item Lua
		3648
		3649	Brian Maher has written a partial interface to libev for lua (at the
		3650	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		3651	L<http://github.com/brimworks/lua-ev>.
		3652
		3653	=back
		3654
		3655
		3656	=head1 MACRO MAGIC
		3657
		3658	Libev can be compiled with a variety of options, the most fundamental
		3659	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3660	functions and callbacks have an initial C<struct ev_loop *> argument.
		3661
		3662	To make it easier to write programs that cope with either variant, the
		3663	following macros are defined:
		3664
		3665	=over 4
		3666
		3667	=item C<EV_A>, C<EV_A_>
		3668
		3669	This provides the loop I<argument> for functions, if one is required ("ev
		3670	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3671	C<EV_A_> is used when other arguments are following. Example:
		3672
		3673	ev_unref (EV_A);
		3674	ev_timer_add (EV_A_ watcher);
		3675	ev_run (EV_A_ 0);
		3676
		3677	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3678	which is often provided by the following macro.
		3679
		3680	=item C<EV_P>, C<EV_P_>
		3681
		3682	This provides the loop I<parameter> for functions, if one is required ("ev
		3683	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3684	C<EV_P_> is used when other parameters are following. Example:
		3685
		3686	// this is how ev_unref is being declared
		3687	static void ev_unref (EV_P);
		3688
		3689	// this is how you can declare your typical callback
		3690	static void cb (EV_P_ ev_timer *w, int revents)
		3691
		3692	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3693	suitable for use with C<EV_A>.
		3694
		3695	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3696
		3697	Similar to the other two macros, this gives you the value of the default
		3698	loop, if multiple loops are supported ("ev loop default").
		3699
		3700	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3701
		3702	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3703	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3704	is undefined when the default loop has not been initialised by a previous
		3705	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3706
		3707	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3708	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3709
		3710	=back
		3711
		3712	Example: Declare and initialise a check watcher, utilising the above
		3713	macros so it will work regardless of whether multiple loops are supported
		3714	or not.
		3715
		3716	static void
		3717	check_cb (EV_P_ ev_timer *w, int revents)
		3718	{
		3719	ev_check_stop (EV_A_ w);
		3720	}
		3721
		3722	ev_check check;
		3723	ev_check_init (&check, check_cb);
		3724	ev_check_start (EV_DEFAULT_ &check);
		3725	ev_run (EV_DEFAULT_ 0);
		3726
		3727	=head1 EMBEDDING
		3728
		3729	Libev can (and often is) directly embedded into host
		3730	applications. Examples of applications that embed it include the Deliantra
		3731	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3732	and rxvt-unicode.
		3733
		3734	The goal is to enable you to just copy the necessary files into your
		3735	source directory without having to change even a single line in them, so
		3736	you can easily upgrade by simply copying (or having a checked-out copy of
		3737	libev somewhere in your source tree).
		3738
		3739	=head2 FILESETS
		3740
		3741	Depending on what features you need you need to include one or more sets of files
		3742	in your application.
		3743
		3744	=head3 CORE EVENT LOOP
		3745
		3746	To include only the libev core (all the C<ev_*> functions), with manual
		3747	configuration (no autoconf):
		3748
		3749	#define EV_STANDALONE 1
		3750	#include "ev.c"
		3751
		3752	This will automatically include F<ev.h>, too, and should be done in a
		3753	single C source file only to provide the function implementations. To use
		3754	it, do the same for F<ev.h> in all files wishing to use this API (best
		3755	done by writing a wrapper around F<ev.h> that you can include instead and
		3756	where you can put other configuration options):
		3757
		3758	#define EV_STANDALONE 1
		3759	#include "ev.h"
		3760
		3761	Both header files and implementation files can be compiled with a C++
		3762	compiler (at least, that's a stated goal, and breakage will be treated
		3763	as a bug).
		3764
		3765	You need the following files in your source tree, or in a directory
		3766	in your include path (e.g. in libev/ when using -Ilibev):
		3767
		3768	ev.h
		3769	ev.c
		3770	ev_vars.h
		3771	ev_wrap.h
		3772
		3773	ev_win32.c required on win32 platforms only
		3774
		3775	ev_select.c only when select backend is enabled (which is enabled by default)
		3776	ev_poll.c only when poll backend is enabled (disabled by default)
		3777	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3778	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3779	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3780
		3781	F<ev.c> includes the backend files directly when enabled, so you only need
		3782	to compile this single file.
		3783
		3784	=head3 LIBEVENT COMPATIBILITY API
		3785
		3786	To include the libevent compatibility API, also include:
		3787
		3788	#include "event.c"
		3789
		3790	in the file including F<ev.c>, and:
		3791
		3792	#include "event.h"
		3793
		3794	in the files that want to use the libevent API. This also includes F<ev.h>.
		3795
		3796	You need the following additional files for this:
		3797
		3798	event.h
		3799	event.c
		3800
		3801	=head3 AUTOCONF SUPPORT
		3802
		3803	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3804	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3805	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3806	include F<config.h> and configure itself accordingly.
		3807
		3808	For this of course you need the m4 file:
		3809
		3810	libev.m4
		3811
		3812	=head2 PREPROCESSOR SYMBOLS/MACROS
		3813
		3814	Libev can be configured via a variety of preprocessor symbols you have to
		3815	define before including (or compiling) any of its files. The default in
		3816	the absence of autoconf is documented for every option.
		3817
		3818	Symbols marked with "(h)" do not change the ABI, and can have different
		3819	values when compiling libev vs. including F<ev.h>, so it is permissible
		3820	to redefine them before including F<ev.h> without breaking compatibility
		3821	to a compiled library. All other symbols change the ABI, which means all
		3822	users of libev and the libev code itself must be compiled with compatible
		3823	settings.
		3824
		3825	=over 4
		3826
		3827	=item EV_COMPAT3 (h)
		3828
		3829	Backwards compatibility is a major concern for libev. This is why this
		3830	release of libev comes with wrappers for the functions and symbols that
		3831	have been renamed between libev version 3 and 4.
		3832
		3833	You can disable these wrappers (to test compatibility with future
		3834	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		3835	sources. This has the additional advantage that you can drop the C<struct>
		3836	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		3837	typedef in that case.
		3838
		3839	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		3840	and in some even more future version the compatibility code will be
		3841	removed completely.
		3842
		3843	=item EV_STANDALONE (h)
		3844
		3845	Must always be C<1> if you do not use autoconf configuration, which
		3846	keeps libev from including F<config.h>, and it also defines dummy
		3847	implementations for some libevent functions (such as logging, which is not
		3848	supported). It will also not define any of the structs usually found in
		3849	F<event.h> that are not directly supported by the libev core alone.
		3850
		3851	In standalone mode, libev will still try to automatically deduce the
		3852	configuration, but has to be more conservative.
		3853
		3854	=item EV_USE_MONOTONIC
		3855
		3856	If defined to be C<1>, libev will try to detect the availability of the
		3857	monotonic clock option at both compile time and runtime. Otherwise no
		3858	use of the monotonic clock option will be attempted. If you enable this,
		3859	you usually have to link against librt or something similar. Enabling it
		3860	when the functionality isn't available is safe, though, although you have
		3861	to make sure you link against any libraries where the C<clock_gettime>
		3862	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		3863
		3864	=item EV_USE_REALTIME
		3865
		3866	If defined to be C<1>, libev will try to detect the availability of the
		3867	real-time clock option at compile time (and assume its availability
		3868	at runtime if successful). Otherwise no use of the real-time clock
		3869	option will be attempted. This effectively replaces C<gettimeofday>
		3870	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		3871	correctness. See the note about libraries in the description of
		3872	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3873	C<EV_USE_CLOCK_SYSCALL>.
		3874
		3875	=item EV_USE_CLOCK_SYSCALL
		3876
		3877	If defined to be C<1>, libev will try to use a direct syscall instead
		3878	of calling the system-provided C<clock_gettime> function. This option
		3879	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3880	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3881	programs needlessly. Using a direct syscall is slightly slower (in
		3882	theory), because no optimised vdso implementation can be used, but avoids
		3883	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3884	higher, as it simplifies linking (no need for C<-lrt>).
		3885
		3886	=item EV_USE_NANOSLEEP
		3887
		3888	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3889	and will use it for delays. Otherwise it will use C<select ()>.
		3890
		3891	=item EV_USE_EVENTFD
		3892
		3893	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3894	available and will probe for kernel support at runtime. This will improve
		3895	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3896	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3897	2.7 or newer, otherwise disabled.
		3898
		3899	=item EV_USE_SELECT
		3900
		3901	If undefined or defined to be C<1>, libev will compile in support for the
		3902	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3903	other method takes over, select will be it. Otherwise the select backend
		3904	will not be compiled in.
		3905
		3906	=item EV_SELECT_USE_FD_SET
		3907
		3908	If defined to C<1>, then the select backend will use the system C<fd_set>
		3909	structure. This is useful if libev doesn't compile due to a missing
		3910	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		3911	on exotic systems. This usually limits the range of file descriptors to
		3912	some low limit such as 1024 or might have other limitations (winsocket
		3913	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		3914	configures the maximum size of the C<fd_set>.
		3915
		3916	=item EV_SELECT_IS_WINSOCKET
		3917
		3918	When defined to C<1>, the select backend will assume that
		3919	select/socket/connect etc. don't understand file descriptors but
		3920	wants osf handles on win32 (this is the case when the select to
		3921	be used is the winsock select). This means that it will call
		3922	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3923	it is assumed that all these functions actually work on fds, even
		3924	on win32. Should not be defined on non-win32 platforms.
		3925
		3926	=item EV_FD_TO_WIN32_HANDLE(fd)
		3927
		3928	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3929	file descriptors to socket handles. When not defining this symbol (the
		3930	default), then libev will call C<_get_osfhandle>, which is usually
		3931	correct. In some cases, programs use their own file descriptor management,
		3932	in which case they can provide this function to map fds to socket handles.
		3933
		3934	=item EV_WIN32_HANDLE_TO_FD(handle)
		3935
		3936	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		3937	using the standard C<_open_osfhandle> function. For programs implementing
		3938	their own fd to handle mapping, overwriting this function makes it easier
		3939	to do so. This can be done by defining this macro to an appropriate value.
		3940
		3941	=item EV_WIN32_CLOSE_FD(fd)
		3942
		3943	If programs implement their own fd to handle mapping on win32, then this
		3944	macro can be used to override the C<close> function, useful to unregister
		3945	file descriptors again. Note that the replacement function has to close
		3946	the underlying OS handle.
		3947
		3948	=item EV_USE_POLL
		3949
		3950	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3951	backend. Otherwise it will be enabled on non-win32 platforms. It
		3952	takes precedence over select.
		3953
		3954	=item EV_USE_EPOLL
		3955
		3956	If defined to be C<1>, libev will compile in support for the Linux
		3957	C<epoll>(7) backend. Its availability will be detected at runtime,
		3958	otherwise another method will be used as fallback. This is the preferred
		3959	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3960	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3961
		3962	=item EV_USE_KQUEUE
		3963
		3964	If defined to be C<1>, libev will compile in support for the BSD style
		3965	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3966	otherwise another method will be used as fallback. This is the preferred
		3967	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3968	supports some types of fds correctly (the only platform we found that
		3969	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3970	not be used unless explicitly requested. The best way to use it is to find
		3971	out whether kqueue supports your type of fd properly and use an embedded
		3972	kqueue loop.
		3973
		3974	=item EV_USE_PORT
		3975
		3976	If defined to be C<1>, libev will compile in support for the Solaris
		3977	10 port style backend. Its availability will be detected at runtime,
		3978	otherwise another method will be used as fallback. This is the preferred
		3979	backend for Solaris 10 systems.
		3980
		3981	=item EV_USE_DEVPOLL
		3982
		3983	Reserved for future expansion, works like the USE symbols above.
		3984
		3985	=item EV_USE_INOTIFY
		3986
		3987	If defined to be C<1>, libev will compile in support for the Linux inotify
		3988	interface to speed up C<ev_stat> watchers. Its actual availability will
		3989	be detected at runtime. If undefined, it will be enabled if the headers
		3990	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3991
		3992	=item EV_ATOMIC_T
		3993
		3994	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3995	access is atomic with respect to other threads or signal contexts. No such
		3996	type is easily found in the C language, so you can provide your own type
		3997	that you know is safe for your purposes. It is used both for signal handler "locking"
		3998	as well as for signal and thread safety in C<ev_async> watchers.
		3999
		4000	In the absence of this define, libev will use C<sig_atomic_t volatile>
		4001	(from F<signal.h>), which is usually good enough on most platforms.
		4002
		4003	=item EV_H (h)
		4004
		4005	The name of the F<ev.h> header file used to include it. The default if
		4006	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		4007	used to virtually rename the F<ev.h> header file in case of conflicts.
		4008
		4009	=item EV_CONFIG_H (h)
		4010
		4011	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		4012	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		4013	C<EV_H>, above.
		4014
		4015	=item EV_EVENT_H (h)
		4016
		4017	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		4018	of how the F<event.h> header can be found, the default is C<"event.h">.
		4019
		4020	=item EV_PROTOTYPES (h)
		4021
		4022	If defined to be C<0>, then F<ev.h> will not define any function
		4023	prototypes, but still define all the structs and other symbols. This is
		4024	occasionally useful if you want to provide your own wrapper functions
		4025	around libev functions.
		4026
		4027	=item EV_MULTIPLICITY
		4028
		4029	If undefined or defined to C<1>, then all event-loop-specific functions
		4030	will have the C<struct ev_loop *> as first argument, and you can create
		4031	additional independent event loops. Otherwise there will be no support
		4032	for multiple event loops and there is no first event loop pointer
		4033	argument. Instead, all functions act on the single default loop.
		4034
		4035	=item EV_MINPRI
		4036
		4037	=item EV_MAXPRI
		4038
		4039	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		4040	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		4041	provide for more priorities by overriding those symbols (usually defined
		4042	to be C<-2> and C<2>, respectively).
		4043
		4044	When doing priority-based operations, libev usually has to linearly search
		4045	all the priorities, so having many of them (hundreds) uses a lot of space
		4046	and time, so using the defaults of five priorities (-2 .. +2) is usually
		4047	fine.
		4048
		4049	If your embedding application does not need any priorities, defining these
		4050	both to C<0> will save some memory and CPU.
		4051
		4052	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4053	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4054	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		4055
		4056	If undefined or defined to be C<1> (and the platform supports it), then
		4057	the respective watcher type is supported. If defined to be C<0>, then it
		4058	is not. Disabling watcher types mainly saves code size.
		4059
		4060	=item EV_FEATURES
		4061
		4062	If you need to shave off some kilobytes of code at the expense of some
		4063	speed (but with the full API), you can define this symbol to request
		4064	certain subsets of functionality. The default is to enable all features
		4065	that can be enabled on the platform.
		4066
		4067	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4068	with some broad features you want) and then selectively re-enable
		4069	additional parts you want, for example if you want everything minimal,
		4070	but multiple event loop support, async and child watchers and the poll
		4071	backend, use this:
		4072
		4073	#define EV_FEATURES 0
		4074	#define EV_MULTIPLICITY 1
		4075	#define EV_USE_POLL 1
		4076	#define EV_CHILD_ENABLE 1
		4077	#define EV_ASYNC_ENABLE 1
		4078
		4079	The actual value is a bitset, it can be a combination of the following
		4080	values:
		4081
		4082	=over 4
		4083
		4084	=item C<1> - faster/larger code
		4085
		4086	Use larger code to speed up some operations.
		4087
		4088	Currently this is used to override some inlining decisions (enlarging the
		4089	code size by roughly 30% on amd64).
		4090
		4091	When optimising for size, use of compiler flags such as C<-Os> with
		4092	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4093	assertions.
		4094
		4095	=item C<2> - faster/larger data structures
		4096
		4097	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4098	hash table sizes and so on. This will usually further increase code size
		4099	and can additionally have an effect on the size of data structures at
		4100	runtime.
		4101
		4102	=item C<4> - full API configuration
		4103
		4104	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4105	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4106
		4107	=item C<8> - full API
		4108
		4109	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4110	details on which parts of the API are still available without this
		4111	feature, and do not complain if this subset changes over time.
		4112
		4113	=item C<16> - enable all optional watcher types
		4114
		4115	Enables all optional watcher types. If you want to selectively enable
		4116	only some watcher types other than I/O and timers (e.g. prepare,
		4117	embed, async, child...) you can enable them manually by defining
		4118	C<EV_watchertype_ENABLE> to C<1> instead.
		4119
		4120	=item C<32> - enable all backends
		4121
		4122	This enables all backends - without this feature, you need to enable at
		4123	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4124
		4125	=item C<64> - enable OS-specific "helper" APIs
		4126
		4127	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4128	default.
		4129
		4130	=back
		4131
		4132	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4133	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4134	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4135	watchers, timers and monotonic clock support.
		4136
		4137	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4138	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4139	your program might be left out as well - a binary starting a timer and an
		4140	I/O watcher then might come out at only 5Kb.
		4141
		4142	=item EV_AVOID_STDIO
		4143
		4144	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4145	functions (printf, scanf, perror etc.). This will increase the code size
		4146	somewhat, but if your program doesn't otherwise depend on stdio and your
		4147	libc allows it, this avoids linking in the stdio library which is quite
		4148	big.
		4149
		4150	Note that error messages might become less precise when this option is
		4151	enabled.
		4152
		4153	=item EV_NSIG
		4154
		4155	The highest supported signal number, +1 (or, the number of
		4156	signals): Normally, libev tries to deduce the maximum number of signals
		4157	automatically, but sometimes this fails, in which case it can be
		4158	specified. Also, using a lower number than detected (C<32> should be
		4159	good for about any system in existence) can save some memory, as libev
		4160	statically allocates some 12-24 bytes per signal number.
		4161
		4162	=item EV_PID_HASHSIZE
		4163
		4164	C<ev_child> watchers use a small hash table to distribute workload by
		4165	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		4166	usually more than enough. If you need to manage thousands of children you
		4167	might want to increase this value (I<must> be a power of two).
		4168
		4169	=item EV_INOTIFY_HASHSIZE
		4170
		4171	C<ev_stat> watchers use a small hash table to distribute workload by
		4172	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		4173	disabled), usually more than enough. If you need to manage thousands of
		4174	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4175	power of two).
		4176
		4177	=item EV_USE_4HEAP
		4178
		4179	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4180	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4181	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4182	faster performance with many (thousands) of watchers.
		4183
		4184	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4185	will be C<0>.
		4186
		4187	=item EV_HEAP_CACHE_AT
		4188
		4189	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4190	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4191	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4192	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4193	but avoids random read accesses on heap changes. This improves performance
		4194	noticeably with many (hundreds) of watchers.
		4195
		4196	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4197	will be C<0>.
		4198
		4199	=item EV_VERIFY
		4200
		4201	Controls how much internal verification (see C<ev_verify ()>) will
		4202	be done: If set to C<0>, no internal verification code will be compiled
		4203	in. If set to C<1>, then verification code will be compiled in, but not
		4204	called. If set to C<2>, then the internal verification code will be
		4205	called once per loop, which can slow down libev. If set to C<3>, then the
		4206	verification code will be called very frequently, which will slow down
		4207	libev considerably.
		4208
		4209	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4210	will be C<0>.
		4211
		4212	=item EV_COMMON
		4213
		4214	By default, all watchers have a C<void *data> member. By redefining
		4215	this macro to something else you can include more and other types of
		4216	members. You have to define it each time you include one of the files,
		4217	though, and it must be identical each time.
		4218
		4219	For example, the perl EV module uses something like this:
		4220
		4221	#define EV_COMMON \
		4222	SV *self; /* contains this struct */ \
		4223	SV *cb_sv, *fh /* note no trailing ";" */
		4224
		4225	=item EV_CB_DECLARE (type)
		4226
		4227	=item EV_CB_INVOKE (watcher, revents)
		4228
		4229	=item ev_set_cb (ev, cb)
		4230
		4231	Can be used to change the callback member declaration in each watcher,
		4232	and the way callbacks are invoked and set. Must expand to a struct member
		4233	definition and a statement, respectively. See the F<ev.h> header file for
		4234	their default definitions. One possible use for overriding these is to
		4235	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		4236	method calls instead of plain function calls in C++.
		4237
		4238	=back
		4239
		4240	=head2 EXPORTED API SYMBOLS
		4241
		4242	If you need to re-export the API (e.g. via a DLL) and you need a list of
		4243	exported symbols, you can use the provided F<Symbol.*> files which list
		4244	all public symbols, one per line:
		4245
		4246	Symbols.ev for libev proper
		4247	Symbols.event for the libevent emulation
		4248
		4249	This can also be used to rename all public symbols to avoid clashes with
		4250	multiple versions of libev linked together (which is obviously bad in
		4251	itself, but sometimes it is inconvenient to avoid this).
		4252
		4253	A sed command like this will create wrapper C<#define>'s that you need to
		4254	include before including F<ev.h>:
		4255
		4256	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		4257
		4258	This would create a file F<wrap.h> which essentially looks like this:
		4259
		4260	#define ev_backend myprefix_ev_backend
		4261	#define ev_check_start myprefix_ev_check_start
		4262	#define ev_check_stop myprefix_ev_check_stop
		4263	...
		4264
		4265	=head2 EXAMPLES
		4266
		4267	For a real-world example of a program the includes libev
		4268	verbatim, you can have a look at the EV perl module
		4269	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		4270	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		4271	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		4272	will be compiled. It is pretty complex because it provides its own header
		4273	file.
		4274
		4275	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		4276	that everybody includes and which overrides some configure choices:
		4277
		4278	#define EV_FEATURES 8
		4279	#define EV_USE_SELECT 1
		4280	#define EV_PREPARE_ENABLE 1
		4281	#define EV_IDLE_ENABLE 1
		4282	#define EV_SIGNAL_ENABLE 1
		4283	#define EV_CHILD_ENABLE 1
		4284	#define EV_USE_STDEXCEPT 0
		4285	#define EV_CONFIG_H <config.h>
		4286
		4287	#include "ev++.h"
		4288
		4289	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		4290
		4291	#include "ev_cpp.h"
		4292	#include "ev.c"
		4293
		4294	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		4295
		4296	=head2 THREADS AND COROUTINES
		4297
		4298	=head3 THREADS
		4299
		4300	All libev functions are reentrant and thread-safe unless explicitly
		4301	documented otherwise, but libev implements no locking itself. This means
		4302	that you can use as many loops as you want in parallel, as long as there
		4303	are no concurrent calls into any libev function with the same loop
		4304	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		4305	of course): libev guarantees that different event loops share no data
		4306	structures that need any locking.
		4307
		4308	Or to put it differently: calls with different loop parameters can be done
		4309	concurrently from multiple threads, calls with the same loop parameter
		4310	must be done serially (but can be done from different threads, as long as
		4311	only one thread ever is inside a call at any point in time, e.g. by using
		4312	a mutex per loop).
		4313
		4314	Specifically to support threads (and signal handlers), libev implements
		4315	so-called C<ev_async> watchers, which allow some limited form of
		4316	concurrency on the same event loop, namely waking it up "from the
		4317	outside".
		4318
		4319	If you want to know which design (one loop, locking, or multiple loops
		4320	without or something else still) is best for your problem, then I cannot
		4321	help you, but here is some generic advice:
		4322
		4323	=over 4
		4324
		4325	=item * most applications have a main thread: use the default libev loop
		4326	in that thread, or create a separate thread running only the default loop.
		4327
		4328	This helps integrating other libraries or software modules that use libev
		4329	themselves and don't care/know about threading.
		4330
		4331	=item * one loop per thread is usually a good model.
		4332
		4333	Doing this is almost never wrong, sometimes a better-performance model
		4334	exists, but it is always a good start.
		4335
		4336	=item * other models exist, such as the leader/follower pattern, where one
		4337	loop is handed through multiple threads in a kind of round-robin fashion.
		4338
		4339	Choosing a model is hard - look around, learn, know that usually you can do
		4340	better than you currently do :-)
		4341
		4342	=item * often you need to talk to some other thread which blocks in the
		4343	event loop.
		4344
		4345	C<ev_async> watchers can be used to wake them up from other threads safely
		4346	(or from signal contexts...).
		4347
		4348	An example use would be to communicate signals or other events that only
		4349	work in the default loop by registering the signal watcher with the
		4350	default loop and triggering an C<ev_async> watcher from the default loop
		4351	watcher callback into the event loop interested in the signal.
		4352
		4353	=back
		4354
		4355	=head4 THREAD LOCKING EXAMPLE
		4356
		4357	Here is a fictitious example of how to run an event loop in a different
		4358	thread than where callbacks are being invoked and watchers are
		4359	created/added/removed.
		4360
		4361	For a real-world example, see the C<EV::Loop::Async> perl module,
		4362	which uses exactly this technique (which is suited for many high-level
		4363	languages).
		4364
		4365	The example uses a pthread mutex to protect the loop data, a condition
		4366	variable to wait for callback invocations, an async watcher to notify the
		4367	event loop thread and an unspecified mechanism to wake up the main thread.
		4368
		4369	First, you need to associate some data with the event loop:
		4370
		4371	typedef struct {
		4372	mutex_t lock; /* global loop lock */
		4373	ev_async async_w;
		4374	thread_t tid;
		4375	cond_t invoke_cv;
		4376	} userdata;
		4377
		4378	void prepare_loop (EV_P)
		4379	{
		4380	// for simplicity, we use a static userdata struct.
		4381	static userdata u;
		4382
		4383	ev_async_init (&u->async_w, async_cb);
		4384	ev_async_start (EV_A_ &u->async_w);
		4385
		4386	pthread_mutex_init (&u->lock, 0);
		4387	pthread_cond_init (&u->invoke_cv, 0);
		4388
		4389	// now associate this with the loop
		4390	ev_set_userdata (EV_A_ u);
		4391	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4392	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4393
		4394	// then create the thread running ev_loop
		4395	pthread_create (&u->tid, 0, l_run, EV_A);
		4396	}
		4397
		4398	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4399	solely to wake up the event loop so it takes notice of any new watchers
		4400	that might have been added:
		4401
		4402	static void
		4403	async_cb (EV_P_ ev_async *w, int revents)
		4404	{
		4405	// just used for the side effects
		4406	}
		4407
		4408	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4409	protecting the loop data, respectively.
		4410
		4411	static void
		4412	l_release (EV_P)
		4413	{
		4414	userdata *u = ev_userdata (EV_A);
		4415	pthread_mutex_unlock (&u->lock);
		4416	}
		4417
		4418	static void
		4419	l_acquire (EV_P)
		4420	{
		4421	userdata *u = ev_userdata (EV_A);
		4422	pthread_mutex_lock (&u->lock);
		4423	}
		4424
		4425	The event loop thread first acquires the mutex, and then jumps straight
		4426	into C<ev_run>:
		4427
		4428	void *
		4429	l_run (void *thr_arg)
		4430	{
		4431	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4432
		4433	l_acquire (EV_A);
		4434	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4435	ev_run (EV_A_ 0);
		4436	l_release (EV_A);
		4437
		4438	return 0;
		4439	}
		4440
		4441	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4442	signal the main thread via some unspecified mechanism (signals? pipe
		4443	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4444	have been called (in a while loop because a) spurious wakeups are possible
		4445	and b) skipping inter-thread-communication when there are no pending
		4446	watchers is very beneficial):
		4447
		4448	static void
		4449	l_invoke (EV_P)
		4450	{
		4451	userdata *u = ev_userdata (EV_A);
		4452
		4453	while (ev_pending_count (EV_A))
		4454	{
		4455	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4456	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4457	}
		4458	}
		4459
		4460	Now, whenever the main thread gets told to invoke pending watchers, it
		4461	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4462	thread to continue:
		4463
		4464	static void
		4465	real_invoke_pending (EV_P)
		4466	{
		4467	userdata *u = ev_userdata (EV_A);
		4468
		4469	pthread_mutex_lock (&u->lock);
		4470	ev_invoke_pending (EV_A);
		4471	pthread_cond_signal (&u->invoke_cv);
		4472	pthread_mutex_unlock (&u->lock);
		4473	}
		4474
		4475	Whenever you want to start/stop a watcher or do other modifications to an
		4476	event loop, you will now have to lock:
		4477
		4478	ev_timer timeout_watcher;
		4479	userdata *u = ev_userdata (EV_A);
		4480
		4481	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4482
		4483	pthread_mutex_lock (&u->lock);
		4484	ev_timer_start (EV_A_ &timeout_watcher);
		4485	ev_async_send (EV_A_ &u->async_w);
		4486	pthread_mutex_unlock (&u->lock);
		4487
		4488	Note that sending the C<ev_async> watcher is required because otherwise
		4489	an event loop currently blocking in the kernel will have no knowledge
		4490	about the newly added timer. By waking up the loop it will pick up any new
		4491	watchers in the next event loop iteration.
		4492
		4493	=head3 COROUTINES
		4494
		4495	Libev is very accommodating to coroutines ("cooperative threads"):
		4496	libev fully supports nesting calls to its functions from different
		4497	coroutines (e.g. you can call C<ev_run> on the same loop from two
		4498	different coroutines, and switch freely between both coroutines running
		4499	the loop, as long as you don't confuse yourself). The only exception is
		4500	that you must not do this from C<ev_periodic> reschedule callbacks.
		4501
		4502	Care has been taken to ensure that libev does not keep local state inside
		4503	C<ev_run>, and other calls do not usually allow for coroutine switches as
		4504	they do not call any callbacks.
		4505
		4506	=head2 COMPILER WARNINGS
		4507
		4508	Depending on your compiler and compiler settings, you might get no or a
		4509	lot of warnings when compiling libev code. Some people are apparently
		4510	scared by this.
		4511
		4512	However, these are unavoidable for many reasons. For one, each compiler
		4513	has different warnings, and each user has different tastes regarding
		4514	warning options. "Warn-free" code therefore cannot be a goal except when
		4515	targeting a specific compiler and compiler-version.
		4516
		4517	Another reason is that some compiler warnings require elaborate
		4518	workarounds, or other changes to the code that make it less clear and less
		4519	maintainable.
		4520
		4521	And of course, some compiler warnings are just plain stupid, or simply
		4522	wrong (because they don't actually warn about the condition their message
		4523	seems to warn about). For example, certain older gcc versions had some
		4524	warnings that resulted in an extreme number of false positives. These have
		4525	been fixed, but some people still insist on making code warn-free with
		4526	such buggy versions.
		4527
		4528	While libev is written to generate as few warnings as possible,
		4529	"warn-free" code is not a goal, and it is recommended not to build libev
		4530	with any compiler warnings enabled unless you are prepared to cope with
		4531	them (e.g. by ignoring them). Remember that warnings are just that:
		4532	warnings, not errors, or proof of bugs.
		4533
		4534
		4535	=head2 VALGRIND
		4536
		4537	Valgrind has a special section here because it is a popular tool that is
		4538	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4539
		4540	If you think you found a bug (memory leak, uninitialised data access etc.)
		4541	in libev, then check twice: If valgrind reports something like:
		4542
		4543	==2274== definitely lost: 0 bytes in 0 blocks.
		4544	==2274== possibly lost: 0 bytes in 0 blocks.
		4545	==2274== still reachable: 256 bytes in 1 blocks.
		4546
		4547	Then there is no memory leak, just as memory accounted to global variables
		4548	is not a memleak - the memory is still being referenced, and didn't leak.
		4549
		4550	Similarly, under some circumstances, valgrind might report kernel bugs
		4551	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4552	although an acceptable workaround has been found here), or it might be
		4553	confused.
		4554
		4555	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4556	make it into some kind of religion.
		4557
		4558	If you are unsure about something, feel free to contact the mailing list
		4559	with the full valgrind report and an explanation on why you think this
		4560	is a bug in libev (best check the archives, too :). However, don't be
		4561	annoyed when you get a brisk "this is no bug" answer and take the chance
		4562	of learning how to interpret valgrind properly.
		4563
		4564	If you need, for some reason, empty reports from valgrind for your project
		4565	I suggest using suppression lists.
		4566
		4567
		4568	=head1 PORTABILITY NOTES
		4569
		4570	=head2 GNU/LINUX 32 BIT LIMITATIONS
		4571
		4572	GNU/Linux is the only common platform that supports 64 bit file/large file
		4573	interfaces but I<disables> them by default.
		4574
		4575	That means that libev compiled in the default environment doesn't support
		4576	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		4577
		4578	Unfortunately, many programs try to work around this GNU/Linux issue
		4579	by enabling the large file API, which makes them incompatible with the
		4580	standard libev compiled for their system.
		4581
		4582	Likewise, libev cannot enable the large file API itself as this would
		4583	suddenly make it incompatible to the default compile time environment,
		4584	i.e. all programs not using special compile switches.
		4585
		4586	=head2 OS/X AND DARWIN BUGS
		4587
		4588	The whole thing is a bug if you ask me - basically any system interface
		4589	you touch is broken, whether it is locales, poll, kqueue or even the
		4590	OpenGL drivers.
		4591
		4592	=head3 C<kqueue> is buggy
		4593
		4594	The kqueue syscall is broken in all known versions - most versions support
		4595	only sockets, many support pipes.
		4596
		4597	Libev tries to work around this by not using C<kqueue> by default on this
		4598	rotten platform, but of course you can still ask for it when creating a
		4599	loop - embedding a socket-only kqueue loop into a select-based one is
		4600	probably going to work well.
		4601
		4602	=head3 C<poll> is buggy
		4603
		4604	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		4605	implementation by something calling C<kqueue> internally around the 10.5.6
		4606	release, so now C<kqueue> I<and> C<poll> are broken.
		4607
		4608	Libev tries to work around this by not using C<poll> by default on
		4609	this rotten platform, but of course you can still ask for it when creating
		4610	a loop.
		4611
		4612	=head3 C<select> is buggy
		4613
		4614	All that's left is C<select>, and of course Apple found a way to fuck this
		4615	one up as well: On OS/X, C<select> actively limits the number of file
		4616	descriptors you can pass in to 1024 - your program suddenly crashes when
		4617	you use more.
		4618
		4619	There is an undocumented "workaround" for this - defining
		4620	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		4621	work on OS/X.
		4622
		4623	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		4624
		4625	=head3 C<errno> reentrancy
		4626
		4627	The default compile environment on Solaris is unfortunately so
		4628	thread-unsafe that you can't even use components/libraries compiled
		4629	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		4630	defined by default. A valid, if stupid, implementation choice.
		4631
		4632	If you want to use libev in threaded environments you have to make sure
		4633	it's compiled with C<_REENTRANT> defined.
		4634
		4635	=head3 Event port backend
		4636
		4637	The scalable event interface for Solaris is called "event
		4638	ports". Unfortunately, this mechanism is very buggy in all major
		4639	releases. If you run into high CPU usage, your program freezes or you get
		4640	a large number of spurious wakeups, make sure you have all the relevant
		4641	and latest kernel patches applied. No, I don't know which ones, but there
		4642	are multiple ones to apply, and afterwards, event ports actually work
		4643	great.
		4644
		4645	If you can't get it to work, you can try running the program by setting
		4646	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		4647	C<select> backends.
		4648
		4649	=head2 AIX POLL BUG
		4650
		4651	AIX unfortunately has a broken C<poll.h> header. Libev works around
		4652	this by trying to avoid the poll backend altogether (i.e. it's not even
		4653	compiled in), which normally isn't a big problem as C<select> works fine
		4654	with large bitsets on AIX, and AIX is dead anyway.
		4655
		4656	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4657
		4658	=head3 General issues
		4659
		4660	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4661	requires, and its I/O model is fundamentally incompatible with the POSIX
		4662	model. Libev still offers limited functionality on this platform in
		4663	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4664	descriptors. This only applies when using Win32 natively, not when using
		4665	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		4666	as every compielr comes with a slightly differently broken/incompatible
		4667	environment.
		4668
		4669	Lifting these limitations would basically require the full
		4670	re-implementation of the I/O system. If you are into this kind of thing,
		4671	then note that glib does exactly that for you in a very portable way (note
		4672	also that glib is the slowest event library known to man).
		4673
		4674	There is no supported compilation method available on windows except
		4675	embedding it into other applications.
		4676
		4677	Sensible signal handling is officially unsupported by Microsoft - libev
		4678	tries its best, but under most conditions, signals will simply not work.
		4679
		4680	Not a libev limitation but worth mentioning: windows apparently doesn't
		4681	accept large writes: instead of resulting in a partial write, windows will
		4682	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4683	so make sure you only write small amounts into your sockets (less than a
		4684	megabyte seems safe, but this apparently depends on the amount of memory
		4685	available).
		4686
		4687	Due to the many, low, and arbitrary limits on the win32 platform and
		4688	the abysmal performance of winsockets, using a large number of sockets
		4689	is not recommended (and not reasonable). If your program needs to use
		4690	more than a hundred or so sockets, then likely it needs to use a totally
		4691	different implementation for windows, as libev offers the POSIX readiness
		4692	notification model, which cannot be implemented efficiently on windows
		4693	(due to Microsoft monopoly games).
		4694
		4695	A typical way to use libev under windows is to embed it (see the embedding
		4696	section for details) and use the following F<evwrap.h> header file instead
		4697	of F<ev.h>:
		4698
		4699	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4700	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4701
		4702	#include "ev.h"
		4703
		4704	And compile the following F<evwrap.c> file into your project (make sure
		4705	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4706
		4707	#include "evwrap.h"
		4708	#include "ev.c"
		4709
		4710	=head3 The winsocket C<select> function
		4711
		4712	The winsocket C<select> function doesn't follow POSIX in that it
		4713	requires socket I<handles> and not socket I<file descriptors> (it is
		4714	also extremely buggy). This makes select very inefficient, and also
		4715	requires a mapping from file descriptors to socket handles (the Microsoft
		4716	C runtime provides the function C<_open_osfhandle> for this). See the
		4717	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4718	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4719
		4720	The configuration for a "naked" win32 using the Microsoft runtime
		4721	libraries and raw winsocket select is:
		4722
		4723	#define EV_USE_SELECT 1
		4724	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4725
		4726	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4727	complexity in the O(n²) range when using win32.
		4728
		4729	=head3 Limited number of file descriptors
		4730
		4731	Windows has numerous arbitrary (and low) limits on things.
		4732
		4733	Early versions of winsocket's select only supported waiting for a maximum
		4734	of C<64> handles (probably owning to the fact that all windows kernels
		4735	can only wait for C<64> things at the same time internally; Microsoft
		4736	recommends spawning a chain of threads and wait for 63 handles and the
		4737	previous thread in each. Sounds great!).
		4738
		4739	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4740	to some high number (e.g. C<2048>) before compiling the winsocket select
		4741	call (which might be in libev or elsewhere, for example, perl and many
		4742	other interpreters do their own select emulation on windows).
		4743
		4744	Another limit is the number of file descriptors in the Microsoft runtime
		4745	libraries, which by default is C<64> (there must be a hidden I<64>
		4746	fetish or something like this inside Microsoft). You can increase this
		4747	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4748	(another arbitrary limit), but is broken in many versions of the Microsoft
		4749	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4750	(depending on windows version and/or the phase of the moon). To get more,
		4751	you need to wrap all I/O functions and provide your own fd management, but
		4752	the cost of calling select (O(n²)) will likely make this unworkable.
		4753
		4754	=head2 PORTABILITY REQUIREMENTS
		4755
		4756	In addition to a working ISO-C implementation and of course the
		4757	backend-specific APIs, libev relies on a few additional extensions:
		4758
		4759	=over 4
		4760
		4761	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4762	calling conventions regardless of C<ev_watcher_type *>.
		4763
		4764	Libev assumes not only that all watcher pointers have the same internal
		4765	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4766	assumes that the same (machine) code can be used to call any watcher
		4767	callback: The watcher callbacks have different type signatures, but libev
		4768	calls them using an C<ev_watcher *> internally.
		4769
		4770	=item pointer accesses must be thread-atomic
		4771
		4772	Accessing a pointer value must be atomic, it must both be readable and
		4773	writable in one piece - this is the case on all current architectures.
		4774
		4775	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4776
		4777	The type C<sig_atomic_t volatile> (or whatever is defined as
		4778	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4779	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4780	believed to be sufficiently portable.
		4781
		4782	=item C<sigprocmask> must work in a threaded environment
		4783
		4784	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4785	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4786	pthread implementations will either allow C<sigprocmask> in the "main
		4787	thread" or will block signals process-wide, both behaviours would
		4788	be compatible with libev. Interaction between C<sigprocmask> and
		4789	C<pthread_sigmask> could complicate things, however.
		4790
		4791	The most portable way to handle signals is to block signals in all threads
		4792	except the initial one, and run the default loop in the initial thread as
		4793	well.
		4794
		4795	=item C<long> must be large enough for common memory allocation sizes
		4796
		4797	To improve portability and simplify its API, libev uses C<long> internally
		4798	instead of C<size_t> when allocating its data structures. On non-POSIX
		4799	systems (Microsoft...) this might be unexpectedly low, but is still at
		4800	least 31 bits everywhere, which is enough for hundreds of millions of
		4801	watchers.
		4802
		4803	=item C<double> must hold a time value in seconds with enough accuracy
		4804
		4805	The type C<double> is used to represent timestamps. It is required to
		4806	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		4807	good enough for at least into the year 4000 with millisecond accuracy
		4808	(the design goal for libev). This requirement is overfulfilled by
		4809	implementations using IEEE 754, which is basically all existing ones. With
		4810	IEEE 754 doubles, you get microsecond accuracy until at least 2200.
		4811
		4812	=back
		4813
		4814	If you know of other additional requirements drop me a note.
		4815
		4816
		4817	=head1 ALGORITHMIC COMPLEXITIES
		4818
		4819	In this section the complexities of (many of) the algorithms used inside
		4820	libev will be documented. For complexity discussions about backends see
		4821	the documentation for C<ev_default_init>.
		4822
		4823	All of the following are about amortised time: If an array needs to be
		4824	extended, libev needs to realloc and move the whole array, but this
		4825	happens asymptotically rarer with higher number of elements, so O(1) might
		4826	mean that libev does a lengthy realloc operation in rare cases, but on
		4827	average it is much faster and asymptotically approaches constant time.
		4828
		4829	=over 4
		4830
		4831	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		4832
		4833	This means that, when you have a watcher that triggers in one hour and
		4834	there are 100 watchers that would trigger before that, then inserting will
		4835	have to skip roughly seven (C<ld 100>) of these watchers.
		4836
		4837	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		4838
		4839	That means that changing a timer costs less than removing/adding them,
		4840	as only the relative motion in the event queue has to be paid for.
		4841
		4842	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		4843
		4844	These just add the watcher into an array or at the head of a list.
		4845
		4846	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		4847
		4848	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		4849
		4850	These watchers are stored in lists, so they need to be walked to find the
		4851	correct watcher to remove. The lists are usually short (you don't usually
		4852	have many watchers waiting for the same fd or signal: one is typical, two
		4853	is rare).
		4854
		4855	=item Finding the next timer in each loop iteration: O(1)
		4856
		4857	By virtue of using a binary or 4-heap, the next timer is always found at a
		4858	fixed position in the storage array.
		4859
		4860	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		4861
		4862	A change means an I/O watcher gets started or stopped, which requires
		4863	libev to recalculate its status (and possibly tell the kernel, depending
		4864	on backend and whether C<ev_io_set> was used).
		4865
		4866	=item Activating one watcher (putting it into the pending state): O(1)
		4867
		4868	=item Priority handling: O(number_of_priorities)
		4869
		4870	Priorities are implemented by allocating some space for each
		4871	priority. When doing priority-based operations, libev usually has to
		4872	linearly search all the priorities, but starting/stopping and activating
		4873	watchers becomes O(1) with respect to priority handling.
		4874
		4875	=item Sending an ev_async: O(1)
		4876
		4877	=item Processing ev_async_send: O(number_of_async_watchers)
		4878
		4879	=item Processing signals: O(max_signal_number)
		4880
		4881	Sending involves a system call I<iff> there were no other C<ev_async_send>
		4882	calls in the current loop iteration. Checking for async and signal events
		4883	involves iterating over all running async watchers or all signal numbers.
		4884
		4885	=back
		4886
		4887
		4888	=head1 PORTING FROM LIBEV 3.X TO 4.X
		4889
		4890	The major version 4 introduced some incompatible changes to the API.
		4891
		4892	At the moment, the C<ev.h> header file provides compatibility definitions
		4893	for all changes, so most programs should still compile. The compatibility
		4894	layer might be removed in later versions of libev, so better update to the
		4895	new API early than late.
		4896
		4897	=over 4
		4898
		4899	=item C<EV_COMPAT3> backwards compatibility mechanism
		4900
		4901	The backward compatibility mechanism can be controlled by
		4902	C<EV_COMPAT3>. See L<PREPROCESSOR SYMBOLS/MACROS> in the L<EMBEDDING>
		4903	section.
		4904
		4905	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		4906
		4907	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		4908
		4909	ev_loop_destroy (EV_DEFAULT_UC);
		4910	ev_loop_fork (EV_DEFAULT);
		4911
		4912	=item function/symbol renames
		4913
		4914	A number of functions and symbols have been renamed:
		4915
		4916	ev_loop => ev_run
		4917	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		4918	EVLOOP_ONESHOT => EVRUN_ONCE
		4919
		4920	ev_unloop => ev_break
		4921	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		4922	EVUNLOOP_ONE => EVBREAK_ONE
		4923	EVUNLOOP_ALL => EVBREAK_ALL
		4924
		4925	EV_TIMEOUT => EV_TIMER
		4926
		4927	ev_loop_count => ev_iteration
		4928	ev_loop_depth => ev_depth
		4929	ev_loop_verify => ev_verify
		4930
		4931	Most functions working on C<struct ev_loop> objects don't have an
		4932	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		4933	associated constants have been renamed to not collide with the C<struct
		4934	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		4935	as all other watcher types. Note that C<ev_loop_fork> is still called
		4936	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		4937	typedef.
		4938
		4939	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		4940
		4941	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		4942	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		4943	and work, but the library code will of course be larger.
		4944
		4945	=back
		4946
		4947
		4948	=head1 GLOSSARY
		4949
		4950	=over 4
		4951
		4952	=item active
		4953
		4954	A watcher is active as long as it has been started and not yet stopped.
		4955	See L<WATCHER STATES> for details.
		4956
		4957	=item application
		4958
		4959	In this document, an application is whatever is using libev.
		4960
		4961	=item backend
		4962
		4963	The part of the code dealing with the operating system interfaces.
		4964
		4965	=item callback
		4966
		4967	The address of a function that is called when some event has been
		4968	detected. Callbacks are being passed the event loop, the watcher that
		4969	received the event, and the actual event bitset.
		4970
		4971	=item callback/watcher invocation
		4972
		4973	The act of calling the callback associated with a watcher.
		4974
		4975	=item event
		4976
		4977	A change of state of some external event, such as data now being available
		4978	for reading on a file descriptor, time having passed or simply not having
		4979	any other events happening anymore.
		4980
		4981	In libev, events are represented as single bits (such as C<EV_READ> or
		4982	C<EV_TIMER>).
		4983
		4984	=item event library
		4985
		4986	A software package implementing an event model and loop.
		4987
		4988	=item event loop
		4989
		4990	An entity that handles and processes external events and converts them
		4991	into callback invocations.
		4992
		4993	=item event model
		4994
		4995	The model used to describe how an event loop handles and processes
		4996	watchers and events.
		4997
		4998	=item pending
		4999
		5000	A watcher is pending as soon as the corresponding event has been
		5001	detected. See L<WATCHER STATES> for details.
		5002
		5003	=item real time
		5004
		5005	The physical time that is observed. It is apparently strictly monotonic :)
		5006
		5007	=item wall-clock time
		5008
		5009	The time and date as shown on clocks. Unlike real time, it can actually
		5010	be wrong and jump forwards and backwards, e.g. when the you adjust your
		5011	clock.
		5012
		5013	=item watcher
		5014
		5015	A data structure that describes interest in certain events. Watchers need
		5016	to be started (attached to an event loop) before they can receive events.
		5017
		5018	=back
825		5019
826	=head1 AUTHOR	5020	=head1 AUTHOR
827		5021
828	Marc Lehmann <libev@schmorp.de>.	5022	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5023	Magnusson and Emanuele Giaquinta.
829		5024

Diff Legend

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