ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.25 by root, Mon Nov 12 21:51:14 2007 UTC vs.
Revision 1.305 by root, Thu Oct 14 05:07:04 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_loop's to stop iterating
		32	ev_unloop (EV_A_ EVUNLOOP_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_loop to stop iterating
		41	ev_unloop (EV_A_ EVUNLOOP_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_loop (0);
		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_loop (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 ABOUT LIBEV
10		84
11	Libev is an event loop: you register interest in certain events (such as a	85	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	86	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	87	these event sources and provide your program with events.
14		88
15	To do this, it must take more or less complete control over your process	89	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	90	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	91	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	93	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	94	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	95	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	96	watcher.
23		97
24	=head1 FEATURES	98	=head2 FEATURES
25		99
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	100	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	101	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	102	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
		103	(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		104	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		105	timers (C<ev_timer>), absolute timers with customised rescheduling
		106	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	107	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	108	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		109	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		110	limited support for fork events (C<ev_fork>).
		111
		112	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	113	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	114	for example).
33		115
34	=head1 CONVENTIONS	116	=head2 CONVENTIONS
35		117
36	Libev is very configurable. In this manual the default configuration	118	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	119	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	120	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	121	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	122	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 *>)	123	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	124	this argument.
43		125
44	=head1 TIME REPRESENTATION	126	=head2 TIME REPRESENTATION
45		127
46	Libev represents time as a single floating point number, representing the	128	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	129	the (fractional) number of seconds since the (POSIX) epoch (in practise
48	the beginning of 1970, details are complicated, don't ask). This type is	130	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	131	ask). This type is called C<ev_tstamp>, which is what you should use
50	to the double type in C.	132	too. It usually aliases to the C<double> type in C. When you need to do
		133	any calculations on it, you should treat it as some floating point value.
		134
		135	Unlike the name component C<stamp> might indicate, it is also used for
		136	time differences (e.g. delays) throughout libev.
		137
		138	=head1 ERROR HANDLING
		139
		140	Libev knows three classes of errors: operating system errors, usage errors
		141	and internal errors (bugs).
		142
		143	When libev catches an operating system error it cannot handle (for example
		144	a system call indicating a condition libev cannot fix), it calls the callback
		145	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		146	abort. The default is to print a diagnostic message and to call C<abort
		147	()>.
		148
		149	When libev detects a usage error such as a negative timer interval, then
		150	it will print a diagnostic message and abort (via the C<assert> mechanism,
		151	so C<NDEBUG> will disable this checking): these are programming errors in
		152	the libev caller and need to be fixed there.
		153
		154	Libev also has a few internal error-checking C<assert>ions, and also has
		155	extensive consistency checking code. These do not trigger under normal
		156	circumstances, as they indicate either a bug in libev or worse.
		157
51		158
52	=head1 GLOBAL FUNCTIONS	159	=head1 GLOBAL FUNCTIONS
53		160
54	These functions can be called anytime, even before initialising the	161	These functions can be called anytime, even before initialising the
55	library in any way.	162	library in any way.
56		163
57	=over 4	164	=over 4
58		165
59	=item ev_tstamp ev_time ()	166	=item ev_tstamp ev_time ()
60		167
61	Returns the current time as libev would use it.	168	Returns the current time as libev would use it. Please note that the
		169	C<ev_now> function is usually faster and also often returns the timestamp
		170	you actually want to know.
		171
		172	=item ev_sleep (ev_tstamp interval)
		173
		174	Sleep for the given interval: The current thread will be blocked until
		175	either it is interrupted or the given time interval has passed. Basically
		176	this is a sub-second-resolution C<sleep ()>.
62		177
63	=item int ev_version_major ()	178	=item int ev_version_major ()
64		179
65	=item int ev_version_minor ()	180	=item int ev_version_minor ()
66		181
67	You can find out the major and minor version numbers of the library	182	You can find out the major and minor ABI version numbers of the library
68	you linked against by calling the functions C<ev_version_major> and	183	you linked against by calling the functions C<ev_version_major> and
69	C<ev_version_minor>. If you want, you can compare against the global	184	C<ev_version_minor>. If you want, you can compare against the global
70	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	185	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
71	version of the library your program was compiled against.	186	version of the library your program was compiled against.
72		187
		188	These version numbers refer to the ABI version of the library, not the
		189	release version.
		190
73	Usually, it's a good idea to terminate if the major versions mismatch,	191	Usually, it's a good idea to terminate if the major versions mismatch,
74	as this indicates an incompatible change. Minor versions are usually	192	as this indicates an incompatible change. Minor versions are usually
75	compatible to older versions, so a larger minor version alone is usually	193	compatible to older versions, so a larger minor version alone is usually
76	not a problem.	194	not a problem.
77		195
		196	Example: Make sure we haven't accidentally been linked against the wrong
		197	version (note, however, that this will not detect ABI mismatches :).
		198
		199	assert (("libev version mismatch",
		200	ev_version_major () == EV_VERSION_MAJOR
		201	&& ev_version_minor () >= EV_VERSION_MINOR));
		202
		203	=item unsigned int ev_supported_backends ()
		204
		205	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		206	value) compiled into this binary of libev (independent of their
		207	availability on the system you are running on). See C<ev_default_loop> for
		208	a description of the set values.
		209
		210	Example: make sure we have the epoll method, because yeah this is cool and
		211	a must have and can we have a torrent of it please!!!11
		212
		213	assert (("sorry, no epoll, no sex",
		214	ev_supported_backends () & EVBACKEND_EPOLL));
		215
		216	=item unsigned int ev_recommended_backends ()
		217
		218	Return the set of all backends compiled into this binary of libev and also
		219	recommended for this platform. This set is often smaller than the one
		220	returned by C<ev_supported_backends>, as for example kqueue is broken on
		221	most BSDs and will not be auto-detected unless you explicitly request it
		222	(assuming you know what you are doing). This is the set of backends that
		223	libev will probe for if you specify no backends explicitly.
		224
		225	=item unsigned int ev_embeddable_backends ()
		226
		227	Returns the set of backends that are embeddable in other event loops. This
		228	is the theoretical, all-platform, value. To find which backends
		229	might be supported on the current system, you would need to look at
		230	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
		231	recommended ones.
		232
		233	See the description of C<ev_embed> watchers for more info.
		234
78	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	235	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
79		236
80	Sets the allocation function to use (the prototype is similar to the	237	Sets the allocation function to use (the prototype is similar - the
81	realloc C function, the semantics are identical). It is used to allocate	238	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
82	and free memory (no surprises here). If it returns zero when memory	239	used to allocate and free memory (no surprises here). If it returns zero
83	needs to be allocated, the library might abort or take some potentially	240	when memory needs to be allocated (C<size != 0>), the library might abort
84	destructive action. The default is your system realloc function.	241	or take some potentially destructive action.
		242
		243	Since some systems (at least OpenBSD and Darwin) fail to implement
		244	correct C<realloc> semantics, libev will use a wrapper around the system
		245	C<realloc> and C<free> functions by default.
85		246
86	You could override this function in high-availability programs to, say,	247	You could override this function in high-availability programs to, say,
87	free some memory if it cannot allocate memory, to use a special allocator,	248	free some memory if it cannot allocate memory, to use a special allocator,
88	or even to sleep a while and retry until some memory is available.	249	or even to sleep a while and retry until some memory is available.
89		250
		251	Example: Replace the libev allocator with one that waits a bit and then
		252	retries (example requires a standards-compliant C<realloc>).
		253
		254	static void *
		255	persistent_realloc (void *ptr, size_t size)
		256	{
		257	for (;;)
		258	{
		259	void *newptr = realloc (ptr, size);
		260
		261	if (newptr)
		262	return newptr;
		263
		264	sleep (60);
		265	}
		266	}
		267
		268	...
		269	ev_set_allocator (persistent_realloc);
		270
90	=item ev_set_syserr_cb (void (*cb)(const char *msg));	271	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
91		272
92	Set the callback function to call on a retryable syscall error (such	273	Set the callback function to call on a retryable system call error (such
93	as failed select, poll, epoll_wait). The message is a printable string	274	as failed select, poll, epoll_wait). The message is a printable string
94	indicating the system call or subsystem causing the problem. If this	275	indicating the system call or subsystem causing the problem. If this
95	callback is set, then libev will expect it to remedy the sitution, no	276	callback is set, then libev will expect it to remedy the situation, no
96	matter what, when it returns. That is, libev will generally retry the	277	matter what, when it returns. That is, libev will generally retry the
97	requested operation, or, if the condition doesn't go away, do bad stuff	278	requested operation, or, if the condition doesn't go away, do bad stuff
98	(such as abort).	279	(such as abort).
99		280
		281	Example: This is basically the same thing that libev does internally, too.
		282
		283	static void
		284	fatal_error (const char *msg)
		285	{
		286	perror (msg);
		287	abort ();
		288	}
		289
		290	...
		291	ev_set_syserr_cb (fatal_error);
		292
100	=back	293	=back
101		294
102	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	295	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
103		296
104	An event loop is described by a C<struct ev_loop *>. The library knows two	297	An event loop is described by a C<struct ev_loop *> (the C<struct>
105	types of such loops, the I<default> loop, which supports signals and child	298	is I<not> optional in this case, as there is also an C<ev_loop>
106	events, and dynamically created loops which do not.	299	I<function>).
107		300
108	If you use threads, a common model is to run the default event loop	301	The library knows two types of such loops, the I<default> loop, which
109	in your main thread (or in a separate thread) and for each thread you	302	supports signals and child events, and dynamically created loops which do
110	create, you also create another event loop. Libev itself does no locking	303	not.
111	whatsoever, so if you mix calls to the same event loop in different
112	threads, make sure you lock (this is usually a bad idea, though, even if
113	done correctly, because it's hideous and inefficient).
114		304
115	=over 4	305	=over 4
116		306
117	=item struct ev_loop *ev_default_loop (unsigned int flags)	307	=item struct ev_loop *ev_default_loop (unsigned int flags)
118		308
119	This will initialise the default event loop if it hasn't been initialised	309	This will initialise the default event loop if it hasn't been initialised
120	yet and return it. If the default loop could not be initialised, returns	310	yet and return it. If the default loop could not be initialised, returns
121	false. If it already was initialised it simply returns it (and ignores the	311	false. If it already was initialised it simply returns it (and ignores the
122	flags).	312	flags. If that is troubling you, check C<ev_backend ()> afterwards).
123		313
124	If you don't know what event loop to use, use the one returned from this	314	If you don't know what event loop to use, use the one returned from this
125	function.	315	function.
126		316
		317	Note that this function is I<not> thread-safe, so if you want to use it
		318	from multiple threads, you have to lock (note also that this is unlikely,
		319	as loops cannot be shared easily between threads anyway).
		320
		321	The default loop is the only loop that can handle C<ev_signal> and
		322	C<ev_child> watchers, and to do this, it always registers a handler
		323	for C<SIGCHLD>. If this is a problem for your application you can either
		324	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
		325	can simply overwrite the C<SIGCHLD> signal handler I<after> calling
		326	C<ev_default_init>.
		327
127	The flags argument can be used to specify special behaviour or specific	328	The flags argument can be used to specify special behaviour or specific
128	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	329	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
129		330
130	It supports the following flags:	331	The following flags are supported:
131		332
132	=over 4	333	=over 4
133		334
134	=item C<EVFLAG_AUTO>	335	=item C<EVFLAG_AUTO>
135		336
136	The default flags value. Use this if you have no clue (it's the right	337	The default flags value. Use this if you have no clue (it's the right
137	thing, believe me).	338	thing, believe me).
138		339
139	=item C<EVFLAG_NOENV>	340	=item C<EVFLAG_NOENV>
140		341
141	If this flag bit is ored into the flag value (or the program runs setuid	342	If this flag bit is or'ed into the flag value (or the program runs setuid
142	or setgid) then libev will I<not> look at the environment variable	343	or setgid) then libev will I<not> look at the environment variable
143	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	344	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
144	override the flags completely if it is found in the environment. This is	345	override the flags completely if it is found in the environment. This is
145	useful to try out specific backends to test their performance, or to work	346	useful to try out specific backends to test their performance, or to work
146	around bugs.	347	around bugs.
147		348
		349	=item C<EVFLAG_FORKCHECK>
		350
		351	Instead of calling C<ev_loop_fork> manually after a fork, you can also
		352	make libev check for a fork in each iteration by enabling this flag.
		353
		354	This works by calling C<getpid ()> on every iteration of the loop,
		355	and thus this might slow down your event loop if you do a lot of loop
		356	iterations and little real work, but is usually not noticeable (on my
		357	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		358	without a system call and thus I<very> fast, but my GNU/Linux system also has
		359	C<pthread_atfork> which is even faster).
		360
		361	The big advantage of this flag is that you can forget about fork (and
		362	forget about forgetting to tell libev about forking) when you use this
		363	flag.
		364
		365	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		366	environment variable.
		367
		368	=item C<EVFLAG_NOINOTIFY>
		369
		370	When this flag is specified, then libev will not attempt to use the
		371	I<inotify> API for it's C<ev_stat> watchers. Apart from debugging and
		372	testing, this flag can be useful to conserve inotify file descriptors, as
		373	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		374
		375	=item C<EVFLAG_SIGNALFD>
		376
		377	When this flag is specified, then libev will attempt to use the
		378	I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This API
		379	delivers signals synchronously, which makes it both faster and might make
		380	it possible to get the queued signal data. It can also simplify signal
		381	handling with threads, as long as you properly block signals in your
		382	threads that are not interested in handling them.
		383
		384	Signalfd will not be used by default as this changes your signal mask, and
		385	there are a lot of shoddy libraries and programs (glib's threadpool for
		386	example) that can't properly initialise their signal masks.
		387
148	=item C<EVMETHOD_SELECT> (portable select backend)	388	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
149		389
		390	This is your standard select(2) backend. Not I<completely> standard, as
		391	libev tries to roll its own fd_set with no limits on the number of fds,
		392	but if that fails, expect a fairly low limit on the number of fds when
		393	using this backend. It doesn't scale too well (O(highest_fd)), but its
		394	usually the fastest backend for a low number of (low-numbered :) fds.
		395
		396	To get good performance out of this backend you need a high amount of
		397	parallelism (most of the file descriptors should be busy). If you are
		398	writing a server, you should C<accept ()> in a loop to accept as many
		399	connections as possible during one iteration. You might also want to have
		400	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		401	readiness notifications you get per iteration.
		402
		403	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		404	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		405	C<exceptfds> set on that platform).
		406
150	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	407	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
151		408
152	=item C<EVMETHOD_EPOLL> (linux only)	409	And this is your standard poll(2) backend. It's more complicated
		410	than select, but handles sparse fds better and has no artificial
		411	limit on the number of fds you can use (except it will slow down
		412	considerably with a lot of inactive fds). It scales similarly to select,
		413	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		414	performance tips.
153		415
154	=item C<EVMETHOD_KQUEUE> (some bsds only)	416	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		417	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
155		418
156	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	419	=item C<EVBACKEND_EPOLL> (value 4, Linux)
157		420
158	=item C<EVMETHOD_PORT> (solaris 10 only)	421	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		422	kernels).
159		423
160	If one or more of these are ored into the flags value, then only these	424	For few fds, this backend is a bit little slower than poll and select,
161	backends will be tried (in the reverse order as given here). If one are	425	but it scales phenomenally better. While poll and select usually scale
162	specified, any backend will do.	426	like O(total_fds) where n is the total number of fds (or the highest fd),
		427	epoll scales either O(1) or O(active_fds).
		428
		429	The epoll mechanism deserves honorable mention as the most misdesigned
		430	of the more advanced event mechanisms: mere annoyances include silently
		431	dropping file descriptors, requiring a system call per change per file
		432	descriptor (and unnecessary guessing of parameters), problems with dup and
		433	so on. The biggest issue is fork races, however - if a program forks then
		434	I<both> parent and child process have to recreate the epoll set, which can
		435	take considerable time (one syscall per file descriptor) and is of course
		436	hard to detect.
		437
		438	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		439	of course I<doesn't>, and epoll just loves to report events for totally
		440	I<different> file descriptors (even already closed ones, so one cannot
		441	even remove them from the set) than registered in the set (especially
		442	on SMP systems). Libev tries to counter these spurious notifications by
		443	employing an additional generation counter and comparing that against the
		444	events to filter out spurious ones, recreating the set when required.
		445
		446	While stopping, setting and starting an I/O watcher in the same iteration
		447	will result in some caching, there is still a system call per such
		448	incident (because the same I<file descriptor> could point to a different
		449	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		450	file descriptors might not work very well if you register events for both
		451	file descriptors.
		452
		453	Best performance from this backend is achieved by not unregistering all
		454	watchers for a file descriptor until it has been closed, if possible,
		455	i.e. keep at least one watcher active per fd at all times. Stopping and
		456	starting a watcher (without re-setting it) also usually doesn't cause
		457	extra overhead. A fork can both result in spurious notifications as well
		458	as in libev having to destroy and recreate the epoll object, which can
		459	take considerable time and thus should be avoided.
		460
		461	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		462	faster than epoll for maybe up to a hundred file descriptors, depending on
		463	the usage. So sad.
		464
		465	While nominally embeddable in other event loops, this feature is broken in
		466	all kernel versions tested so far.
		467
		468	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		469	C<EVBACKEND_POLL>.
		470
		471	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		472
		473	Kqueue deserves special mention, as at the time of this writing, it
		474	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		475	with anything but sockets and pipes, except on Darwin, where of course
		476	it's completely useless). Unlike epoll, however, whose brokenness
		477	is by design, these kqueue bugs can (and eventually will) be fixed
		478	without API changes to existing programs. For this reason it's not being
		479	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		480	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		481	system like NetBSD.
		482
		483	You still can embed kqueue into a normal poll or select backend and use it
		484	only for sockets (after having made sure that sockets work with kqueue on
		485	the target platform). See C<ev_embed> watchers for more info.
		486
		487	It scales in the same way as the epoll backend, but the interface to the
		488	kernel is more efficient (which says nothing about its actual speed, of
		489	course). While stopping, setting and starting an I/O watcher does never
		490	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		491	two event changes per incident. Support for C<fork ()> is very bad (but
		492	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		493	cases
		494
		495	This backend usually performs well under most conditions.
		496
		497	While nominally embeddable in other event loops, this doesn't work
		498	everywhere, so you might need to test for this. And since it is broken
		499	almost everywhere, you should only use it when you have a lot of sockets
		500	(for which it usually works), by embedding it into another event loop
		501	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		502	also broken on OS X)) and, did I mention it, using it only for sockets.
		503
		504	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		505	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		506	C<NOTE_EOF>.
		507
		508	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		509
		510	This is not implemented yet (and might never be, unless you send me an
		511	implementation). According to reports, C</dev/poll> only supports sockets
		512	and is not embeddable, which would limit the usefulness of this backend
		513	immensely.
		514
		515	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		516
		517	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		518	it's really slow, but it still scales very well (O(active_fds)).
		519
		520	Please note that Solaris event ports can deliver a lot of spurious
		521	notifications, so you need to use non-blocking I/O or other means to avoid
		522	blocking when no data (or space) is available.
		523
		524	While this backend scales well, it requires one system call per active
		525	file descriptor per loop iteration. For small and medium numbers of file
		526	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		527	might perform better.
		528
		529	On the positive side, with the exception of the spurious readiness
		530	notifications, this backend actually performed fully to specification
		531	in all tests and is fully embeddable, which is a rare feat among the
		532	OS-specific backends (I vastly prefer correctness over speed hacks).
		533
		534	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		535	C<EVBACKEND_POLL>.
		536
		537	=item C<EVBACKEND_ALL>
		538
		539	Try all backends (even potentially broken ones that wouldn't be tried
		540	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		541	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		542
		543	It is definitely not recommended to use this flag.
163		544
164	=back	545	=back
165		546
		547	If one or more of the backend flags are or'ed into the flags value,
		548	then only these backends will be tried (in the reverse order as listed
		549	here). If none are specified, all backends in C<ev_recommended_backends
		550	()> will be tried.
		551
		552	Example: This is the most typical usage.
		553
		554	if (!ev_default_loop (0))
		555	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		556
		557	Example: Restrict libev to the select and poll backends, and do not allow
		558	environment settings to be taken into account:
		559
		560	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		561
		562	Example: Use whatever libev has to offer, but make sure that kqueue is
		563	used if available (warning, breaks stuff, best use only with your own
		564	private event loop and only if you know the OS supports your types of
		565	fds):
		566
		567	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
		568
166	=item struct ev_loop *ev_loop_new (unsigned int flags)	569	=item struct ev_loop *ev_loop_new (unsigned int flags)
167		570
168	Similar to C<ev_default_loop>, but always creates a new event loop that is	571	Similar to C<ev_default_loop>, but always creates a new event loop that is
169	always distinct from the default loop. Unlike the default loop, it cannot	572	always distinct from the default loop.
170	handle signal and child watchers, and attempts to do so will be greeted by	573
171	undefined behaviour (or a failed assertion if assertions are enabled).	574	Note that this function I<is> thread-safe, and one common way to use
		575	libev with threads is indeed to create one loop per thread, and using the
		576	default loop in the "main" or "initial" thread.
		577
		578	Example: Try to create a event loop that uses epoll and nothing else.
		579
		580	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		581	if (!epoller)
		582	fatal ("no epoll found here, maybe it hides under your chair");
172		583
173	=item ev_default_destroy ()	584	=item ev_default_destroy ()
174		585
175	Destroys the default loop again (frees all memory and kernel state	586	Destroys the default loop (frees all memory and kernel state etc.). None
176	etc.). This stops all registered event watchers (by not touching them in	587	of the active event watchers will be stopped in the normal sense, so
177	any way whatsoever, although you cannot rely on this :).	588	e.g. C<ev_is_active> might still return true. It is your responsibility to
		589	either stop all watchers cleanly yourself I<before> calling this function,
		590	or cope with the fact afterwards (which is usually the easiest thing, you
		591	can just ignore the watchers and/or C<free ()> them for example).
		592
		593	Note that certain global state, such as signal state (and installed signal
		594	handlers), will not be freed by this function, and related watchers (such
		595	as signal and child watchers) would need to be stopped manually.
		596
		597	In general it is not advisable to call this function except in the
		598	rare occasion where you really need to free e.g. the signal handling
		599	pipe fds. If you need dynamically allocated loops it is better to use
		600	C<ev_loop_new> and C<ev_loop_destroy>.
178		601
179	=item ev_loop_destroy (loop)	602	=item ev_loop_destroy (loop)
180		603
181	Like C<ev_default_destroy>, but destroys an event loop created by an	604	Like C<ev_default_destroy>, but destroys an event loop created by an
182	earlier call to C<ev_loop_new>.	605	earlier call to C<ev_loop_new>.
183		606
184	=item ev_default_fork ()	607	=item ev_default_fork ()
185		608
		609	This function sets a flag that causes subsequent C<ev_loop> iterations
186	This function reinitialises the kernel state for backends that have	610	to reinitialise the kernel state for backends that have one. Despite the
187	one. Despite the name, you can call it anytime, but it makes most sense	611	name, you can call it anytime, but it makes most sense after forking, in
188	after forking, in either the parent or child process (or both, but that	612	the child process (or both child and parent, but that again makes little
189	again makes little sense).	613	sense). You I<must> call it in the child before using any of the libev
		614	functions, and it will only take effect at the next C<ev_loop> iteration.
190		615
191	You I<must> call this function after forking if and only if you want to	616	Again, you I<have> to call it on I<any> loop that you want to re-use after
192	use the event library in both processes. If you just fork+exec, you don't	617	a fork, I<even if you do not plan to use the loop in the parent>. This is
193	have to call it.	618	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		619	during fork.
		620
		621	On the other hand, you only need to call this function in the child
		622	process if and only if you want to use the event loop in the child. If you
		623	just fork+exec or create a new loop in the child, you don't have to call
		624	it at all.
194		625
195	The function itself is quite fast and it's usually not a problem to call	626	The function itself is quite fast and it's usually not a problem to call
196	it just in case after a fork. To make this easy, the function will fit in	627	it just in case after a fork. To make this easy, the function will fit in
197	quite nicely into a call to C<pthread_atfork>:	628	quite nicely into a call to C<pthread_atfork>:
198		629
…		…
200		631
201	=item ev_loop_fork (loop)	632	=item ev_loop_fork (loop)
202		633
203	Like C<ev_default_fork>, but acts on an event loop created by	634	Like C<ev_default_fork>, but acts on an event loop created by
204	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	635	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
205	after fork, and how you do this is entirely your own problem.	636	after fork that you want to re-use in the child, and how you keep track of
		637	them is entirely your own problem.
206		638
		639	=item int ev_is_default_loop (loop)
		640
		641	Returns true when the given loop is, in fact, the default loop, and false
		642	otherwise.
		643
		644	=item unsigned int ev_iteration (loop)
		645
		646	Returns the current iteration count for the loop, which is identical to
		647	the number of times libev did poll for new events. It starts at C<0> and
		648	happily wraps around with enough iterations.
		649
		650	This value can sometimes be useful as a generation counter of sorts (it
		651	"ticks" the number of loop iterations), as it roughly corresponds with
		652	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		653	prepare and check phases.
		654
207	=item unsigned int ev_method (loop)	655	=item unsigned int ev_depth (loop)
208		656
		657	Returns the number of times C<ev_loop> was entered minus the number of
		658	times C<ev_loop> was exited, in other words, the recursion depth.
		659
		660	Outside C<ev_loop>, this number is zero. In a callback, this number is
		661	C<1>, unless C<ev_loop> was invoked recursively (or from another thread),
		662	in which case it is higher.
		663
		664	Leaving C<ev_loop> abnormally (setjmp/longjmp, cancelling the thread
		665	etc.), doesn't count as "exit" - consider this as a hint to avoid such
		666	ungentleman behaviour unless it's really convenient.
		667
		668	=item unsigned int ev_backend (loop)
		669
209	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	670	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
210	use.	671	use.
211		672
212	=item ev_tstamp ev_now (loop)	673	=item ev_tstamp ev_now (loop)
213		674
214	Returns the current "event loop time", which is the time the event loop	675	Returns the current "event loop time", which is the time the event loop
215	got events and started processing them. This timestamp does not change	676	received events and started processing them. This timestamp does not
216	as long as callbacks are being processed, and this is also the base time	677	change as long as callbacks are being processed, and this is also the base
217	used for relative timers. You can treat it as the timestamp of the event	678	time used for relative timers. You can treat it as the timestamp of the
218	occuring (or more correctly, the mainloop finding out about it).	679	event occurring (or more correctly, libev finding out about it).
		680
		681	=item ev_now_update (loop)
		682
		683	Establishes the current time by querying the kernel, updating the time
		684	returned by C<ev_now ()> in the progress. This is a costly operation and
		685	is usually done automatically within C<ev_loop ()>.
		686
		687	This function is rarely useful, but when some event callback runs for a
		688	very long time without entering the event loop, updating libev's idea of
		689	the current time is a good idea.
		690
		691	See also L<The special problem of time updates> in the C<ev_timer> section.
		692
		693	=item ev_suspend (loop)
		694
		695	=item ev_resume (loop)
		696
		697	These two functions suspend and resume a loop, for use when the loop is
		698	not used for a while and timeouts should not be processed.
		699
		700	A typical use case would be an interactive program such as a game: When
		701	the user presses C<^Z> to suspend the game and resumes it an hour later it
		702	would be best to handle timeouts as if no time had actually passed while
		703	the program was suspended. This can be achieved by calling C<ev_suspend>
		704	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		705	C<ev_resume> directly afterwards to resume timer processing.
		706
		707	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		708	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		709	will be rescheduled (that is, they will lose any events that would have
		710	occurred while suspended).
		711
		712	After calling C<ev_suspend> you B<must not> call I<any> function on the
		713	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		714	without a previous call to C<ev_suspend>.
		715
		716	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		717	event loop time (see C<ev_now_update>).
219		718
220	=item ev_loop (loop, int flags)	719	=item ev_loop (loop, int flags)
221		720
222	Finally, this is it, the event handler. This function usually is called	721	Finally, this is it, the event handler. This function usually is called
223	after you initialised all your watchers and you want to start handling	722	after you have initialised all your watchers and you want to start
224	events.	723	handling events.
225		724
226	If the flags argument is specified as 0, it will not return until either	725	If the flags argument is specified as C<0>, it will not return until
227	no event watchers are active anymore or C<ev_unloop> was called.	726	either no event watchers are active anymore or C<ev_unloop> was called.
		727
		728	Please note that an explicit C<ev_unloop> is usually better than
		729	relying on all watchers to be stopped when deciding when a program has
		730	finished (especially in interactive programs), but having a program
		731	that automatically loops as long as it has to and no longer by virtue
		732	of relying on its watchers stopping correctly, that is truly a thing of
		733	beauty.
228		734
229	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	735	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
230	those events and any outstanding ones, but will not block your process in	736	those events and any already outstanding ones, but will not block your
231	case there are no events and will return after one iteration of the loop.	737	process in case there are no events and will return after one iteration of
		738	the loop.
232		739
233	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	740	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
234	neccessary) and will handle those and any outstanding ones. It will block	741	necessary) and will handle those and any already outstanding ones. It
235	your process until at least one new event arrives, and will return after	742	will block your process until at least one new event arrives (which could
		743	be an event internal to libev itself, so there is no guarantee that a
		744	user-registered callback will be called), and will return after one
236	one iteration of the loop.	745	iteration of the loop.
237		746
238	This flags value could be used to implement alternative looping	747	This is useful if you are waiting for some external event in conjunction
239	constructs, but the C<prepare> and C<check> watchers provide a better and	748	with something not expressible using other libev watchers (i.e. "roll your
240	more generic mechanism.	749	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		750	usually a better approach for this kind of thing.
		751
		752	Here are the gory details of what C<ev_loop> does:
		753
		754	- Before the first iteration, call any pending watchers.
		755	* If EVFLAG_FORKCHECK was used, check for a fork.
		756	- If a fork was detected (by any means), queue and call all fork watchers.
		757	- Queue and call all prepare watchers.
		758	- If we have been forked, detach and recreate the kernel state
		759	as to not disturb the other process.
		760	- Update the kernel state with all outstanding changes.
		761	- Update the "event loop time" (ev_now ()).
		762	- Calculate for how long to sleep or block, if at all
		763	(active idle watchers, EVLOOP_NONBLOCK or not having
		764	any active watchers at all will result in not sleeping).
		765	- Sleep if the I/O and timer collect interval say so.
		766	- Block the process, waiting for any events.
		767	- Queue all outstanding I/O (fd) events.
		768	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		769	- Queue all expired timers.
		770	- Queue all expired periodics.
		771	- Unless any events are pending now, queue all idle watchers.
		772	- Queue all check watchers.
		773	- Call all queued watchers in reverse order (i.e. check watchers first).
		774	Signals and child watchers are implemented as I/O watchers, and will
		775	be handled here by queueing them when their watcher gets executed.
		776	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
		777	were used, or there are no active watchers, return, otherwise
		778	continue with step *.
		779
		780	Example: Queue some jobs and then loop until no events are outstanding
		781	anymore.
		782
		783	... queue jobs here, make sure they register event watchers as long
		784	... as they still have work to do (even an idle watcher will do..)
		785	ev_loop (my_loop, 0);
		786	... jobs done or somebody called unloop. yeah!
241		787
242	=item ev_unloop (loop, how)	788	=item ev_unloop (loop, how)
243		789
244	Can be used to make a call to C<ev_loop> return early (but only after it	790	Can be used to make a call to C<ev_loop> return early (but only after it
245	has processed all outstanding events). The C<how> argument must be either	791	has processed all outstanding events). The C<how> argument must be either
246	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	792	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
247	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	793	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
248		794
		795	This "unloop state" will be cleared when entering C<ev_loop> again.
		796
		797	It is safe to call C<ev_unloop> from outside any C<ev_loop> calls.
		798
249	=item ev_ref (loop)	799	=item ev_ref (loop)
250		800
251	=item ev_unref (loop)	801	=item ev_unref (loop)
252		802
253	Ref/unref can be used to add or remove a reference count on the event	803	Ref/unref can be used to add or remove a reference count on the event
254	loop: Every watcher keeps one reference, and as long as the reference	804	loop: Every watcher keeps one reference, and as long as the reference
255	count is nonzero, C<ev_loop> will not return on its own. If you have	805	count is nonzero, C<ev_loop> will not return on its own.
256	a watcher you never unregister that should not keep C<ev_loop> from	806
257	returning, ev_unref() after starting, and ev_ref() before stopping it. For	807	This is useful when you have a watcher that you never intend to
		808	unregister, but that nevertheless should not keep C<ev_loop> from
		809	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		810	before stopping it.
		811
258	example, libev itself uses this for its internal signal pipe: It is not	812	As an example, libev itself uses this for its internal signal pipe: It
259	visible to the libev user and should not keep C<ev_loop> from exiting if	813	is not visible to the libev user and should not keep C<ev_loop> from
260	no event watchers registered by it are active. It is also an excellent	814	exiting if no event watchers registered by it are active. It is also an
261	way to do this for generic recurring timers or from within third-party	815	excellent way to do this for generic recurring timers or from within
262	libraries. Just remember to I<unref after start> and I<ref before stop>.	816	third-party libraries. Just remember to I<unref after start> and I<ref
		817	before stop> (but only if the watcher wasn't active before, or was active
		818	before, respectively. Note also that libev might stop watchers itself
		819	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		820	in the callback).
		821
		822	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		823	running when nothing else is active.
		824
		825	ev_signal exitsig;
		826	ev_signal_init (&exitsig, sig_cb, SIGINT);
		827	ev_signal_start (loop, &exitsig);
		828	evf_unref (loop);
		829
		830	Example: For some weird reason, unregister the above signal handler again.
		831
		832	ev_ref (loop);
		833	ev_signal_stop (loop, &exitsig);
		834
		835	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		836
		837	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		838
		839	These advanced functions influence the time that libev will spend waiting
		840	for events. Both time intervals are by default C<0>, meaning that libev
		841	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		842	latency.
		843
		844	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		845	allows libev to delay invocation of I/O and timer/periodic callbacks
		846	to increase efficiency of loop iterations (or to increase power-saving
		847	opportunities).
		848
		849	The idea is that sometimes your program runs just fast enough to handle
		850	one (or very few) event(s) per loop iteration. While this makes the
		851	program responsive, it also wastes a lot of CPU time to poll for new
		852	events, especially with backends like C<select ()> which have a high
		853	overhead for the actual polling but can deliver many events at once.
		854
		855	By setting a higher I<io collect interval> you allow libev to spend more
		856	time collecting I/O events, so you can handle more events per iteration,
		857	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		858	C<ev_timer>) will be not affected. Setting this to a non-null value will
		859	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		860	sleep time ensures that libev will not poll for I/O events more often then
		861	once per this interval, on average.
		862
		863	Likewise, by setting a higher I<timeout collect interval> you allow libev
		864	to spend more time collecting timeouts, at the expense of increased
		865	latency/jitter/inexactness (the watcher callback will be called
		866	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		867	value will not introduce any overhead in libev.
		868
		869	Many (busy) programs can usually benefit by setting the I/O collect
		870	interval to a value near C<0.1> or so, which is often enough for
		871	interactive servers (of course not for games), likewise for timeouts. It
		872	usually doesn't make much sense to set it to a lower value than C<0.01>,
		873	as this approaches the timing granularity of most systems. Note that if
		874	you do transactions with the outside world and you can't increase the
		875	parallelity, then this setting will limit your transaction rate (if you
		876	need to poll once per transaction and the I/O collect interval is 0.01,
		877	then you can't do more than 100 transactions per second).
		878
		879	Setting the I<timeout collect interval> can improve the opportunity for
		880	saving power, as the program will "bundle" timer callback invocations that
		881	are "near" in time together, by delaying some, thus reducing the number of
		882	times the process sleeps and wakes up again. Another useful technique to
		883	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		884	they fire on, say, one-second boundaries only.
		885
		886	Example: we only need 0.1s timeout granularity, and we wish not to poll
		887	more often than 100 times per second:
		888
		889	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		890	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		891
		892	=item ev_invoke_pending (loop)
		893
		894	This call will simply invoke all pending watchers while resetting their
		895	pending state. Normally, C<ev_loop> does this automatically when required,
		896	but when overriding the invoke callback this call comes handy.
		897
		898	=item int ev_pending_count (loop)
		899
		900	Returns the number of pending watchers - zero indicates that no watchers
		901	are pending.
		902
		903	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		904
		905	This overrides the invoke pending functionality of the loop: Instead of
		906	invoking all pending watchers when there are any, C<ev_loop> will call
		907	this callback instead. This is useful, for example, when you want to
		908	invoke the actual watchers inside another context (another thread etc.).
		909
		910	If you want to reset the callback, use C<ev_invoke_pending> as new
		911	callback.
		912
		913	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		914
		915	Sometimes you want to share the same loop between multiple threads. This
		916	can be done relatively simply by putting mutex_lock/unlock calls around
		917	each call to a libev function.
		918
		919	However, C<ev_loop> can run an indefinite time, so it is not feasible to
		920	wait for it to return. One way around this is to wake up the loop via
		921	C<ev_unloop> and C<av_async_send>, another way is to set these I<release>
		922	and I<acquire> callbacks on the loop.
		923
		924	When set, then C<release> will be called just before the thread is
		925	suspended waiting for new events, and C<acquire> is called just
		926	afterwards.
		927
		928	Ideally, C<release> will just call your mutex_unlock function, and
		929	C<acquire> will just call the mutex_lock function again.
		930
		931	While event loop modifications are allowed between invocations of
		932	C<release> and C<acquire> (that's their only purpose after all), no
		933	modifications done will affect the event loop, i.e. adding watchers will
		934	have no effect on the set of file descriptors being watched, or the time
		935	waited. Use an C<ev_async> watcher to wake up C<ev_loop> when you want it
		936	to take note of any changes you made.
		937
		938	In theory, threads executing C<ev_loop> will be async-cancel safe between
		939	invocations of C<release> and C<acquire>.
		940
		941	See also the locking example in the C<THREADS> section later in this
		942	document.
		943
		944	=item ev_set_userdata (loop, void *data)
		945
		946	=item ev_userdata (loop)
		947
		948	Set and retrieve a single C<void *> associated with a loop. When
		949	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		950	C<0.>
		951
		952	These two functions can be used to associate arbitrary data with a loop,
		953	and are intended solely for the C<invoke_pending_cb>, C<release> and
		954	C<acquire> callbacks described above, but of course can be (ab-)used for
		955	any other purpose as well.
		956
		957	=item ev_loop_verify (loop)
		958
		959	This function only does something when C<EV_VERIFY> support has been
		960	compiled in, which is the default for non-minimal builds. It tries to go
		961	through all internal structures and checks them for validity. If anything
		962	is found to be inconsistent, it will print an error message to standard
		963	error and call C<abort ()>.
		964
		965	This can be used to catch bugs inside libev itself: under normal
		966	circumstances, this function will never abort as of course libev keeps its
		967	data structures consistent.
263		968
264	=back	969	=back
265		970
		971
266	=head1 ANATOMY OF A WATCHER	972	=head1 ANATOMY OF A WATCHER
		973
		974	In the following description, uppercase C<TYPE> in names stands for the
		975	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		976	watchers and C<ev_io_start> for I/O watchers.
267		977
268	A watcher is a structure that you create and register to record your	978	A watcher is a structure that you create and register to record your
269	interest in some event. For instance, if you want to wait for STDIN to	979	interest in some event. For instance, if you want to wait for STDIN to
270	become readable, you would create an C<ev_io> watcher for that:	980	become readable, you would create an C<ev_io> watcher for that:
271		981
272	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	982	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
273	{	983	{
274	ev_io_stop (w);	984	ev_io_stop (w);
275	ev_unloop (loop, EVUNLOOP_ALL);	985	ev_unloop (loop, EVUNLOOP_ALL);
276	}	986	}
277		987
278	struct ev_loop *loop = ev_default_loop (0);	988	struct ev_loop *loop = ev_default_loop (0);
		989
279	struct ev_io stdin_watcher;	990	ev_io stdin_watcher;
		991
280	ev_init (&stdin_watcher, my_cb);	992	ev_init (&stdin_watcher, my_cb);
281	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	993	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
282	ev_io_start (loop, &stdin_watcher);	994	ev_io_start (loop, &stdin_watcher);
		995
283	ev_loop (loop, 0);	996	ev_loop (loop, 0);
284		997
285	As you can see, you are responsible for allocating the memory for your	998	As you can see, you are responsible for allocating the memory for your
286	watcher structures (and it is usually a bad idea to do this on the stack,	999	watcher structures (and it is I<usually> a bad idea to do this on the
287	although this can sometimes be quite valid).	1000	stack).
		1001
		1002	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1003	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
288		1004
289	Each watcher structure must be initialised by a call to C<ev_init	1005	Each watcher structure must be initialised by a call to C<ev_init
290	(watcher *, callback)>, which expects a callback to be provided. This	1006	(watcher *, callback)>, which expects a callback to be provided. This
291	callback gets invoked each time the event occurs (or, in the case of io	1007	callback gets invoked each time the event occurs (or, in the case of I/O
292	watchers, each time the event loop detects that the file descriptor given	1008	watchers, each time the event loop detects that the file descriptor given
293	is readable and/or writable).	1009	is readable and/or writable).
294		1010
295	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1011	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
296	with arguments specific to this watcher type. There is also a macro	1012	macro to configure it, with arguments specific to the watcher type. There
297	to combine initialisation and setting in one call: C<< ev_<type>_init	1013	is also a macro to combine initialisation and setting in one call: C<<
298	(watcher *, callback, ...) >>.	1014	ev_TYPE_init (watcher *, callback, ...) >>.
299		1015
300	To make the watcher actually watch out for events, you have to start it	1016	To make the watcher actually watch out for events, you have to start it
301	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1017	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
302	*) >>), and you can stop watching for events at any time by calling the	1018	*) >>), and you can stop watching for events at any time by calling the
303	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1019	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
304		1020
305	As long as your watcher is active (has been started but not stopped) you	1021	As long as your watcher is active (has been started but not stopped) you
306	must not touch the values stored in it. Most specifically you must never	1022	must not touch the values stored in it. Most specifically you must never
307	reinitialise it or call its set method.	1023	reinitialise it or call its C<ev_TYPE_set> macro.
308
309	You can check whether an event is active by calling the C<ev_is_active
310	(watcher *)> macro. To see whether an event is outstanding (but the
311	callback for it has not been called yet) you can use the C<ev_is_pending
312	(watcher *)> macro.
313		1024
314	Each and every callback receives the event loop pointer as first, the	1025	Each and every callback receives the event loop pointer as first, the
315	registered watcher structure as second, and a bitset of received events as	1026	registered watcher structure as second, and a bitset of received events as
316	third argument.	1027	third argument.
317		1028
…		…
326	=item C<EV_WRITE>	1037	=item C<EV_WRITE>
327		1038
328	The file descriptor in the C<ev_io> watcher has become readable and/or	1039	The file descriptor in the C<ev_io> watcher has become readable and/or
329	writable.	1040	writable.
330		1041
331	=item C<EV_TIMEOUT>	1042	=item C<EV_TIMER>
332		1043
333	The C<ev_timer> watcher has timed out.	1044	The C<ev_timer> watcher has timed out.
334		1045
335	=item C<EV_PERIODIC>	1046	=item C<EV_PERIODIC>
336		1047
…		…
341	The signal specified in the C<ev_signal> watcher has been received by a thread.	1052	The signal specified in the C<ev_signal> watcher has been received by a thread.
342		1053
343	=item C<EV_CHILD>	1054	=item C<EV_CHILD>
344		1055
345	The pid specified in the C<ev_child> watcher has received a status change.	1056	The pid specified in the C<ev_child> watcher has received a status change.
		1057
		1058	=item C<EV_STAT>
		1059
		1060	The path specified in the C<ev_stat> watcher changed its attributes somehow.
346		1061
347	=item C<EV_IDLE>	1062	=item C<EV_IDLE>
348		1063
349	The C<ev_idle> watcher has determined that you have nothing better to do.	1064	The C<ev_idle> watcher has determined that you have nothing better to do.
350		1065
…		…
358	received events. Callbacks of both watcher types can start and stop as	1073	received events. Callbacks of both watcher types can start and stop as
359	many watchers as they want, and all of them will be taken into account	1074	many watchers as they want, and all of them will be taken into account
360	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1075	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
361	C<ev_loop> from blocking).	1076	C<ev_loop> from blocking).
362		1077
		1078	=item C<EV_EMBED>
		1079
		1080	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1081
		1082	=item C<EV_FORK>
		1083
		1084	The event loop has been resumed in the child process after fork (see
		1085	C<ev_fork>).
		1086
		1087	=item C<EV_ASYNC>
		1088
		1089	The given async watcher has been asynchronously notified (see C<ev_async>).
		1090
		1091	=item C<EV_CUSTOM>
		1092
		1093	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1094	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1095
363	=item C<EV_ERROR>	1096	=item C<EV_ERROR>
364		1097
365	An unspecified error has occured, the watcher has been stopped. This might	1098	An unspecified error has occurred, the watcher has been stopped. This might
366	happen because the watcher could not be properly started because libev	1099	happen because the watcher could not be properly started because libev
367	ran out of memory, a file descriptor was found to be closed or any other	1100	ran out of memory, a file descriptor was found to be closed or any other
		1101	problem. Libev considers these application bugs.
		1102
368	problem. You best act on it by reporting the problem and somehow coping	1103	You best act on it by reporting the problem and somehow coping with the
369	with the watcher being stopped.	1104	watcher being stopped. Note that well-written programs should not receive
		1105	an error ever, so when your watcher receives it, this usually indicates a
		1106	bug in your program.
370		1107
371	Libev will usually signal a few "dummy" events together with an error,	1108	Libev will usually signal a few "dummy" events together with an error, for
372	for example it might indicate that a fd is readable or writable, and if	1109	example it might indicate that a fd is readable or writable, and if your
373	your callbacks is well-written it can just attempt the operation and cope	1110	callbacks is well-written it can just attempt the operation and cope with
374	with the error from read() or write(). This will not work in multithreaded	1111	the error from read() or write(). This will not work in multi-threaded
375	programs, though, so beware.	1112	programs, though, as the fd could already be closed and reused for another
		1113	thing, so beware.
376		1114
377	=back	1115	=back
378		1116
		1117	=head2 GENERIC WATCHER FUNCTIONS
		1118
		1119	=over 4
		1120
		1121	=item C<ev_init> (ev_TYPE *watcher, callback)
		1122
		1123	This macro initialises the generic portion of a watcher. The contents
		1124	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1125	the generic parts of the watcher are initialised, you I<need> to call
		1126	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1127	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1128	which rolls both calls into one.
		1129
		1130	You can reinitialise a watcher at any time as long as it has been stopped
		1131	(or never started) and there are no pending events outstanding.
		1132
		1133	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1134	int revents)>.
		1135
		1136	Example: Initialise an C<ev_io> watcher in two steps.
		1137
		1138	ev_io w;
		1139	ev_init (&w, my_cb);
		1140	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1141
		1142	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1143
		1144	This macro initialises the type-specific parts of a watcher. You need to
		1145	call C<ev_init> at least once before you call this macro, but you can
		1146	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1147	macro on a watcher that is active (it can be pending, however, which is a
		1148	difference to the C<ev_init> macro).
		1149
		1150	Although some watcher types do not have type-specific arguments
		1151	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1152
		1153	See C<ev_init>, above, for an example.
		1154
		1155	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1156
		1157	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1158	calls into a single call. This is the most convenient method to initialise
		1159	a watcher. The same limitations apply, of course.
		1160
		1161	Example: Initialise and set an C<ev_io> watcher in one step.
		1162
		1163	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1164
		1165	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1166
		1167	Starts (activates) the given watcher. Only active watchers will receive
		1168	events. If the watcher is already active nothing will happen.
		1169
		1170	Example: Start the C<ev_io> watcher that is being abused as example in this
		1171	whole section.
		1172
		1173	ev_io_start (EV_DEFAULT_UC, &w);
		1174
		1175	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1176
		1177	Stops the given watcher if active, and clears the pending status (whether
		1178	the watcher was active or not).
		1179
		1180	It is possible that stopped watchers are pending - for example,
		1181	non-repeating timers are being stopped when they become pending - but
		1182	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1183	pending. If you want to free or reuse the memory used by the watcher it is
		1184	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1185
		1186	=item bool ev_is_active (ev_TYPE *watcher)
		1187
		1188	Returns a true value iff the watcher is active (i.e. it has been started
		1189	and not yet been stopped). As long as a watcher is active you must not modify
		1190	it.
		1191
		1192	=item bool ev_is_pending (ev_TYPE *watcher)
		1193
		1194	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1195	events but its callback has not yet been invoked). As long as a watcher
		1196	is pending (but not active) you must not call an init function on it (but
		1197	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1198	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1199	it).
		1200
		1201	=item callback ev_cb (ev_TYPE *watcher)
		1202
		1203	Returns the callback currently set on the watcher.
		1204
		1205	=item ev_cb_set (ev_TYPE *watcher, callback)
		1206
		1207	Change the callback. You can change the callback at virtually any time
		1208	(modulo threads).
		1209
		1210	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1211
		1212	=item int ev_priority (ev_TYPE *watcher)
		1213
		1214	Set and query the priority of the watcher. The priority is a small
		1215	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1216	(default: C<-2>). Pending watchers with higher priority will be invoked
		1217	before watchers with lower priority, but priority will not keep watchers
		1218	from being executed (except for C<ev_idle> watchers).
		1219
		1220	If you need to suppress invocation when higher priority events are pending
		1221	you need to look at C<ev_idle> watchers, which provide this functionality.
		1222
		1223	You I<must not> change the priority of a watcher as long as it is active or
		1224	pending.
		1225
		1226	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1227	fine, as long as you do not mind that the priority value you query might
		1228	or might not have been clamped to the valid range.
		1229
		1230	The default priority used by watchers when no priority has been set is
		1231	always C<0>, which is supposed to not be too high and not be too low :).
		1232
		1233	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1234	priorities.
		1235
		1236	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1237
		1238	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1239	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1240	can deal with that fact, as both are simply passed through to the
		1241	callback.
		1242
		1243	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1244
		1245	If the watcher is pending, this function clears its pending status and
		1246	returns its C<revents> bitset (as if its callback was invoked). If the
		1247	watcher isn't pending it does nothing and returns C<0>.
		1248
		1249	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1250	callback to be invoked, which can be accomplished with this function.
		1251
		1252	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1253
		1254	Feeds the given event set into the event loop, as if the specified event
		1255	had happened for the specified watcher (which must be a pointer to an
		1256	initialised but not necessarily started event watcher). Obviously you must
		1257	not free the watcher as long as it has pending events.
		1258
		1259	Stopping the watcher, letting libev invoke it, or calling
		1260	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1261	not started in the first place.
		1262
		1263	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1264	functions that do not need a watcher.
		1265
		1266	=back
		1267
		1268
379	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1269	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
380		1270
381	Each watcher has, by default, a member C<void *data> that you can change	1271	Each watcher has, by default, a member C<void *data> that you can change
382	and read at any time, libev will completely ignore it. This can be used	1272	and read at any time: libev will completely ignore it. This can be used
383	to associate arbitrary data with your watcher. If you need more data and	1273	to associate arbitrary data with your watcher. If you need more data and
384	don't want to allocate memory and store a pointer to it in that data	1274	don't want to allocate memory and store a pointer to it in that data
385	member, you can also "subclass" the watcher type and provide your own	1275	member, you can also "subclass" the watcher type and provide your own
386	data:	1276	data:
387		1277
388	struct my_io	1278	struct my_io
389	{	1279	{
390	struct ev_io io;	1280	ev_io io;
391	int otherfd;	1281	int otherfd;
392	void *somedata;	1282	void *somedata;
393	struct whatever *mostinteresting;	1283	struct whatever *mostinteresting;
394	}	1284	};
		1285
		1286	...
		1287	struct my_io w;
		1288	ev_io_init (&w.io, my_cb, fd, EV_READ);
395		1289
396	And since your callback will be called with a pointer to the watcher, you	1290	And since your callback will be called with a pointer to the watcher, you
397	can cast it back to your own type:	1291	can cast it back to your own type:
398		1292
399	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1293	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
400	{	1294	{
401	struct my_io *w = (struct my_io *)w_;	1295	struct my_io *w = (struct my_io *)w_;
402	...	1296	...
403	}	1297	}
404		1298
405	More interesting and less C-conformant ways of catsing your callback type	1299	More interesting and less C-conformant ways of casting your callback type
406	have been omitted....	1300	instead have been omitted.
		1301
		1302	Another common scenario is to use some data structure with multiple
		1303	embedded watchers:
		1304
		1305	struct my_biggy
		1306	{
		1307	int some_data;
		1308	ev_timer t1;
		1309	ev_timer t2;
		1310	}
		1311
		1312	In this case getting the pointer to C<my_biggy> is a bit more
		1313	complicated: Either you store the address of your C<my_biggy> struct
		1314	in the C<data> member of the watcher (for woozies), or you need to use
		1315	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1316	programmers):
		1317
		1318	#include <stddef.h>
		1319
		1320	static void
		1321	t1_cb (EV_P_ ev_timer *w, int revents)
		1322	{
		1323	struct my_biggy big = (struct my_biggy *)
		1324	(((char *)w) - offsetof (struct my_biggy, t1));
		1325	}
		1326
		1327	static void
		1328	t2_cb (EV_P_ ev_timer *w, int revents)
		1329	{
		1330	struct my_biggy big = (struct my_biggy *)
		1331	(((char *)w) - offsetof (struct my_biggy, t2));
		1332	}
		1333
		1334	=head2 WATCHER PRIORITY MODELS
		1335
		1336	Many event loops support I<watcher priorities>, which are usually small
		1337	integers that influence the ordering of event callback invocation
		1338	between watchers in some way, all else being equal.
		1339
		1340	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1341	description for the more technical details such as the actual priority
		1342	range.
		1343
		1344	There are two common ways how these these priorities are being interpreted
		1345	by event loops:
		1346
		1347	In the more common lock-out model, higher priorities "lock out" invocation
		1348	of lower priority watchers, which means as long as higher priority
		1349	watchers receive events, lower priority watchers are not being invoked.
		1350
		1351	The less common only-for-ordering model uses priorities solely to order
		1352	callback invocation within a single event loop iteration: Higher priority
		1353	watchers are invoked before lower priority ones, but they all get invoked
		1354	before polling for new events.
		1355
		1356	Libev uses the second (only-for-ordering) model for all its watchers
		1357	except for idle watchers (which use the lock-out model).
		1358
		1359	The rationale behind this is that implementing the lock-out model for
		1360	watchers is not well supported by most kernel interfaces, and most event
		1361	libraries will just poll for the same events again and again as long as
		1362	their callbacks have not been executed, which is very inefficient in the
		1363	common case of one high-priority watcher locking out a mass of lower
		1364	priority ones.
		1365
		1366	Static (ordering) priorities are most useful when you have two or more
		1367	watchers handling the same resource: a typical usage example is having an
		1368	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1369	timeouts. Under load, data might be received while the program handles
		1370	other jobs, but since timers normally get invoked first, the timeout
		1371	handler will be executed before checking for data. In that case, giving
		1372	the timer a lower priority than the I/O watcher ensures that I/O will be
		1373	handled first even under adverse conditions (which is usually, but not
		1374	always, what you want).
		1375
		1376	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1377	will only be executed when no same or higher priority watchers have
		1378	received events, they can be used to implement the "lock-out" model when
		1379	required.
		1380
		1381	For example, to emulate how many other event libraries handle priorities,
		1382	you can associate an C<ev_idle> watcher to each such watcher, and in
		1383	the normal watcher callback, you just start the idle watcher. The real
		1384	processing is done in the idle watcher callback. This causes libev to
		1385	continuously poll and process kernel event data for the watcher, but when
		1386	the lock-out case is known to be rare (which in turn is rare :), this is
		1387	workable.
		1388
		1389	Usually, however, the lock-out model implemented that way will perform
		1390	miserably under the type of load it was designed to handle. In that case,
		1391	it might be preferable to stop the real watcher before starting the
		1392	idle watcher, so the kernel will not have to process the event in case
		1393	the actual processing will be delayed for considerable time.
		1394
		1395	Here is an example of an I/O watcher that should run at a strictly lower
		1396	priority than the default, and which should only process data when no
		1397	other events are pending:
		1398
		1399	ev_idle idle; // actual processing watcher
		1400	ev_io io; // actual event watcher
		1401
		1402	static void
		1403	io_cb (EV_P_ ev_io *w, int revents)
		1404	{
		1405	// stop the I/O watcher, we received the event, but
		1406	// are not yet ready to handle it.
		1407	ev_io_stop (EV_A_ w);
		1408
		1409	// start the idle watcher to handle the actual event.
		1410	// it will not be executed as long as other watchers
		1411	// with the default priority are receiving events.
		1412	ev_idle_start (EV_A_ &idle);
		1413	}
		1414
		1415	static void
		1416	idle_cb (EV_P_ ev_idle *w, int revents)
		1417	{
		1418	// actual processing
		1419	read (STDIN_FILENO, ...);
		1420
		1421	// have to start the I/O watcher again, as
		1422	// we have handled the event
		1423	ev_io_start (EV_P_ &io);
		1424	}
		1425
		1426	// initialisation
		1427	ev_idle_init (&idle, idle_cb);
		1428	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1429	ev_io_start (EV_DEFAULT_ &io);
		1430
		1431	In the "real" world, it might also be beneficial to start a timer, so that
		1432	low-priority connections can not be locked out forever under load. This
		1433	enables your program to keep a lower latency for important connections
		1434	during short periods of high load, while not completely locking out less
		1435	important ones.
407		1436
408		1437
409	=head1 WATCHER TYPES	1438	=head1 WATCHER TYPES
410		1439
411	This section describes each watcher in detail, but will not repeat	1440	This section describes each watcher in detail, but will not repeat
412	information given in the last section.	1441	information given in the last section. Any initialisation/set macros,
		1442	functions and members specific to the watcher type are explained.
413		1443
		1444	Members are additionally marked with either I<[read-only]>, meaning that,
		1445	while the watcher is active, you can look at the member and expect some
		1446	sensible content, but you must not modify it (you can modify it while the
		1447	watcher is stopped to your hearts content), or I<[read-write]>, which
		1448	means you can expect it to have some sensible content while the watcher
		1449	is active, but you can also modify it. Modifying it may not do something
		1450	sensible or take immediate effect (or do anything at all), but libev will
		1451	not crash or malfunction in any way.
		1452
		1453
414	=head2 C<ev_io> - is this file descriptor readable or writable	1454	=head2 C<ev_io> - is this file descriptor readable or writable?
415		1455
416	I/O watchers check whether a file descriptor is readable or writable	1456	I/O watchers check whether a file descriptor is readable or writable
417	in each iteration of the event loop (This behaviour is called	1457	in each iteration of the event loop, or, more precisely, when reading
418	level-triggering because you keep receiving events as long as the	1458	would not block the process and writing would at least be able to write
419	condition persists. Remember you can stop the watcher if you don't want to	1459	some data. This behaviour is called level-triggering because you keep
420	act on the event and neither want to receive future events).	1460	receiving events as long as the condition persists. Remember you can stop
		1461	the watcher if you don't want to act on the event and neither want to
		1462	receive future events.
421		1463
422	In general you can register as many read and/or write event watchers per	1464	In general you can register as many read and/or write event watchers per
423	fd as you want (as long as you don't confuse yourself). Setting all file	1465	fd as you want (as long as you don't confuse yourself). Setting all file
424	descriptors to non-blocking mode is also usually a good idea (but not	1466	descriptors to non-blocking mode is also usually a good idea (but not
425	required if you know what you are doing).	1467	required if you know what you are doing).
426		1468
427	You have to be careful with dup'ed file descriptors, though. Some backends	1469	If you cannot use non-blocking mode, then force the use of a
428	(the linux epoll backend is a notable example) cannot handle dup'ed file	1470	known-to-be-good backend (at the time of this writing, this includes only
429	descriptors correctly if you register interest in two or more fds pointing	1471	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
430	to the same underlying file/socket etc. description (that is, they share	1472	descriptors for which non-blocking operation makes no sense (such as
431	the same underlying "file open").	1473	files) - libev doesn't guarantee any specific behaviour in that case.
432		1474
433	If you must do this, then force the use of a known-to-be-good backend	1475	Another thing you have to watch out for is that it is quite easy to
434	(at the time of this writing, this includes only EVMETHOD_SELECT and	1476	receive "spurious" readiness notifications, that is your callback might
435	EVMETHOD_POLL).	1477	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1478	because there is no data. Not only are some backends known to create a
		1479	lot of those (for example Solaris ports), it is very easy to get into
		1480	this situation even with a relatively standard program structure. Thus
		1481	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1482	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1483
		1484	If you cannot run the fd in non-blocking mode (for example you should
		1485	not play around with an Xlib connection), then you have to separately
		1486	re-test whether a file descriptor is really ready with a known-to-be good
		1487	interface such as poll (fortunately in our Xlib example, Xlib already
		1488	does this on its own, so its quite safe to use). Some people additionally
		1489	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1490	indefinitely.
		1491
		1492	But really, best use non-blocking mode.
		1493
		1494	=head3 The special problem of disappearing file descriptors
		1495
		1496	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1497	descriptor (either due to calling C<close> explicitly or any other means,
		1498	such as C<dup2>). The reason is that you register interest in some file
		1499	descriptor, but when it goes away, the operating system will silently drop
		1500	this interest. If another file descriptor with the same number then is
		1501	registered with libev, there is no efficient way to see that this is, in
		1502	fact, a different file descriptor.
		1503
		1504	To avoid having to explicitly tell libev about such cases, libev follows
		1505	the following policy: Each time C<ev_io_set> is being called, libev
		1506	will assume that this is potentially a new file descriptor, otherwise
		1507	it is assumed that the file descriptor stays the same. That means that
		1508	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1509	descriptor even if the file descriptor number itself did not change.
		1510
		1511	This is how one would do it normally anyway, the important point is that
		1512	the libev application should not optimise around libev but should leave
		1513	optimisations to libev.
		1514
		1515	=head3 The special problem of dup'ed file descriptors
		1516
		1517	Some backends (e.g. epoll), cannot register events for file descriptors,
		1518	but only events for the underlying file descriptions. That means when you
		1519	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1520	events for them, only one file descriptor might actually receive events.
		1521
		1522	There is no workaround possible except not registering events
		1523	for potentially C<dup ()>'ed file descriptors, or to resort to
		1524	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1525
		1526	=head3 The special problem of fork
		1527
		1528	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1529	useless behaviour. Libev fully supports fork, but needs to be told about
		1530	it in the child.
		1531
		1532	To support fork in your programs, you either have to call
		1533	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1534	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1535	C<EVBACKEND_POLL>.
		1536
		1537	=head3 The special problem of SIGPIPE
		1538
		1539	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1540	when writing to a pipe whose other end has been closed, your program gets
		1541	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1542	this is sensible behaviour, for daemons, this is usually undesirable.
		1543
		1544	So when you encounter spurious, unexplained daemon exits, make sure you
		1545	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1546	somewhere, as that would have given you a big clue).
		1547
		1548	=head3 The special problem of accept()ing when you can't
		1549
		1550	Many implementations of the POSIX C<accept> function (for example,
		1551	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1552	connection from the pending queue in all error cases.
		1553
		1554	For example, larger servers often run out of file descriptors (because
		1555	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1556	rejecting the connection, leading to libev signalling readiness on
		1557	the next iteration again (the connection still exists after all), and
		1558	typically causing the program to loop at 100% CPU usage.
		1559
		1560	Unfortunately, the set of errors that cause this issue differs between
		1561	operating systems, there is usually little the app can do to remedy the
		1562	situation, and no known thread-safe method of removing the connection to
		1563	cope with overload is known (to me).
		1564
		1565	One of the easiest ways to handle this situation is to just ignore it
		1566	- when the program encounters an overload, it will just loop until the
		1567	situation is over. While this is a form of busy waiting, no OS offers an
		1568	event-based way to handle this situation, so it's the best one can do.
		1569
		1570	A better way to handle the situation is to log any errors other than
		1571	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1572	messages, and continue as usual, which at least gives the user an idea of
		1573	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1574	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1575	usage.
		1576
		1577	If your program is single-threaded, then you could also keep a dummy file
		1578	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1579	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1580	close that fd, and create a new dummy fd. This will gracefully refuse
		1581	clients under typical overload conditions.
		1582
		1583	The last way to handle it is to simply log the error and C<exit>, as
		1584	is often done with C<malloc> failures, but this results in an easy
		1585	opportunity for a DoS attack.
		1586
		1587	=head3 Watcher-Specific Functions
436		1588
437	=over 4	1589	=over 4
438		1590
439	=item ev_io_init (ev_io *, callback, int fd, int events)	1591	=item ev_io_init (ev_io *, callback, int fd, int events)
440		1592
441	=item ev_io_set (ev_io *, int fd, int events)	1593	=item ev_io_set (ev_io *, int fd, int events)
442		1594
443	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1595	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
444	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1596	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
445	EV_WRITE> to receive the given events.	1597	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1598
		1599	=item int fd [read-only]
		1600
		1601	The file descriptor being watched.
		1602
		1603	=item int events [read-only]
		1604
		1605	The events being watched.
446		1606
447	=back	1607	=back
448		1608
		1609	=head3 Examples
		1610
		1611	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1612	readable, but only once. Since it is likely line-buffered, you could
		1613	attempt to read a whole line in the callback.
		1614
		1615	static void
		1616	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1617	{
		1618	ev_io_stop (loop, w);
		1619	.. read from stdin here (or from w->fd) and handle any I/O errors
		1620	}
		1621
		1622	...
		1623	struct ev_loop *loop = ev_default_init (0);
		1624	ev_io stdin_readable;
		1625	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1626	ev_io_start (loop, &stdin_readable);
		1627	ev_loop (loop, 0);
		1628
		1629
449	=head2 C<ev_timer> - relative and optionally recurring timeouts	1630	=head2 C<ev_timer> - relative and optionally repeating timeouts
450		1631
451	Timer watchers are simple relative timers that generate an event after a	1632	Timer watchers are simple relative timers that generate an event after a
452	given time, and optionally repeating in regular intervals after that.	1633	given time, and optionally repeating in regular intervals after that.
453		1634
454	The timers are based on real time, that is, if you register an event that	1635	The timers are based on real time, that is, if you register an event that
455	times out after an hour and you reset your system clock to last years	1636	times out after an hour and you reset your system clock to January last
456	time, it will still time out after (roughly) and hour. "Roughly" because	1637	year, it will still time out after (roughly) one hour. "Roughly" because
457	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1638	detecting time jumps is hard, and some inaccuracies are unavoidable (the
458	monotonic clock option helps a lot here).	1639	monotonic clock option helps a lot here).
		1640
		1641	The callback is guaranteed to be invoked only I<after> its timeout has
		1642	passed (not I<at>, so on systems with very low-resolution clocks this
		1643	might introduce a small delay). If multiple timers become ready during the
		1644	same loop iteration then the ones with earlier time-out values are invoked
		1645	before ones of the same priority with later time-out values (but this is
		1646	no longer true when a callback calls C<ev_loop> recursively).
		1647
		1648	=head3 Be smart about timeouts
		1649
		1650	Many real-world problems involve some kind of timeout, usually for error
		1651	recovery. A typical example is an HTTP request - if the other side hangs,
		1652	you want to raise some error after a while.
		1653
		1654	What follows are some ways to handle this problem, from obvious and
		1655	inefficient to smart and efficient.
		1656
		1657	In the following, a 60 second activity timeout is assumed - a timeout that
		1658	gets reset to 60 seconds each time there is activity (e.g. each time some
		1659	data or other life sign was received).
		1660
		1661	=over 4
		1662
		1663	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1664
		1665	This is the most obvious, but not the most simple way: In the beginning,
		1666	start the watcher:
		1667
		1668	ev_timer_init (timer, callback, 60., 0.);
		1669	ev_timer_start (loop, timer);
		1670
		1671	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1672	and start it again:
		1673
		1674	ev_timer_stop (loop, timer);
		1675	ev_timer_set (timer, 60., 0.);
		1676	ev_timer_start (loop, timer);
		1677
		1678	This is relatively simple to implement, but means that each time there is
		1679	some activity, libev will first have to remove the timer from its internal
		1680	data structure and then add it again. Libev tries to be fast, but it's
		1681	still not a constant-time operation.
		1682
		1683	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1684
		1685	This is the easiest way, and involves using C<ev_timer_again> instead of
		1686	C<ev_timer_start>.
		1687
		1688	To implement this, configure an C<ev_timer> with a C<repeat> value
		1689	of C<60> and then call C<ev_timer_again> at start and each time you
		1690	successfully read or write some data. If you go into an idle state where
		1691	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1692	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1693
		1694	That means you can ignore both the C<ev_timer_start> function and the
		1695	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1696	member and C<ev_timer_again>.
		1697
		1698	At start:
		1699
		1700	ev_init (timer, callback);
		1701	timer->repeat = 60.;
		1702	ev_timer_again (loop, timer);
		1703
		1704	Each time there is some activity:
		1705
		1706	ev_timer_again (loop, timer);
		1707
		1708	It is even possible to change the time-out on the fly, regardless of
		1709	whether the watcher is active or not:
		1710
		1711	timer->repeat = 30.;
		1712	ev_timer_again (loop, timer);
		1713
		1714	This is slightly more efficient then stopping/starting the timer each time
		1715	you want to modify its timeout value, as libev does not have to completely
		1716	remove and re-insert the timer from/into its internal data structure.
		1717
		1718	It is, however, even simpler than the "obvious" way to do it.
		1719
		1720	=item 3. Let the timer time out, but then re-arm it as required.
		1721
		1722	This method is more tricky, but usually most efficient: Most timeouts are
		1723	relatively long compared to the intervals between other activity - in
		1724	our example, within 60 seconds, there are usually many I/O events with
		1725	associated activity resets.
		1726
		1727	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1728	but remember the time of last activity, and check for a real timeout only
		1729	within the callback:
		1730
		1731	ev_tstamp last_activity; // time of last activity
		1732
		1733	static void
		1734	callback (EV_P_ ev_timer *w, int revents)
		1735	{
		1736	ev_tstamp now = ev_now (EV_A);
		1737	ev_tstamp timeout = last_activity + 60.;
		1738
		1739	// if last_activity + 60. is older than now, we did time out
		1740	if (timeout < now)
		1741	{
		1742	// timeout occurred, take action
		1743	}
		1744	else
		1745	{
		1746	// callback was invoked, but there was some activity, re-arm
		1747	// the watcher to fire in last_activity + 60, which is
		1748	// guaranteed to be in the future, so "again" is positive:
		1749	w->repeat = timeout - now;
		1750	ev_timer_again (EV_A_ w);
		1751	}
		1752	}
		1753
		1754	To summarise the callback: first calculate the real timeout (defined
		1755	as "60 seconds after the last activity"), then check if that time has
		1756	been reached, which means something I<did>, in fact, time out. Otherwise
		1757	the callback was invoked too early (C<timeout> is in the future), so
		1758	re-schedule the timer to fire at that future time, to see if maybe we have
		1759	a timeout then.
		1760
		1761	Note how C<ev_timer_again> is used, taking advantage of the
		1762	C<ev_timer_again> optimisation when the timer is already running.
		1763
		1764	This scheme causes more callback invocations (about one every 60 seconds
		1765	minus half the average time between activity), but virtually no calls to
		1766	libev to change the timeout.
		1767
		1768	To start the timer, simply initialise the watcher and set C<last_activity>
		1769	to the current time (meaning we just have some activity :), then call the
		1770	callback, which will "do the right thing" and start the timer:
		1771
		1772	ev_init (timer, callback);
		1773	last_activity = ev_now (loop);
		1774	callback (loop, timer, EV_TIMER);
		1775
		1776	And when there is some activity, simply store the current time in
		1777	C<last_activity>, no libev calls at all:
		1778
		1779	last_activity = ev_now (loop);
		1780
		1781	This technique is slightly more complex, but in most cases where the
		1782	time-out is unlikely to be triggered, much more efficient.
		1783
		1784	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1785	callback :) - just change the timeout and invoke the callback, which will
		1786	fix things for you.
		1787
		1788	=item 4. Wee, just use a double-linked list for your timeouts.
		1789
		1790	If there is not one request, but many thousands (millions...), all
		1791	employing some kind of timeout with the same timeout value, then one can
		1792	do even better:
		1793
		1794	When starting the timeout, calculate the timeout value and put the timeout
		1795	at the I<end> of the list.
		1796
		1797	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1798	the list is expected to fire (for example, using the technique #3).
		1799
		1800	When there is some activity, remove the timer from the list, recalculate
		1801	the timeout, append it to the end of the list again, and make sure to
		1802	update the C<ev_timer> if it was taken from the beginning of the list.
		1803
		1804	This way, one can manage an unlimited number of timeouts in O(1) time for
		1805	starting, stopping and updating the timers, at the expense of a major
		1806	complication, and having to use a constant timeout. The constant timeout
		1807	ensures that the list stays sorted.
		1808
		1809	=back
		1810
		1811	So which method the best?
		1812
		1813	Method #2 is a simple no-brain-required solution that is adequate in most
		1814	situations. Method #3 requires a bit more thinking, but handles many cases
		1815	better, and isn't very complicated either. In most case, choosing either
		1816	one is fine, with #3 being better in typical situations.
		1817
		1818	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1819	rather complicated, but extremely efficient, something that really pays
		1820	off after the first million or so of active timers, i.e. it's usually
		1821	overkill :)
		1822
		1823	=head3 The special problem of time updates
		1824
		1825	Establishing the current time is a costly operation (it usually takes at
		1826	least two system calls): EV therefore updates its idea of the current
		1827	time only before and after C<ev_loop> collects new events, which causes a
		1828	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1829	lots of events in one iteration.
459		1830
460	The relative timeouts are calculated relative to the C<ev_now ()>	1831	The relative timeouts are calculated relative to the C<ev_now ()>
461	time. This is usually the right thing as this timestamp refers to the time	1832	time. This is usually the right thing as this timestamp refers to the time
462	of the event triggering whatever timeout you are modifying/starting. If	1833	of the event triggering whatever timeout you are modifying/starting. If
463	you suspect event processing to be delayed and you *need* to base the timeout	1834	you suspect event processing to be delayed and you I<need> to base the
464	on the current time, use something like this to adjust for this:	1835	timeout on the current time, use something like this to adjust for this:
465		1836
466	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1837	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
467		1838
		1839	If the event loop is suspended for a long time, you can also force an
		1840	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1841	()>.
		1842
		1843	=head3 The special problems of suspended animation
		1844
		1845	When you leave the server world it is quite customary to hit machines that
		1846	can suspend/hibernate - what happens to the clocks during such a suspend?
		1847
		1848	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1849	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1850	to run until the system is suspended, but they will not advance while the
		1851	system is suspended. That means, on resume, it will be as if the program
		1852	was frozen for a few seconds, but the suspend time will not be counted
		1853	towards C<ev_timer> when a monotonic clock source is used. The real time
		1854	clock advanced as expected, but if it is used as sole clocksource, then a
		1855	long suspend would be detected as a time jump by libev, and timers would
		1856	be adjusted accordingly.
		1857
		1858	I would not be surprised to see different behaviour in different between
		1859	operating systems, OS versions or even different hardware.
		1860
		1861	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1862	time jump in the monotonic clocks and the realtime clock. If the program
		1863	is suspended for a very long time, and monotonic clock sources are in use,
		1864	then you can expect C<ev_timer>s to expire as the full suspension time
		1865	will be counted towards the timers. When no monotonic clock source is in
		1866	use, then libev will again assume a timejump and adjust accordingly.
		1867
		1868	It might be beneficial for this latter case to call C<ev_suspend>
		1869	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1870	deterministic behaviour in this case (you can do nothing against
		1871	C<SIGSTOP>).
		1872
		1873	=head3 Watcher-Specific Functions and Data Members
		1874
468	=over 4	1875	=over 4
469		1876
470	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1877	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
471		1878
472	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1879	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
473		1880
474	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1881	Configure the timer to trigger after C<after> seconds. If C<repeat>
475	C<0.>, then it will automatically be stopped. If it is positive, then the	1882	is C<0.>, then it will automatically be stopped once the timeout is
476	timer will automatically be configured to trigger again C<repeat> seconds	1883	reached. If it is positive, then the timer will automatically be
477	later, again, and again, until stopped manually.	1884	configured to trigger again C<repeat> seconds later, again, and again,
		1885	until stopped manually.
478		1886
479	The timer itself will do a best-effort at avoiding drift, that is, if you	1887	The timer itself will do a best-effort at avoiding drift, that is, if
480	configure a timer to trigger every 10 seconds, then it will trigger at	1888	you configure a timer to trigger every 10 seconds, then it will normally
481	exactly 10 second intervals. If, however, your program cannot keep up with	1889	trigger at exactly 10 second intervals. If, however, your program cannot
482	the timer (because it takes longer than those 10 seconds to do stuff) the	1890	keep up with the timer (because it takes longer than those 10 seconds to
483	timer will not fire more than once per event loop iteration.	1891	do stuff) the timer will not fire more than once per event loop iteration.
484		1892
485	=item ev_timer_again (loop)	1893	=item ev_timer_again (loop, ev_timer *)
486		1894
487	This will act as if the timer timed out and restart it again if it is	1895	This will act as if the timer timed out and restart it again if it is
488	repeating. The exact semantics are:	1896	repeating. The exact semantics are:
489		1897
		1898	If the timer is pending, its pending status is cleared.
		1899
490	If the timer is started but nonrepeating, stop it.	1900	If the timer is started but non-repeating, stop it (as if it timed out).
491		1901
492	If the timer is repeating, either start it if necessary (with the repeat	1902	If the timer is repeating, either start it if necessary (with the
493	value), or reset the running timer to the repeat value.	1903	C<repeat> value), or reset the running timer to the C<repeat> value.
494		1904
495	This sounds a bit complicated, but here is a useful and typical	1905	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
496	example: Imagine you have a tcp connection and you want a so-called idle	1906	usage example.
497	timeout, that is, you want to be called when there have been, say, 60	1907
498	seconds of inactivity on the socket. The easiest way to do this is to	1908	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
499	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	1909
500	time you successfully read or write some data. If you go into an idle	1910	Returns the remaining time until a timer fires. If the timer is active,
501	state where you do not expect data to travel on the socket, you can stop	1911	then this time is relative to the current event loop time, otherwise it's
502	the timer, and again will automatically restart it if need be.	1912	the timeout value currently configured.
		1913
		1914	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		1915	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		1916	will return C<4>. When the timer expires and is restarted, it will return
		1917	roughly C<7> (likely slightly less as callback invocation takes some time,
		1918	too), and so on.
		1919
		1920	=item ev_tstamp repeat [read-write]
		1921
		1922	The current C<repeat> value. Will be used each time the watcher times out
		1923	or C<ev_timer_again> is called, and determines the next timeout (if any),
		1924	which is also when any modifications are taken into account.
503		1925
504	=back	1926	=back
505		1927
		1928	=head3 Examples
		1929
		1930	Example: Create a timer that fires after 60 seconds.
		1931
		1932	static void
		1933	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1934	{
		1935	.. one minute over, w is actually stopped right here
		1936	}
		1937
		1938	ev_timer mytimer;
		1939	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1940	ev_timer_start (loop, &mytimer);
		1941
		1942	Example: Create a timeout timer that times out after 10 seconds of
		1943	inactivity.
		1944
		1945	static void
		1946	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1947	{
		1948	.. ten seconds without any activity
		1949	}
		1950
		1951	ev_timer mytimer;
		1952	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1953	ev_timer_again (&mytimer); /* start timer */
		1954	ev_loop (loop, 0);
		1955
		1956	// and in some piece of code that gets executed on any "activity":
		1957	// reset the timeout to start ticking again at 10 seconds
		1958	ev_timer_again (&mytimer);
		1959
		1960
506	=head2 C<ev_periodic> - to cron or not to cron	1961	=head2 C<ev_periodic> - to cron or not to cron?
507		1962
508	Periodic watchers are also timers of a kind, but they are very versatile	1963	Periodic watchers are also timers of a kind, but they are very versatile
509	(and unfortunately a bit complex).	1964	(and unfortunately a bit complex).
510		1965
511	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1966	Unlike C<ev_timer>, periodic watchers are not based on real time (or
512	but on wallclock time (absolute time). You can tell a periodic watcher	1967	relative time, the physical time that passes) but on wall clock time
513	to trigger "at" some specific point in time. For example, if you tell a	1968	(absolute time, the thing you can read on your calender or clock). The
514	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1969	difference is that wall clock time can run faster or slower than real
515	+ 10.>) and then reset your system clock to the last year, then it will	1970	time, and time jumps are not uncommon (e.g. when you adjust your
516	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	1971	wrist-watch).
517	roughly 10 seconds later and of course not if you reset your system time
518	again).
519		1972
520	They can also be used to implement vastly more complex timers, such as	1973	You can tell a periodic watcher to trigger after some specific point
		1974	in time: for example, if you tell a periodic watcher to trigger "in 10
		1975	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		1976	not a delay) and then reset your system clock to January of the previous
		1977	year, then it will take a year or more to trigger the event (unlike an
		1978	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		1979	it, as it uses a relative timeout).
		1980
		1981	C<ev_periodic> watchers can also be used to implement vastly more complex
521	triggering an event on eahc midnight, local time.	1982	timers, such as triggering an event on each "midnight, local time", or
		1983	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		1984	those cannot react to time jumps.
		1985
		1986	As with timers, the callback is guaranteed to be invoked only when the
		1987	point in time where it is supposed to trigger has passed. If multiple
		1988	timers become ready during the same loop iteration then the ones with
		1989	earlier time-out values are invoked before ones with later time-out values
		1990	(but this is no longer true when a callback calls C<ev_loop> recursively).
		1991
		1992	=head3 Watcher-Specific Functions and Data Members
522		1993
523	=over 4	1994	=over 4
524		1995
525	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1996	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
526		1997
527	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1998	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
528		1999
529	Lots of arguments, lets sort it out... There are basically three modes of	2000	Lots of arguments, let's sort it out... There are basically three modes of
530	operation, and we will explain them from simplest to complex:	2001	operation, and we will explain them from simplest to most complex:
531
532		2002
533	=over 4	2003	=over 4
534		2004
535	=item * absolute timer (interval = reschedule_cb = 0)	2005	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
536		2006
537	In this configuration the watcher triggers an event at the wallclock time	2007	In this configuration the watcher triggers an event after the wall clock
538	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2008	time C<offset> has passed. It will not repeat and will not adjust when a
539	that is, if it is to be run at January 1st 2011 then it will run when the	2009	time jump occurs, that is, if it is to be run at January 1st 2011 then it
540	system time reaches or surpasses this time.	2010	will be stopped and invoked when the system clock reaches or surpasses
		2011	this point in time.
541		2012
542	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2013	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
543		2014
544	In this mode the watcher will always be scheduled to time out at the next	2015	In this mode the watcher will always be scheduled to time out at the next
545	C<at + N * interval> time (for some integer N) and then repeat, regardless	2016	C<offset + N * interval> time (for some integer N, which can also be
546	of any time jumps.	2017	negative) and then repeat, regardless of any time jumps. The C<offset>
		2018	argument is merely an offset into the C<interval> periods.
547		2019
548	This can be used to create timers that do not drift with respect to system	2020	This can be used to create timers that do not drift with respect to the
549	time:	2021	system clock, for example, here is an C<ev_periodic> that triggers each
		2022	hour, on the hour (with respect to UTC):
550		2023
551	ev_periodic_set (&periodic, 0., 3600., 0);	2024	ev_periodic_set (&periodic, 0., 3600., 0);
552		2025
553	This doesn't mean there will always be 3600 seconds in between triggers,	2026	This doesn't mean there will always be 3600 seconds in between triggers,
554	but only that the the callback will be called when the system time shows a	2027	but only that the callback will be called when the system time shows a
555	full hour (UTC), or more correctly, when the system time is evenly divisible	2028	full hour (UTC), or more correctly, when the system time is evenly divisible
556	by 3600.	2029	by 3600.
557		2030
558	Another way to think about it (for the mathematically inclined) is that	2031	Another way to think about it (for the mathematically inclined) is that
559	C<ev_periodic> will try to run the callback in this mode at the next possible	2032	C<ev_periodic> will try to run the callback in this mode at the next possible
560	time where C<time = at (mod interval)>, regardless of any time jumps.	2033	time where C<time = offset (mod interval)>, regardless of any time jumps.
561		2034
		2035	For numerical stability it is preferable that the C<offset> value is near
		2036	C<ev_now ()> (the current time), but there is no range requirement for
		2037	this value, and in fact is often specified as zero.
		2038
		2039	Note also that there is an upper limit to how often a timer can fire (CPU
		2040	speed for example), so if C<interval> is very small then timing stability
		2041	will of course deteriorate. Libev itself tries to be exact to be about one
		2042	millisecond (if the OS supports it and the machine is fast enough).
		2043
562	=item * manual reschedule mode (reschedule_cb = callback)	2044	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
563		2045
564	In this mode the values for C<interval> and C<at> are both being	2046	In this mode the values for C<interval> and C<offset> are both being
565	ignored. Instead, each time the periodic watcher gets scheduled, the	2047	ignored. Instead, each time the periodic watcher gets scheduled, the
566	reschedule callback will be called with the watcher as first, and the	2048	reschedule callback will be called with the watcher as first, and the
567	current time as second argument.	2049	current time as second argument.
568		2050
569	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2051	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
570	ever, or make any event loop modifications>. If you need to stop it,	2052	or make ANY other event loop modifications whatsoever, unless explicitly
571	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by	2053	allowed by documentation here>.
572	starting a prepare watcher).
573		2054
		2055	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2056	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2057	only event loop modification you are allowed to do).
		2058
574	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2059	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
575	ev_tstamp now)>, e.g.:	2060	*w, ev_tstamp now)>, e.g.:
576		2061
		2062	static ev_tstamp
577	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2063	my_rescheduler (ev_periodic *w, ev_tstamp now)
578	{	2064	{
579	return now + 60.;	2065	return now + 60.;
580	}	2066	}
581		2067
582	It must return the next time to trigger, based on the passed time value	2068	It must return the next time to trigger, based on the passed time value
583	(that is, the lowest time value larger than to the second argument). It	2069	(that is, the lowest time value larger than to the second argument). It
584	will usually be called just before the callback will be triggered, but	2070	will usually be called just before the callback will be triggered, but
585	might be called at other times, too.	2071	might be called at other times, too.
586		2072
587	NOTE: I<< This callback must always return a time that is later than the	2073	NOTE: I<< This callback must always return a time that is higher than or
588	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	2074	equal to the passed C<now> value >>.
589		2075
590	This can be used to create very complex timers, such as a timer that	2076	This can be used to create very complex timers, such as a timer that
591	triggers on each midnight, local time. To do this, you would calculate the	2077	triggers on "next midnight, local time". To do this, you would calculate the
592	next midnight after C<now> and return the timestamp value for this. How	2078	next midnight after C<now> and return the timestamp value for this. How
593	you do this is, again, up to you (but it is not trivial, which is the main	2079	you do this is, again, up to you (but it is not trivial, which is the main
594	reason I omitted it as an example).	2080	reason I omitted it as an example).
595		2081
596	=back	2082	=back
…		…
600	Simply stops and restarts the periodic watcher again. This is only useful	2086	Simply stops and restarts the periodic watcher again. This is only useful
601	when you changed some parameters or the reschedule callback would return	2087	when you changed some parameters or the reschedule callback would return
602	a different time than the last time it was called (e.g. in a crond like	2088	a different time than the last time it was called (e.g. in a crond like
603	program when the crontabs have changed).	2089	program when the crontabs have changed).
604		2090
		2091	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2092
		2093	When active, returns the absolute time that the watcher is supposed
		2094	to trigger next. This is not the same as the C<offset> argument to
		2095	C<ev_periodic_set>, but indeed works even in interval and manual
		2096	rescheduling modes.
		2097
		2098	=item ev_tstamp offset [read-write]
		2099
		2100	When repeating, this contains the offset value, otherwise this is the
		2101	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2102	although libev might modify this value for better numerical stability).
		2103
		2104	Can be modified any time, but changes only take effect when the periodic
		2105	timer fires or C<ev_periodic_again> is being called.
		2106
		2107	=item ev_tstamp interval [read-write]
		2108
		2109	The current interval value. Can be modified any time, but changes only
		2110	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2111	called.
		2112
		2113	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2114
		2115	The current reschedule callback, or C<0>, if this functionality is
		2116	switched off. Can be changed any time, but changes only take effect when
		2117	the periodic timer fires or C<ev_periodic_again> is being called.
		2118
605	=back	2119	=back
606		2120
		2121	=head3 Examples
		2122
		2123	Example: Call a callback every hour, or, more precisely, whenever the
		2124	system time is divisible by 3600. The callback invocation times have
		2125	potentially a lot of jitter, but good long-term stability.
		2126
		2127	static void
		2128	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
		2129	{
		2130	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2131	}
		2132
		2133	ev_periodic hourly_tick;
		2134	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2135	ev_periodic_start (loop, &hourly_tick);
		2136
		2137	Example: The same as above, but use a reschedule callback to do it:
		2138
		2139	#include <math.h>
		2140
		2141	static ev_tstamp
		2142	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2143	{
		2144	return now + (3600. - fmod (now, 3600.));
		2145	}
		2146
		2147	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2148
		2149	Example: Call a callback every hour, starting now:
		2150
		2151	ev_periodic hourly_tick;
		2152	ev_periodic_init (&hourly_tick, clock_cb,
		2153	fmod (ev_now (loop), 3600.), 3600., 0);
		2154	ev_periodic_start (loop, &hourly_tick);
		2155
		2156
607	=head2 C<ev_signal> - signal me when a signal gets signalled	2157	=head2 C<ev_signal> - signal me when a signal gets signalled!
608		2158
609	Signal watchers will trigger an event when the process receives a specific	2159	Signal watchers will trigger an event when the process receives a specific
610	signal one or more times. Even though signals are very asynchronous, libev	2160	signal one or more times. Even though signals are very asynchronous, libev
611	will try it's best to deliver signals synchronously, i.e. as part of the	2161	will try it's best to deliver signals synchronously, i.e. as part of the
612	normal event processing, like any other event.	2162	normal event processing, like any other event.
613		2163
		2164	If you want signals to be delivered truly asynchronously, just use
		2165	C<sigaction> as you would do without libev and forget about sharing
		2166	the signal. You can even use C<ev_async> from a signal handler to
		2167	synchronously wake up an event loop.
		2168
614	You can configure as many watchers as you like per signal. Only when the	2169	You can configure as many watchers as you like for the same signal, but
		2170	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2171	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2172	C<SIGINT> in both the default loop and another loop at the same time. At
		2173	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2174
615	first watcher gets started will libev actually register a signal watcher	2175	When the first watcher gets started will libev actually register something
616	with the kernel (thus it coexists with your own signal handlers as long	2176	with the kernel (thus it coexists with your own signal handlers as long as
617	as you don't register any with libev). Similarly, when the last signal	2177	you don't register any with libev for the same signal).
618	watcher for a signal is stopped libev will reset the signal handler to	2178
619	SIG_DFL (regardless of what it was set to before).	2179	If possible and supported, libev will install its handlers with
		2180	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2181	not be unduly interrupted. If you have a problem with system calls getting
		2182	interrupted by signals you can block all signals in an C<ev_check> watcher
		2183	and unblock them in an C<ev_prepare> watcher.
		2184
		2185	=head3 The special problem of inheritance over fork/execve/pthread_create
		2186
		2187	Both the signal mask (C<sigprocmask>) and the signal disposition
		2188	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2189	stopping it again), that is, libev might or might not block the signal,
		2190	and might or might not set or restore the installed signal handler.
		2191
		2192	While this does not matter for the signal disposition (libev never
		2193	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2194	C<execve>), this matters for the signal mask: many programs do not expect
		2195	certain signals to be blocked.
		2196
		2197	This means that before calling C<exec> (from the child) you should reset
		2198	the signal mask to whatever "default" you expect (all clear is a good
		2199	choice usually).
		2200
		2201	The simplest way to ensure that the signal mask is reset in the child is
		2202	to install a fork handler with C<pthread_atfork> that resets it. That will
		2203	catch fork calls done by libraries (such as the libc) as well.
		2204
		2205	In current versions of libev, the signal will not be blocked indefinitely
		2206	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2207	the window of opportunity for problems, it will not go away, as libev
		2208	I<has> to modify the signal mask, at least temporarily.
		2209
		2210	So I can't stress this enough: I<If you do not reset your signal mask when
		2211	you expect it to be empty, you have a race condition in your code>. This
		2212	is not a libev-specific thing, this is true for most event libraries.
		2213
		2214	=head3 Watcher-Specific Functions and Data Members
620		2215
621	=over 4	2216	=over 4
622		2217
623	=item ev_signal_init (ev_signal *, callback, int signum)	2218	=item ev_signal_init (ev_signal *, callback, int signum)
624		2219
625	=item ev_signal_set (ev_signal *, int signum)	2220	=item ev_signal_set (ev_signal *, int signum)
626		2221
627	Configures the watcher to trigger on the given signal number (usually one	2222	Configures the watcher to trigger on the given signal number (usually one
628	of the C<SIGxxx> constants).	2223	of the C<SIGxxx> constants).
629		2224
		2225	=item int signum [read-only]
		2226
		2227	The signal the watcher watches out for.
		2228
630	=back	2229	=back
631		2230
		2231	=head3 Examples
		2232
		2233	Example: Try to exit cleanly on SIGINT.
		2234
		2235	static void
		2236	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2237	{
		2238	ev_unloop (loop, EVUNLOOP_ALL);
		2239	}
		2240
		2241	ev_signal signal_watcher;
		2242	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2243	ev_signal_start (loop, &signal_watcher);
		2244
		2245
632	=head2 C<ev_child> - wait for pid status changes	2246	=head2 C<ev_child> - watch out for process status changes
633		2247
634	Child watchers trigger when your process receives a SIGCHLD in response to	2248	Child watchers trigger when your process receives a SIGCHLD in response to
635	some child status changes (most typically when a child of yours dies).	2249	some child status changes (most typically when a child of yours dies or
		2250	exits). It is permissible to install a child watcher I<after> the child
		2251	has been forked (which implies it might have already exited), as long
		2252	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2253	forking and then immediately registering a watcher for the child is fine,
		2254	but forking and registering a watcher a few event loop iterations later or
		2255	in the next callback invocation is not.
		2256
		2257	Only the default event loop is capable of handling signals, and therefore
		2258	you can only register child watchers in the default event loop.
		2259
		2260	Due to some design glitches inside libev, child watchers will always be
		2261	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2262	libev)
		2263
		2264	=head3 Process Interaction
		2265
		2266	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2267	initialised. This is necessary to guarantee proper behaviour even if the
		2268	first child watcher is started after the child exits. The occurrence
		2269	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2270	synchronously as part of the event loop processing. Libev always reaps all
		2271	children, even ones not watched.
		2272
		2273	=head3 Overriding the Built-In Processing
		2274
		2275	Libev offers no special support for overriding the built-in child
		2276	processing, but if your application collides with libev's default child
		2277	handler, you can override it easily by installing your own handler for
		2278	C<SIGCHLD> after initialising the default loop, and making sure the
		2279	default loop never gets destroyed. You are encouraged, however, to use an
		2280	event-based approach to child reaping and thus use libev's support for
		2281	that, so other libev users can use C<ev_child> watchers freely.
		2282
		2283	=head3 Stopping the Child Watcher
		2284
		2285	Currently, the child watcher never gets stopped, even when the
		2286	child terminates, so normally one needs to stop the watcher in the
		2287	callback. Future versions of libev might stop the watcher automatically
		2288	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2289	problem).
		2290
		2291	=head3 Watcher-Specific Functions and Data Members
636		2292
637	=over 4	2293	=over 4
638		2294
639	=item ev_child_init (ev_child *, callback, int pid)	2295	=item ev_child_init (ev_child *, callback, int pid, int trace)
640		2296
641	=item ev_child_set (ev_child *, int pid)	2297	=item ev_child_set (ev_child *, int pid, int trace)
642		2298
643	Configures the watcher to wait for status changes of process C<pid> (or	2299	Configures the watcher to wait for status changes of process C<pid> (or
644	I<any> process if C<pid> is specified as C<0>). The callback can look	2300	I<any> process if C<pid> is specified as C<0>). The callback can look
645	at the C<rstatus> member of the C<ev_child> watcher structure to see	2301	at the C<rstatus> member of the C<ev_child> watcher structure to see
646	the status word (use the macros from C<sys/wait.h> and see your systems	2302	the status word (use the macros from C<sys/wait.h> and see your systems
647	C<waitpid> documentation). The C<rpid> member contains the pid of the	2303	C<waitpid> documentation). The C<rpid> member contains the pid of the
648	process causing the status change.	2304	process causing the status change. C<trace> must be either C<0> (only
		2305	activate the watcher when the process terminates) or C<1> (additionally
		2306	activate the watcher when the process is stopped or continued).
		2307
		2308	=item int pid [read-only]
		2309
		2310	The process id this watcher watches out for, or C<0>, meaning any process id.
		2311
		2312	=item int rpid [read-write]
		2313
		2314	The process id that detected a status change.
		2315
		2316	=item int rstatus [read-write]
		2317
		2318	The process exit/trace status caused by C<rpid> (see your systems
		2319	C<waitpid> and C<sys/wait.h> documentation for details).
649		2320
650	=back	2321	=back
651		2322
		2323	=head3 Examples
		2324
		2325	Example: C<fork()> a new process and install a child handler to wait for
		2326	its completion.
		2327
		2328	ev_child cw;
		2329
		2330	static void
		2331	child_cb (EV_P_ ev_child *w, int revents)
		2332	{
		2333	ev_child_stop (EV_A_ w);
		2334	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2335	}
		2336
		2337	pid_t pid = fork ();
		2338
		2339	if (pid < 0)
		2340	// error
		2341	else if (pid == 0)
		2342	{
		2343	// the forked child executes here
		2344	exit (1);
		2345	}
		2346	else
		2347	{
		2348	ev_child_init (&cw, child_cb, pid, 0);
		2349	ev_child_start (EV_DEFAULT_ &cw);
		2350	}
		2351
		2352
		2353	=head2 C<ev_stat> - did the file attributes just change?
		2354
		2355	This watches a file system path for attribute changes. That is, it calls
		2356	C<stat> on that path in regular intervals (or when the OS says it changed)
		2357	and sees if it changed compared to the last time, invoking the callback if
		2358	it did.
		2359
		2360	The path does not need to exist: changing from "path exists" to "path does
		2361	not exist" is a status change like any other. The condition "path does not
		2362	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2363	C<st_nlink> field being zero (which is otherwise always forced to be at
		2364	least one) and all the other fields of the stat buffer having unspecified
		2365	contents.
		2366
		2367	The path I<must not> end in a slash or contain special components such as
		2368	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2369	your working directory changes, then the behaviour is undefined.
		2370
		2371	Since there is no portable change notification interface available, the
		2372	portable implementation simply calls C<stat(2)> regularly on the path
		2373	to see if it changed somehow. You can specify a recommended polling
		2374	interval for this case. If you specify a polling interval of C<0> (highly
		2375	recommended!) then a I<suitable, unspecified default> value will be used
		2376	(which you can expect to be around five seconds, although this might
		2377	change dynamically). Libev will also impose a minimum interval which is
		2378	currently around C<0.1>, but that's usually overkill.
		2379
		2380	This watcher type is not meant for massive numbers of stat watchers,
		2381	as even with OS-supported change notifications, this can be
		2382	resource-intensive.
		2383
		2384	At the time of this writing, the only OS-specific interface implemented
		2385	is the Linux inotify interface (implementing kqueue support is left as an
		2386	exercise for the reader. Note, however, that the author sees no way of
		2387	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2388
		2389	=head3 ABI Issues (Largefile Support)
		2390
		2391	Libev by default (unless the user overrides this) uses the default
		2392	compilation environment, which means that on systems with large file
		2393	support disabled by default, you get the 32 bit version of the stat
		2394	structure. When using the library from programs that change the ABI to
		2395	use 64 bit file offsets the programs will fail. In that case you have to
		2396	compile libev with the same flags to get binary compatibility. This is
		2397	obviously the case with any flags that change the ABI, but the problem is
		2398	most noticeably displayed with ev_stat and large file support.
		2399
		2400	The solution for this is to lobby your distribution maker to make large
		2401	file interfaces available by default (as e.g. FreeBSD does) and not
		2402	optional. Libev cannot simply switch on large file support because it has
		2403	to exchange stat structures with application programs compiled using the
		2404	default compilation environment.
		2405
		2406	=head3 Inotify and Kqueue
		2407
		2408	When C<inotify (7)> support has been compiled into libev and present at
		2409	runtime, it will be used to speed up change detection where possible. The
		2410	inotify descriptor will be created lazily when the first C<ev_stat>
		2411	watcher is being started.
		2412
		2413	Inotify presence does not change the semantics of C<ev_stat> watchers
		2414	except that changes might be detected earlier, and in some cases, to avoid
		2415	making regular C<stat> calls. Even in the presence of inotify support
		2416	there are many cases where libev has to resort to regular C<stat> polling,
		2417	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2418	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2419	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2420	xfs are fully working) libev usually gets away without polling.
		2421
		2422	There is no support for kqueue, as apparently it cannot be used to
		2423	implement this functionality, due to the requirement of having a file
		2424	descriptor open on the object at all times, and detecting renames, unlinks
		2425	etc. is difficult.
		2426
		2427	=head3 C<stat ()> is a synchronous operation
		2428
		2429	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2430	the process. The exception are C<ev_stat> watchers - those call C<stat
		2431	()>, which is a synchronous operation.
		2432
		2433	For local paths, this usually doesn't matter: unless the system is very
		2434	busy or the intervals between stat's are large, a stat call will be fast,
		2435	as the path data is usually in memory already (except when starting the
		2436	watcher).
		2437
		2438	For networked file systems, calling C<stat ()> can block an indefinite
		2439	time due to network issues, and even under good conditions, a stat call
		2440	often takes multiple milliseconds.
		2441
		2442	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2443	paths, although this is fully supported by libev.
		2444
		2445	=head3 The special problem of stat time resolution
		2446
		2447	The C<stat ()> system call only supports full-second resolution portably,
		2448	and even on systems where the resolution is higher, most file systems
		2449	still only support whole seconds.
		2450
		2451	That means that, if the time is the only thing that changes, you can
		2452	easily miss updates: on the first update, C<ev_stat> detects a change and
		2453	calls your callback, which does something. When there is another update
		2454	within the same second, C<ev_stat> will be unable to detect unless the
		2455	stat data does change in other ways (e.g. file size).
		2456
		2457	The solution to this is to delay acting on a change for slightly more
		2458	than a second (or till slightly after the next full second boundary), using
		2459	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2460	ev_timer_again (loop, w)>).
		2461
		2462	The C<.02> offset is added to work around small timing inconsistencies
		2463	of some operating systems (where the second counter of the current time
		2464	might be be delayed. One such system is the Linux kernel, where a call to
		2465	C<gettimeofday> might return a timestamp with a full second later than
		2466	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2467	update file times then there will be a small window where the kernel uses
		2468	the previous second to update file times but libev might already execute
		2469	the timer callback).
		2470
		2471	=head3 Watcher-Specific Functions and Data Members
		2472
		2473	=over 4
		2474
		2475	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2476
		2477	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2478
		2479	Configures the watcher to wait for status changes of the given
		2480	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2481	be detected and should normally be specified as C<0> to let libev choose
		2482	a suitable value. The memory pointed to by C<path> must point to the same
		2483	path for as long as the watcher is active.
		2484
		2485	The callback will receive an C<EV_STAT> event when a change was detected,
		2486	relative to the attributes at the time the watcher was started (or the
		2487	last change was detected).
		2488
		2489	=item ev_stat_stat (loop, ev_stat *)
		2490
		2491	Updates the stat buffer immediately with new values. If you change the
		2492	watched path in your callback, you could call this function to avoid
		2493	detecting this change (while introducing a race condition if you are not
		2494	the only one changing the path). Can also be useful simply to find out the
		2495	new values.
		2496
		2497	=item ev_statdata attr [read-only]
		2498
		2499	The most-recently detected attributes of the file. Although the type is
		2500	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2501	suitable for your system, but you can only rely on the POSIX-standardised
		2502	members to be present. If the C<st_nlink> member is C<0>, then there was
		2503	some error while C<stat>ing the file.
		2504
		2505	=item ev_statdata prev [read-only]
		2506
		2507	The previous attributes of the file. The callback gets invoked whenever
		2508	C<prev> != C<attr>, or, more precisely, one or more of these members
		2509	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2510	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2511
		2512	=item ev_tstamp interval [read-only]
		2513
		2514	The specified interval.
		2515
		2516	=item const char *path [read-only]
		2517
		2518	The file system path that is being watched.
		2519
		2520	=back
		2521
		2522	=head3 Examples
		2523
		2524	Example: Watch C</etc/passwd> for attribute changes.
		2525
		2526	static void
		2527	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2528	{
		2529	/* /etc/passwd changed in some way */
		2530	if (w->attr.st_nlink)
		2531	{
		2532	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2533	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2534	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2535	}
		2536	else
		2537	/* you shalt not abuse printf for puts */
		2538	puts ("wow, /etc/passwd is not there, expect problems. "
		2539	"if this is windows, they already arrived\n");
		2540	}
		2541
		2542	...
		2543	ev_stat passwd;
		2544
		2545	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2546	ev_stat_start (loop, &passwd);
		2547
		2548	Example: Like above, but additionally use a one-second delay so we do not
		2549	miss updates (however, frequent updates will delay processing, too, so
		2550	one might do the work both on C<ev_stat> callback invocation I<and> on
		2551	C<ev_timer> callback invocation).
		2552
		2553	static ev_stat passwd;
		2554	static ev_timer timer;
		2555
		2556	static void
		2557	timer_cb (EV_P_ ev_timer *w, int revents)
		2558	{
		2559	ev_timer_stop (EV_A_ w);
		2560
		2561	/* now it's one second after the most recent passwd change */
		2562	}
		2563
		2564	static void
		2565	stat_cb (EV_P_ ev_stat *w, int revents)
		2566	{
		2567	/* reset the one-second timer */
		2568	ev_timer_again (EV_A_ &timer);
		2569	}
		2570
		2571	...
		2572	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2573	ev_stat_start (loop, &passwd);
		2574	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2575
		2576
652	=head2 C<ev_idle> - when you've got nothing better to do	2577	=head2 C<ev_idle> - when you've got nothing better to do...
653		2578
654	Idle watchers trigger events when there are no other events are pending	2579	Idle watchers trigger events when no other events of the same or higher
655	(prepare, check and other idle watchers do not count). That is, as long	2580	priority are pending (prepare, check and other idle watchers do not count
656	as your process is busy handling sockets or timeouts (or even signals,	2581	as receiving "events").
657	imagine) it will not be triggered. But when your process is idle all idle	2582
658	watchers are being called again and again, once per event loop iteration -	2583	That is, as long as your process is busy handling sockets or timeouts
		2584	(or even signals, imagine) of the same or higher priority it will not be
		2585	triggered. But when your process is idle (or only lower-priority watchers
		2586	are pending), the idle watchers are being called once per event loop
659	until stopped, that is, or your process receives more events and becomes	2587	iteration - until stopped, that is, or your process receives more events
660	busy.	2588	and becomes busy again with higher priority stuff.
661		2589
662	The most noteworthy effect is that as long as any idle watchers are	2590	The most noteworthy effect is that as long as any idle watchers are
663	active, the process will not block when waiting for new events.	2591	active, the process will not block when waiting for new events.
664		2592
665	Apart from keeping your process non-blocking (which is a useful	2593	Apart from keeping your process non-blocking (which is a useful
666	effect on its own sometimes), idle watchers are a good place to do	2594	effect on its own sometimes), idle watchers are a good place to do
667	"pseudo-background processing", or delay processing stuff to after the	2595	"pseudo-background processing", or delay processing stuff to after the
668	event loop has handled all outstanding events.	2596	event loop has handled all outstanding events.
669		2597
		2598	=head3 Watcher-Specific Functions and Data Members
		2599
670	=over 4	2600	=over 4
671		2601
672	=item ev_idle_init (ev_signal *, callback)	2602	=item ev_idle_init (ev_idle *, callback)
673		2603
674	Initialises and configures the idle watcher - it has no parameters of any	2604	Initialises and configures the idle watcher - it has no parameters of any
675	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2605	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
676	believe me.	2606	believe me.
677		2607
678	=back	2608	=back
679		2609
		2610	=head3 Examples
		2611
		2612	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2613	callback, free it. Also, use no error checking, as usual.
		2614
		2615	static void
		2616	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2617	{
		2618	free (w);
		2619	// now do something you wanted to do when the program has
		2620	// no longer anything immediate to do.
		2621	}
		2622
		2623	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2624	ev_idle_init (idle_watcher, idle_cb);
		2625	ev_idle_start (loop, idle_watcher);
		2626
		2627
680	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2628	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
681		2629
682	Prepare and check watchers are usually (but not always) used in tandem:	2630	Prepare and check watchers are usually (but not always) used in pairs:
683	prepare watchers get invoked before the process blocks and check watchers	2631	prepare watchers get invoked before the process blocks and check watchers
684	afterwards.	2632	afterwards.
685		2633
		2634	You I<must not> call C<ev_loop> or similar functions that enter
		2635	the current event loop from either C<ev_prepare> or C<ev_check>
		2636	watchers. Other loops than the current one are fine, however. The
		2637	rationale behind this is that you do not need to check for recursion in
		2638	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2639	C<ev_check> so if you have one watcher of each kind they will always be
		2640	called in pairs bracketing the blocking call.
		2641
686	Their main purpose is to integrate other event mechanisms into libev. This	2642	Their main purpose is to integrate other event mechanisms into libev and
687	could be used, for example, to track variable changes, implement your own	2643	their use is somewhat advanced. They could be used, for example, to track
688	watchers, integrate net-snmp or a coroutine library and lots more.	2644	variable changes, implement your own watchers, integrate net-snmp or a
		2645	coroutine library and lots more. They are also occasionally useful if
		2646	you cache some data and want to flush it before blocking (for example,
		2647	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2648	watcher).
689		2649
690	This is done by examining in each prepare call which file descriptors need	2650	This is done by examining in each prepare call which file descriptors
691	to be watched by the other library, registering C<ev_io> watchers for	2651	need to be watched by the other library, registering C<ev_io> watchers
692	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2652	for them and starting an C<ev_timer> watcher for any timeouts (many
693	provide just this functionality). Then, in the check watcher you check for	2653	libraries provide exactly this functionality). Then, in the check watcher,
694	any events that occured (by checking the pending status of all watchers	2654	you check for any events that occurred (by checking the pending status
695	and stopping them) and call back into the library. The I/O and timer	2655	of all watchers and stopping them) and call back into the library. The
696	callbacks will never actually be called (but must be valid nevertheless,	2656	I/O and timer callbacks will never actually be called (but must be valid
697	because you never know, you know?).	2657	nevertheless, because you never know, you know?).
698		2658
699	As another example, the Perl Coro module uses these hooks to integrate	2659	As another example, the Perl Coro module uses these hooks to integrate
700	coroutines into libev programs, by yielding to other active coroutines	2660	coroutines into libev programs, by yielding to other active coroutines
701	during each prepare and only letting the process block if no coroutines	2661	during each prepare and only letting the process block if no coroutines
702	are ready to run (it's actually more complicated: it only runs coroutines	2662	are ready to run (it's actually more complicated: it only runs coroutines
703	with priority higher than or equal to the event loop and one coroutine	2663	with priority higher than or equal to the event loop and one coroutine
704	of lower priority, but only once, using idle watchers to keep the event	2664	of lower priority, but only once, using idle watchers to keep the event
705	loop from blocking if lower-priority coroutines are active, thus mapping	2665	loop from blocking if lower-priority coroutines are active, thus mapping
706	low-priority coroutines to idle/background tasks).	2666	low-priority coroutines to idle/background tasks).
707		2667
		2668	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2669	priority, to ensure that they are being run before any other watchers
		2670	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2671
		2672	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2673	activate ("feed") events into libev. While libev fully supports this, they
		2674	might get executed before other C<ev_check> watchers did their job. As
		2675	C<ev_check> watchers are often used to embed other (non-libev) event
		2676	loops those other event loops might be in an unusable state until their
		2677	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2678	others).
		2679
		2680	=head3 Watcher-Specific Functions and Data Members
		2681
708	=over 4	2682	=over 4
709		2683
710	=item ev_prepare_init (ev_prepare *, callback)	2684	=item ev_prepare_init (ev_prepare *, callback)
711		2685
712	=item ev_check_init (ev_check *, callback)	2686	=item ev_check_init (ev_check *, callback)
713		2687
714	Initialises and configures the prepare or check watcher - they have no	2688	Initialises and configures the prepare or check watcher - they have no
715	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2689	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
716	macros, but using them is utterly, utterly and completely pointless.	2690	macros, but using them is utterly, utterly, utterly and completely
		2691	pointless.
717		2692
718	=back	2693	=back
719		2694
		2695	=head3 Examples
		2696
		2697	There are a number of principal ways to embed other event loops or modules
		2698	into libev. Here are some ideas on how to include libadns into libev
		2699	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2700	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2701	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2702	Glib event loop).
		2703
		2704	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2705	and in a check watcher, destroy them and call into libadns. What follows
		2706	is pseudo-code only of course. This requires you to either use a low
		2707	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2708	the callbacks for the IO/timeout watchers might not have been called yet.
		2709
		2710	static ev_io iow [nfd];
		2711	static ev_timer tw;
		2712
		2713	static void
		2714	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2715	{
		2716	}
		2717
		2718	// create io watchers for each fd and a timer before blocking
		2719	static void
		2720	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2721	{
		2722	int timeout = 3600000;
		2723	struct pollfd fds [nfd];
		2724	// actual code will need to loop here and realloc etc.
		2725	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2726
		2727	/* the callback is illegal, but won't be called as we stop during check */
		2728	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		2729	ev_timer_start (loop, &tw);
		2730
		2731	// create one ev_io per pollfd
		2732	for (int i = 0; i < nfd; ++i)
		2733	{
		2734	ev_io_init (iow + i, io_cb, fds [i].fd,
		2735	((fds [i].events & POLLIN ? EV_READ : 0)
		2736	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2737
		2738	fds [i].revents = 0;
		2739	ev_io_start (loop, iow + i);
		2740	}
		2741	}
		2742
		2743	// stop all watchers after blocking
		2744	static void
		2745	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2746	{
		2747	ev_timer_stop (loop, &tw);
		2748
		2749	for (int i = 0; i < nfd; ++i)
		2750	{
		2751	// set the relevant poll flags
		2752	// could also call adns_processreadable etc. here
		2753	struct pollfd *fd = fds + i;
		2754	int revents = ev_clear_pending (iow + i);
		2755	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2756	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2757
		2758	// now stop the watcher
		2759	ev_io_stop (loop, iow + i);
		2760	}
		2761
		2762	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2763	}
		2764
		2765	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2766	in the prepare watcher and would dispose of the check watcher.
		2767
		2768	Method 3: If the module to be embedded supports explicit event
		2769	notification (libadns does), you can also make use of the actual watcher
		2770	callbacks, and only destroy/create the watchers in the prepare watcher.
		2771
		2772	static void
		2773	timer_cb (EV_P_ ev_timer *w, int revents)
		2774	{
		2775	adns_state ads = (adns_state)w->data;
		2776	update_now (EV_A);
		2777
		2778	adns_processtimeouts (ads, &tv_now);
		2779	}
		2780
		2781	static void
		2782	io_cb (EV_P_ ev_io *w, int revents)
		2783	{
		2784	adns_state ads = (adns_state)w->data;
		2785	update_now (EV_A);
		2786
		2787	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2788	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2789	}
		2790
		2791	// do not ever call adns_afterpoll
		2792
		2793	Method 4: Do not use a prepare or check watcher because the module you
		2794	want to embed is not flexible enough to support it. Instead, you can
		2795	override their poll function. The drawback with this solution is that the
		2796	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2797	this approach, effectively embedding EV as a client into the horrible
		2798	libglib event loop.
		2799
		2800	static gint
		2801	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2802	{
		2803	int got_events = 0;
		2804
		2805	for (n = 0; n < nfds; ++n)
		2806	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2807
		2808	if (timeout >= 0)
		2809	// create/start timer
		2810
		2811	// poll
		2812	ev_loop (EV_A_ 0);
		2813
		2814	// stop timer again
		2815	if (timeout >= 0)
		2816	ev_timer_stop (EV_A_ &to);
		2817
		2818	// stop io watchers again - their callbacks should have set
		2819	for (n = 0; n < nfds; ++n)
		2820	ev_io_stop (EV_A_ iow [n]);
		2821
		2822	return got_events;
		2823	}
		2824
		2825
		2826	=head2 C<ev_embed> - when one backend isn't enough...
		2827
		2828	This is a rather advanced watcher type that lets you embed one event loop
		2829	into another (currently only C<ev_io> events are supported in the embedded
		2830	loop, other types of watchers might be handled in a delayed or incorrect
		2831	fashion and must not be used).
		2832
		2833	There are primarily two reasons you would want that: work around bugs and
		2834	prioritise I/O.
		2835
		2836	As an example for a bug workaround, the kqueue backend might only support
		2837	sockets on some platform, so it is unusable as generic backend, but you
		2838	still want to make use of it because you have many sockets and it scales
		2839	so nicely. In this case, you would create a kqueue-based loop and embed
		2840	it into your default loop (which might use e.g. poll). Overall operation
		2841	will be a bit slower because first libev has to call C<poll> and then
		2842	C<kevent>, but at least you can use both mechanisms for what they are
		2843	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2844
		2845	As for prioritising I/O: under rare circumstances you have the case where
		2846	some fds have to be watched and handled very quickly (with low latency),
		2847	and even priorities and idle watchers might have too much overhead. In
		2848	this case you would put all the high priority stuff in one loop and all
		2849	the rest in a second one, and embed the second one in the first.
		2850
		2851	As long as the watcher is active, the callback will be invoked every
		2852	time there might be events pending in the embedded loop. The callback
		2853	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		2854	sweep and invoke their callbacks (the callback doesn't need to invoke the
		2855	C<ev_embed_sweep> function directly, it could also start an idle watcher
		2856	to give the embedded loop strictly lower priority for example).
		2857
		2858	You can also set the callback to C<0>, in which case the embed watcher
		2859	will automatically execute the embedded loop sweep whenever necessary.
		2860
		2861	Fork detection will be handled transparently while the C<ev_embed> watcher
		2862	is active, i.e., the embedded loop will automatically be forked when the
		2863	embedding loop forks. In other cases, the user is responsible for calling
		2864	C<ev_loop_fork> on the embedded loop.
		2865
		2866	Unfortunately, not all backends are embeddable: only the ones returned by
		2867	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2868	portable one.
		2869
		2870	So when you want to use this feature you will always have to be prepared
		2871	that you cannot get an embeddable loop. The recommended way to get around
		2872	this is to have a separate variables for your embeddable loop, try to
		2873	create it, and if that fails, use the normal loop for everything.
		2874
		2875	=head3 C<ev_embed> and fork
		2876
		2877	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2878	automatically be applied to the embedded loop as well, so no special
		2879	fork handling is required in that case. When the watcher is not running,
		2880	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2881	as applicable.
		2882
		2883	=head3 Watcher-Specific Functions and Data Members
		2884
		2885	=over 4
		2886
		2887	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2888
		2889	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2890
		2891	Configures the watcher to embed the given loop, which must be
		2892	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2893	invoked automatically, otherwise it is the responsibility of the callback
		2894	to invoke it (it will continue to be called until the sweep has been done,
		2895	if you do not want that, you need to temporarily stop the embed watcher).
		2896
		2897	=item ev_embed_sweep (loop, ev_embed *)
		2898
		2899	Make a single, non-blocking sweep over the embedded loop. This works
		2900	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2901	appropriate way for embedded loops.
		2902
		2903	=item struct ev_loop *other [read-only]
		2904
		2905	The embedded event loop.
		2906
		2907	=back
		2908
		2909	=head3 Examples
		2910
		2911	Example: Try to get an embeddable event loop and embed it into the default
		2912	event loop. If that is not possible, use the default loop. The default
		2913	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2914	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2915	used).
		2916
		2917	struct ev_loop *loop_hi = ev_default_init (0);
		2918	struct ev_loop *loop_lo = 0;
		2919	ev_embed embed;
		2920
		2921	// see if there is a chance of getting one that works
		2922	// (remember that a flags value of 0 means autodetection)
		2923	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2924	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2925	: 0;
		2926
		2927	// if we got one, then embed it, otherwise default to loop_hi
		2928	if (loop_lo)
		2929	{
		2930	ev_embed_init (&embed, 0, loop_lo);
		2931	ev_embed_start (loop_hi, &embed);
		2932	}
		2933	else
		2934	loop_lo = loop_hi;
		2935
		2936	Example: Check if kqueue is available but not recommended and create
		2937	a kqueue backend for use with sockets (which usually work with any
		2938	kqueue implementation). Store the kqueue/socket-only event loop in
		2939	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2940
		2941	struct ev_loop *loop = ev_default_init (0);
		2942	struct ev_loop *loop_socket = 0;
		2943	ev_embed embed;
		2944
		2945	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2946	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2947	{
		2948	ev_embed_init (&embed, 0, loop_socket);
		2949	ev_embed_start (loop, &embed);
		2950	}
		2951
		2952	if (!loop_socket)
		2953	loop_socket = loop;
		2954
		2955	// now use loop_socket for all sockets, and loop for everything else
		2956
		2957
		2958	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2959
		2960	Fork watchers are called when a C<fork ()> was detected (usually because
		2961	whoever is a good citizen cared to tell libev about it by calling
		2962	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2963	event loop blocks next and before C<ev_check> watchers are being called,
		2964	and only in the child after the fork. If whoever good citizen calling
		2965	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2966	handlers will be invoked, too, of course.
		2967
		2968	=head3 The special problem of life after fork - how is it possible?
		2969
		2970	Most uses of C<fork()> consist of forking, then some simple calls to set
		2971	up/change the process environment, followed by a call to C<exec()>. This
		2972	sequence should be handled by libev without any problems.
		2973
		2974	This changes when the application actually wants to do event handling
		2975	in the child, or both parent in child, in effect "continuing" after the
		2976	fork.
		2977
		2978	The default mode of operation (for libev, with application help to detect
		2979	forks) is to duplicate all the state in the child, as would be expected
		2980	when I<either> the parent I<or> the child process continues.
		2981
		2982	When both processes want to continue using libev, then this is usually the
		2983	wrong result. In that case, usually one process (typically the parent) is
		2984	supposed to continue with all watchers in place as before, while the other
		2985	process typically wants to start fresh, i.e. without any active watchers.
		2986
		2987	The cleanest and most efficient way to achieve that with libev is to
		2988	simply create a new event loop, which of course will be "empty", and
		2989	use that for new watchers. This has the advantage of not touching more
		2990	memory than necessary, and thus avoiding the copy-on-write, and the
		2991	disadvantage of having to use multiple event loops (which do not support
		2992	signal watchers).
		2993
		2994	When this is not possible, or you want to use the default loop for
		2995	other reasons, then in the process that wants to start "fresh", call
		2996	C<ev_default_destroy ()> followed by C<ev_default_loop (...)>. Destroying
		2997	the default loop will "orphan" (not stop) all registered watchers, so you
		2998	have to be careful not to execute code that modifies those watchers. Note
		2999	also that in that case, you have to re-register any signal watchers.
		3000
		3001	=head3 Watcher-Specific Functions and Data Members
		3002
		3003	=over 4
		3004
		3005	=item ev_fork_init (ev_signal *, callback)
		3006
		3007	Initialises and configures the fork watcher - it has no parameters of any
		3008	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3009	believe me.
		3010
		3011	=back
		3012
		3013
		3014	=head2 C<ev_async> - how to wake up an event loop
		3015
		3016	In general, you cannot use an C<ev_loop> from multiple threads or other
		3017	asynchronous sources such as signal handlers (as opposed to multiple event
		3018	loops - those are of course safe to use in different threads).
		3019
		3020	Sometimes, however, you need to wake up an event loop you do not control,
		3021	for example because it belongs to another thread. This is what C<ev_async>
		3022	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3023	it by calling C<ev_async_send>, which is thread- and signal safe.
		3024
		3025	This functionality is very similar to C<ev_signal> watchers, as signals,
		3026	too, are asynchronous in nature, and signals, too, will be compressed
		3027	(i.e. the number of callback invocations may be less than the number of
		3028	C<ev_async_sent> calls).
		3029
		3030	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		3031	just the default loop.
		3032
		3033	=head3 Queueing
		3034
		3035	C<ev_async> does not support queueing of data in any way. The reason
		3036	is that the author does not know of a simple (or any) algorithm for a
		3037	multiple-writer-single-reader queue that works in all cases and doesn't
		3038	need elaborate support such as pthreads or unportable memory access
		3039	semantics.
		3040
		3041	That means that if you want to queue data, you have to provide your own
		3042	queue. But at least I can tell you how to implement locking around your
		3043	queue:
		3044
		3045	=over 4
		3046
		3047	=item queueing from a signal handler context
		3048
		3049	To implement race-free queueing, you simply add to the queue in the signal
		3050	handler but you block the signal handler in the watcher callback. Here is
		3051	an example that does that for some fictitious SIGUSR1 handler:
		3052
		3053	static ev_async mysig;
		3054
		3055	static void
		3056	sigusr1_handler (void)
		3057	{
		3058	sometype data;
		3059
		3060	// no locking etc.
		3061	queue_put (data);
		3062	ev_async_send (EV_DEFAULT_ &mysig);
		3063	}
		3064
		3065	static void
		3066	mysig_cb (EV_P_ ev_async *w, int revents)
		3067	{
		3068	sometype data;
		3069	sigset_t block, prev;
		3070
		3071	sigemptyset (&block);
		3072	sigaddset (&block, SIGUSR1);
		3073	sigprocmask (SIG_BLOCK, &block, &prev);
		3074
		3075	while (queue_get (&data))
		3076	process (data);
		3077
		3078	if (sigismember (&prev, SIGUSR1)
		3079	sigprocmask (SIG_UNBLOCK, &block, 0);
		3080	}
		3081
		3082	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3083	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3084	either...).
		3085
		3086	=item queueing from a thread context
		3087
		3088	The strategy for threads is different, as you cannot (easily) block
		3089	threads but you can easily preempt them, so to queue safely you need to
		3090	employ a traditional mutex lock, such as in this pthread example:
		3091
		3092	static ev_async mysig;
		3093	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3094
		3095	static void
		3096	otherthread (void)
		3097	{
		3098	// only need to lock the actual queueing operation
		3099	pthread_mutex_lock (&mymutex);
		3100	queue_put (data);
		3101	pthread_mutex_unlock (&mymutex);
		3102
		3103	ev_async_send (EV_DEFAULT_ &mysig);
		3104	}
		3105
		3106	static void
		3107	mysig_cb (EV_P_ ev_async *w, int revents)
		3108	{
		3109	pthread_mutex_lock (&mymutex);
		3110
		3111	while (queue_get (&data))
		3112	process (data);
		3113
		3114	pthread_mutex_unlock (&mymutex);
		3115	}
		3116
		3117	=back
		3118
		3119
		3120	=head3 Watcher-Specific Functions and Data Members
		3121
		3122	=over 4
		3123
		3124	=item ev_async_init (ev_async *, callback)
		3125
		3126	Initialises and configures the async watcher - it has no parameters of any
		3127	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3128	trust me.
		3129
		3130	=item ev_async_send (loop, ev_async *)
		3131
		3132	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3133	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3134	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3135	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3136	section below on what exactly this means).
		3137
		3138	Note that, as with other watchers in libev, multiple events might get
		3139	compressed into a single callback invocation (another way to look at this
		3140	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3141	reset when the event loop detects that).
		3142
		3143	This call incurs the overhead of a system call only once per event loop
		3144	iteration, so while the overhead might be noticeable, it doesn't apply to
		3145	repeated calls to C<ev_async_send> for the same event loop.
		3146
		3147	=item bool = ev_async_pending (ev_async *)
		3148
		3149	Returns a non-zero value when C<ev_async_send> has been called on the
		3150	watcher but the event has not yet been processed (or even noted) by the
		3151	event loop.
		3152
		3153	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3154	the loop iterates next and checks for the watcher to have become active,
		3155	it will reset the flag again. C<ev_async_pending> can be used to very
		3156	quickly check whether invoking the loop might be a good idea.
		3157
		3158	Not that this does I<not> check whether the watcher itself is pending,
		3159	only whether it has been requested to make this watcher pending: there
		3160	is a time window between the event loop checking and resetting the async
		3161	notification, and the callback being invoked.
		3162
		3163	=back
		3164
		3165
720	=head1 OTHER FUNCTIONS	3166	=head1 OTHER FUNCTIONS
721		3167
722	There are some other functions of possible interest. Described. Here. Now.	3168	There are some other functions of possible interest. Described. Here. Now.
723		3169
724	=over 4	3170	=over 4
725		3171
726	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3172	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
727		3173
728	This function combines a simple timer and an I/O watcher, calls your	3174	This function combines a simple timer and an I/O watcher, calls your
729	callback on whichever event happens first and automatically stop both	3175	callback on whichever event happens first and automatically stops both
730	watchers. This is useful if you want to wait for a single event on an fd	3176	watchers. This is useful if you want to wait for a single event on an fd
731	or timeout without having to allocate/configure/start/stop/free one or	3177	or timeout without having to allocate/configure/start/stop/free one or
732	more watchers yourself.	3178	more watchers yourself.
733		3179
734	If C<fd> is less than 0, then no I/O watcher will be started and events	3180	If C<fd> is less than 0, then no I/O watcher will be started and the
735	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3181	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
736	C<events> set will be craeted and started.	3182	the given C<fd> and C<events> set will be created and started.
737		3183
738	If C<timeout> is less than 0, then no timeout watcher will be	3184	If C<timeout> is less than 0, then no timeout watcher will be
739	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3185	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
740	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3186	repeat = 0) will be started. C<0> is a valid timeout.
741	dubious value.
742		3187
743	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3188	The callback has the type C<void (*cb)(int revents, void *arg)> and is
744	passed an C<revents> set like normal event callbacks (a combination of	3189	passed an C<revents> set like normal event callbacks (a combination of
745	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3190	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
746	value passed to C<ev_once>:	3191	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3192	a timeout and an io event at the same time - you probably should give io
		3193	events precedence.
747		3194
		3195	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3196
748	static void stdin_ready (int revents, void *arg)	3197	static void stdin_ready (int revents, void *arg)
749	{	3198	{
750	if (revents & EV_TIMEOUT)
751	/* doh, nothing entered */;
752	else if (revents & EV_READ)	3199	if (revents & EV_READ)
753	/* stdin might have data for us, joy! */;	3200	/* stdin might have data for us, joy! */;
		3201	else if (revents & EV_TIMER)
		3202	/* doh, nothing entered */;
754	}	3203	}
755		3204
756	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3205	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
757
758	=item ev_feed_event (loop, watcher, int events)
759
760	Feeds the given event set into the event loop, as if the specified event
761	had happened for the specified watcher (which must be a pointer to an
762	initialised but not necessarily started event watcher).
763		3206
764	=item ev_feed_fd_event (loop, int fd, int revents)	3207	=item ev_feed_fd_event (loop, int fd, int revents)
765		3208
766	Feed an event on the given fd, as if a file descriptor backend detected	3209	Feed an event on the given fd, as if a file descriptor backend detected
767	the given events it.	3210	the given events it.
768		3211
769	=item ev_feed_signal_event (loop, int signum)	3212	=item ev_feed_signal_event (loop, int signum)
770		3213
771	Feed an event as if the given signal occured (loop must be the default loop!).	3214	Feed an event as if the given signal occurred (C<loop> must be the default
		3215	loop!).
772		3216
773	=back	3217	=back
		3218
774		3219
775	=head1 LIBEVENT EMULATION	3220	=head1 LIBEVENT EMULATION
776		3221
777	Libev offers a compatibility emulation layer for libevent. It cannot	3222	Libev offers a compatibility emulation layer for libevent. It cannot
778	emulate the internals of libevent, so here are some usage hints:	3223	emulate the internals of libevent, so here are some usage hints:
…		…
790		3235
791	=item * Priorities are not currently supported. Initialising priorities	3236	=item * Priorities are not currently supported. Initialising priorities
792	will fail and all watchers will have the same priority, even though there	3237	will fail and all watchers will have the same priority, even though there
793	is an ev_pri field.	3238	is an ev_pri field.
794		3239
		3240	=item * In libevent, the last base created gets the signals, in libev, the
		3241	first base created (== the default loop) gets the signals.
		3242
795	=item * Other members are not supported.	3243	=item * Other members are not supported.
796		3244
797	=item * The libev emulation is I<not> ABI compatible to libevent, you need	3245	=item * The libev emulation is I<not> ABI compatible to libevent, you need
798	to use the libev header file and library.	3246	to use the libev header file and library.
799		3247
800	=back	3248	=back
801		3249
802	=head1 C++ SUPPORT	3250	=head1 C++ SUPPORT
803		3251
804	TBD.	3252	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		3253	you to use some convenience methods to start/stop watchers and also change
		3254	the callback model to a model using method callbacks on objects.
		3255
		3256	To use it,
		3257
		3258	#include <ev++.h>
		3259
		3260	This automatically includes F<ev.h> and puts all of its definitions (many
		3261	of them macros) into the global namespace. All C++ specific things are
		3262	put into the C<ev> namespace. It should support all the same embedding
		3263	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		3264
		3265	Care has been taken to keep the overhead low. The only data member the C++
		3266	classes add (compared to plain C-style watchers) is the event loop pointer
		3267	that the watcher is associated with (or no additional members at all if
		3268	you disable C<EV_MULTIPLICITY> when embedding libev).
		3269
		3270	Currently, functions, and static and non-static member functions can be
		3271	used as callbacks. Other types should be easy to add as long as they only
		3272	need one additional pointer for context. If you need support for other
		3273	types of functors please contact the author (preferably after implementing
		3274	it).
		3275
		3276	Here is a list of things available in the C<ev> namespace:
		3277
		3278	=over 4
		3279
		3280	=item C<ev::READ>, C<ev::WRITE> etc.
		3281
		3282	These are just enum values with the same values as the C<EV_READ> etc.
		3283	macros from F<ev.h>.
		3284
		3285	=item C<ev::tstamp>, C<ev::now>
		3286
		3287	Aliases to the same types/functions as with the C<ev_> prefix.
		3288
		3289	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		3290
		3291	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		3292	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		3293	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		3294	defines by many implementations.
		3295
		3296	All of those classes have these methods:
		3297
		3298	=over 4
		3299
		3300	=item ev::TYPE::TYPE ()
		3301
		3302	=item ev::TYPE::TYPE (loop)
		3303
		3304	=item ev::TYPE::~TYPE
		3305
		3306	The constructor (optionally) takes an event loop to associate the watcher
		3307	with. If it is omitted, it will use C<EV_DEFAULT>.
		3308
		3309	The constructor calls C<ev_init> for you, which means you have to call the
		3310	C<set> method before starting it.
		3311
		3312	It will not set a callback, however: You have to call the templated C<set>
		3313	method to set a callback before you can start the watcher.
		3314
		3315	(The reason why you have to use a method is a limitation in C++ which does
		3316	not allow explicit template arguments for constructors).
		3317
		3318	The destructor automatically stops the watcher if it is active.
		3319
		3320	=item w->set<class, &class::method> (object *)
		3321
		3322	This method sets the callback method to call. The method has to have a
		3323	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		3324	first argument and the C<revents> as second. The object must be given as
		3325	parameter and is stored in the C<data> member of the watcher.
		3326
		3327	This method synthesizes efficient thunking code to call your method from
		3328	the C callback that libev requires. If your compiler can inline your
		3329	callback (i.e. it is visible to it at the place of the C<set> call and
		3330	your compiler is good :), then the method will be fully inlined into the
		3331	thunking function, making it as fast as a direct C callback.
		3332
		3333	Example: simple class declaration and watcher initialisation
		3334
		3335	struct myclass
		3336	{
		3337	void io_cb (ev::io &w, int revents) { }
		3338	}
		3339
		3340	myclass obj;
		3341	ev::io iow;
		3342	iow.set <myclass, &myclass::io_cb> (&obj);
		3343
		3344	=item w->set (object *)
		3345
		3346	This is a variation of a method callback - leaving out the method to call
		3347	will default the method to C<operator ()>, which makes it possible to use
		3348	functor objects without having to manually specify the C<operator ()> all
		3349	the time. Incidentally, you can then also leave out the template argument
		3350	list.
		3351
		3352	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3353	int revents)>.
		3354
		3355	See the method-C<set> above for more details.
		3356
		3357	Example: use a functor object as callback.
		3358
		3359	struct myfunctor
		3360	{
		3361	void operator() (ev::io &w, int revents)
		3362	{
		3363	...
		3364	}
		3365	}
		3366
		3367	myfunctor f;
		3368
		3369	ev::io w;
		3370	w.set (&f);
		3371
		3372	=item w->set<function> (void *data = 0)
		3373
		3374	Also sets a callback, but uses a static method or plain function as
		3375	callback. The optional C<data> argument will be stored in the watcher's
		3376	C<data> member and is free for you to use.
		3377
		3378	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		3379
		3380	See the method-C<set> above for more details.
		3381
		3382	Example: Use a plain function as callback.
		3383
		3384	static void io_cb (ev::io &w, int revents) { }
		3385	iow.set <io_cb> ();
		3386
		3387	=item w->set (loop)
		3388
		3389	Associates a different C<struct ev_loop> with this watcher. You can only
		3390	do this when the watcher is inactive (and not pending either).
		3391
		3392	=item w->set ([arguments])
		3393
		3394	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		3395	called at least once. Unlike the C counterpart, an active watcher gets
		3396	automatically stopped and restarted when reconfiguring it with this
		3397	method.
		3398
		3399	=item w->start ()
		3400
		3401	Starts the watcher. Note that there is no C<loop> argument, as the
		3402	constructor already stores the event loop.
		3403
		3404	=item w->stop ()
		3405
		3406	Stops the watcher if it is active. Again, no C<loop> argument.
		3407
		3408	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		3409
		3410	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		3411	C<ev_TYPE_again> function.
		3412
		3413	=item w->sweep () (C<ev::embed> only)
		3414
		3415	Invokes C<ev_embed_sweep>.
		3416
		3417	=item w->update () (C<ev::stat> only)
		3418
		3419	Invokes C<ev_stat_stat>.
		3420
		3421	=back
		3422
		3423	=back
		3424
		3425	Example: Define a class with an IO and idle watcher, start one of them in
		3426	the constructor.
		3427
		3428	class myclass
		3429	{
		3430	ev::io io ; void io_cb (ev::io &w, int revents);
		3431	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		3432
		3433	myclass (int fd)
		3434	{
		3435	io .set <myclass, &myclass::io_cb > (this);
		3436	idle.set <myclass, &myclass::idle_cb> (this);
		3437
		3438	io.start (fd, ev::READ);
		3439	}
		3440	};
		3441
		3442
		3443	=head1 OTHER LANGUAGE BINDINGS
		3444
		3445	Libev does not offer other language bindings itself, but bindings for a
		3446	number of languages exist in the form of third-party packages. If you know
		3447	any interesting language binding in addition to the ones listed here, drop
		3448	me a note.
		3449
		3450	=over 4
		3451
		3452	=item Perl
		3453
		3454	The EV module implements the full libev API and is actually used to test
		3455	libev. EV is developed together with libev. Apart from the EV core module,
		3456	there are additional modules that implement libev-compatible interfaces
		3457	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3458	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3459	and C<EV::Glib>).
		3460
		3461	It can be found and installed via CPAN, its homepage is at
		3462	L<http://software.schmorp.de/pkg/EV>.
		3463
		3464	=item Python
		3465
		3466	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3467	seems to be quite complete and well-documented.
		3468
		3469	=item Ruby
		3470
		3471	Tony Arcieri has written a ruby extension that offers access to a subset
		3472	of the libev API and adds file handle abstractions, asynchronous DNS and
		3473	more on top of it. It can be found via gem servers. Its homepage is at
		3474	L<http://rev.rubyforge.org/>.
		3475
		3476	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3477	makes rev work even on mingw.
		3478
		3479	=item Haskell
		3480
		3481	A haskell binding to libev is available at
		3482	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3483
		3484	=item D
		3485
		3486	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3487	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3488
		3489	=item Ocaml
		3490
		3491	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3492	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3493
		3494	=item Lua
		3495
		3496	Brian Maher has written a partial interface to libev for lua (at the
		3497	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		3498	L<http://github.com/brimworks/lua-ev>.
		3499
		3500	=back
		3501
		3502
		3503	=head1 MACRO MAGIC
		3504
		3505	Libev can be compiled with a variety of options, the most fundamental
		3506	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3507	functions and callbacks have an initial C<struct ev_loop *> argument.
		3508
		3509	To make it easier to write programs that cope with either variant, the
		3510	following macros are defined:
		3511
		3512	=over 4
		3513
		3514	=item C<EV_A>, C<EV_A_>
		3515
		3516	This provides the loop I<argument> for functions, if one is required ("ev
		3517	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3518	C<EV_A_> is used when other arguments are following. Example:
		3519
		3520	ev_unref (EV_A);
		3521	ev_timer_add (EV_A_ watcher);
		3522	ev_loop (EV_A_ 0);
		3523
		3524	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3525	which is often provided by the following macro.
		3526
		3527	=item C<EV_P>, C<EV_P_>
		3528
		3529	This provides the loop I<parameter> for functions, if one is required ("ev
		3530	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3531	C<EV_P_> is used when other parameters are following. Example:
		3532
		3533	// this is how ev_unref is being declared
		3534	static void ev_unref (EV_P);
		3535
		3536	// this is how you can declare your typical callback
		3537	static void cb (EV_P_ ev_timer *w, int revents)
		3538
		3539	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3540	suitable for use with C<EV_A>.
		3541
		3542	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3543
		3544	Similar to the other two macros, this gives you the value of the default
		3545	loop, if multiple loops are supported ("ev loop default").
		3546
		3547	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3548
		3549	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3550	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3551	is undefined when the default loop has not been initialised by a previous
		3552	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3553
		3554	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3555	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3556
		3557	=back
		3558
		3559	Example: Declare and initialise a check watcher, utilising the above
		3560	macros so it will work regardless of whether multiple loops are supported
		3561	or not.
		3562
		3563	static void
		3564	check_cb (EV_P_ ev_timer *w, int revents)
		3565	{
		3566	ev_check_stop (EV_A_ w);
		3567	}
		3568
		3569	ev_check check;
		3570	ev_check_init (&check, check_cb);
		3571	ev_check_start (EV_DEFAULT_ &check);
		3572	ev_loop (EV_DEFAULT_ 0);
		3573
		3574	=head1 EMBEDDING
		3575
		3576	Libev can (and often is) directly embedded into host
		3577	applications. Examples of applications that embed it include the Deliantra
		3578	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3579	and rxvt-unicode.
		3580
		3581	The goal is to enable you to just copy the necessary files into your
		3582	source directory without having to change even a single line in them, so
		3583	you can easily upgrade by simply copying (or having a checked-out copy of
		3584	libev somewhere in your source tree).
		3585
		3586	=head2 FILESETS
		3587
		3588	Depending on what features you need you need to include one or more sets of files
		3589	in your application.
		3590
		3591	=head3 CORE EVENT LOOP
		3592
		3593	To include only the libev core (all the C<ev_*> functions), with manual
		3594	configuration (no autoconf):
		3595
		3596	#define EV_STANDALONE 1
		3597	#include "ev.c"
		3598
		3599	This will automatically include F<ev.h>, too, and should be done in a
		3600	single C source file only to provide the function implementations. To use
		3601	it, do the same for F<ev.h> in all files wishing to use this API (best
		3602	done by writing a wrapper around F<ev.h> that you can include instead and
		3603	where you can put other configuration options):
		3604
		3605	#define EV_STANDALONE 1
		3606	#include "ev.h"
		3607
		3608	Both header files and implementation files can be compiled with a C++
		3609	compiler (at least, that's a stated goal, and breakage will be treated
		3610	as a bug).
		3611
		3612	You need the following files in your source tree, or in a directory
		3613	in your include path (e.g. in libev/ when using -Ilibev):
		3614
		3615	ev.h
		3616	ev.c
		3617	ev_vars.h
		3618	ev_wrap.h
		3619
		3620	ev_win32.c required on win32 platforms only
		3621
		3622	ev_select.c only when select backend is enabled (which is enabled by default)
		3623	ev_poll.c only when poll backend is enabled (disabled by default)
		3624	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3625	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3626	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3627
		3628	F<ev.c> includes the backend files directly when enabled, so you only need
		3629	to compile this single file.
		3630
		3631	=head3 LIBEVENT COMPATIBILITY API
		3632
		3633	To include the libevent compatibility API, also include:
		3634
		3635	#include "event.c"
		3636
		3637	in the file including F<ev.c>, and:
		3638
		3639	#include "event.h"
		3640
		3641	in the files that want to use the libevent API. This also includes F<ev.h>.
		3642
		3643	You need the following additional files for this:
		3644
		3645	event.h
		3646	event.c
		3647
		3648	=head3 AUTOCONF SUPPORT
		3649
		3650	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3651	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3652	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3653	include F<config.h> and configure itself accordingly.
		3654
		3655	For this of course you need the m4 file:
		3656
		3657	libev.m4
		3658
		3659	=head2 PREPROCESSOR SYMBOLS/MACROS
		3660
		3661	Libev can be configured via a variety of preprocessor symbols you have to
		3662	define before including (or compiling) any of its files. The default in
		3663	the absence of autoconf is documented for every option.
		3664
		3665	Symbols marked with "(h)" do not change the ABI, and can have different
		3666	values when compiling libev vs. including F<ev.h>, so it is permissible
		3667	to redefine them before including F<ev.h> without breaking compatibility
		3668	to a compiled library. All other symbols change the ABI, which means all
		3669	users of libev and the libev code itself must be compiled with compatible
		3670	settings.
		3671
		3672	=over 4
		3673
		3674	=item EV_STANDALONE (h)
		3675
		3676	Must always be C<1> if you do not use autoconf configuration, which
		3677	keeps libev from including F<config.h>, and it also defines dummy
		3678	implementations for some libevent functions (such as logging, which is not
		3679	supported). It will also not define any of the structs usually found in
		3680	F<event.h> that are not directly supported by the libev core alone.
		3681
		3682	In standalone mode, libev will still try to automatically deduce the
		3683	configuration, but has to be more conservative.
		3684
		3685	=item EV_USE_MONOTONIC
		3686
		3687	If defined to be C<1>, libev will try to detect the availability of the
		3688	monotonic clock option at both compile time and runtime. Otherwise no
		3689	use of the monotonic clock option will be attempted. If you enable this,
		3690	you usually have to link against librt or something similar. Enabling it
		3691	when the functionality isn't available is safe, though, although you have
		3692	to make sure you link against any libraries where the C<clock_gettime>
		3693	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		3694
		3695	=item EV_USE_REALTIME
		3696
		3697	If defined to be C<1>, libev will try to detect the availability of the
		3698	real-time clock option at compile time (and assume its availability
		3699	at runtime if successful). Otherwise no use of the real-time clock
		3700	option will be attempted. This effectively replaces C<gettimeofday>
		3701	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		3702	correctness. See the note about libraries in the description of
		3703	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3704	C<EV_USE_CLOCK_SYSCALL>.
		3705
		3706	=item EV_USE_CLOCK_SYSCALL
		3707
		3708	If defined to be C<1>, libev will try to use a direct syscall instead
		3709	of calling the system-provided C<clock_gettime> function. This option
		3710	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3711	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3712	programs needlessly. Using a direct syscall is slightly slower (in
		3713	theory), because no optimised vdso implementation can be used, but avoids
		3714	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3715	higher, as it simplifies linking (no need for C<-lrt>).
		3716
		3717	=item EV_USE_NANOSLEEP
		3718
		3719	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3720	and will use it for delays. Otherwise it will use C<select ()>.
		3721
		3722	=item EV_USE_EVENTFD
		3723
		3724	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3725	available and will probe for kernel support at runtime. This will improve
		3726	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3727	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3728	2.7 or newer, otherwise disabled.
		3729
		3730	=item EV_USE_SELECT
		3731
		3732	If undefined or defined to be C<1>, libev will compile in support for the
		3733	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3734	other method takes over, select will be it. Otherwise the select backend
		3735	will not be compiled in.
		3736
		3737	=item EV_SELECT_USE_FD_SET
		3738
		3739	If defined to C<1>, then the select backend will use the system C<fd_set>
		3740	structure. This is useful if libev doesn't compile due to a missing
		3741	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		3742	on exotic systems. This usually limits the range of file descriptors to
		3743	some low limit such as 1024 or might have other limitations (winsocket
		3744	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		3745	configures the maximum size of the C<fd_set>.
		3746
		3747	=item EV_SELECT_IS_WINSOCKET
		3748
		3749	When defined to C<1>, the select backend will assume that
		3750	select/socket/connect etc. don't understand file descriptors but
		3751	wants osf handles on win32 (this is the case when the select to
		3752	be used is the winsock select). This means that it will call
		3753	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3754	it is assumed that all these functions actually work on fds, even
		3755	on win32. Should not be defined on non-win32 platforms.
		3756
		3757	=item EV_FD_TO_WIN32_HANDLE(fd)
		3758
		3759	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3760	file descriptors to socket handles. When not defining this symbol (the
		3761	default), then libev will call C<_get_osfhandle>, which is usually
		3762	correct. In some cases, programs use their own file descriptor management,
		3763	in which case they can provide this function to map fds to socket handles.
		3764
		3765	=item EV_WIN32_HANDLE_TO_FD(handle)
		3766
		3767	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		3768	using the standard C<_open_osfhandle> function. For programs implementing
		3769	their own fd to handle mapping, overwriting this function makes it easier
		3770	to do so. This can be done by defining this macro to an appropriate value.
		3771
		3772	=item EV_WIN32_CLOSE_FD(fd)
		3773
		3774	If programs implement their own fd to handle mapping on win32, then this
		3775	macro can be used to override the C<close> function, useful to unregister
		3776	file descriptors again. Note that the replacement function has to close
		3777	the underlying OS handle.
		3778
		3779	=item EV_USE_POLL
		3780
		3781	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3782	backend. Otherwise it will be enabled on non-win32 platforms. It
		3783	takes precedence over select.
		3784
		3785	=item EV_USE_EPOLL
		3786
		3787	If defined to be C<1>, libev will compile in support for the Linux
		3788	C<epoll>(7) backend. Its availability will be detected at runtime,
		3789	otherwise another method will be used as fallback. This is the preferred
		3790	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3791	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3792
		3793	=item EV_USE_KQUEUE
		3794
		3795	If defined to be C<1>, libev will compile in support for the BSD style
		3796	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3797	otherwise another method will be used as fallback. This is the preferred
		3798	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3799	supports some types of fds correctly (the only platform we found that
		3800	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3801	not be used unless explicitly requested. The best way to use it is to find
		3802	out whether kqueue supports your type of fd properly and use an embedded
		3803	kqueue loop.
		3804
		3805	=item EV_USE_PORT
		3806
		3807	If defined to be C<1>, libev will compile in support for the Solaris
		3808	10 port style backend. Its availability will be detected at runtime,
		3809	otherwise another method will be used as fallback. This is the preferred
		3810	backend for Solaris 10 systems.
		3811
		3812	=item EV_USE_DEVPOLL
		3813
		3814	Reserved for future expansion, works like the USE symbols above.
		3815
		3816	=item EV_USE_INOTIFY
		3817
		3818	If defined to be C<1>, libev will compile in support for the Linux inotify
		3819	interface to speed up C<ev_stat> watchers. Its actual availability will
		3820	be detected at runtime. If undefined, it will be enabled if the headers
		3821	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3822
		3823	=item EV_ATOMIC_T
		3824
		3825	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3826	access is atomic with respect to other threads or signal contexts. No such
		3827	type is easily found in the C language, so you can provide your own type
		3828	that you know is safe for your purposes. It is used both for signal handler "locking"
		3829	as well as for signal and thread safety in C<ev_async> watchers.
		3830
		3831	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3832	(from F<signal.h>), which is usually good enough on most platforms.
		3833
		3834	=item EV_H (h)
		3835
		3836	The name of the F<ev.h> header file used to include it. The default if
		3837	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3838	used to virtually rename the F<ev.h> header file in case of conflicts.
		3839
		3840	=item EV_CONFIG_H (h)
		3841
		3842	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3843	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3844	C<EV_H>, above.
		3845
		3846	=item EV_EVENT_H (h)
		3847
		3848	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3849	of how the F<event.h> header can be found, the default is C<"event.h">.
		3850
		3851	=item EV_PROTOTYPES (h)
		3852
		3853	If defined to be C<0>, then F<ev.h> will not define any function
		3854	prototypes, but still define all the structs and other symbols. This is
		3855	occasionally useful if you want to provide your own wrapper functions
		3856	around libev functions.
		3857
		3858	=item EV_MULTIPLICITY
		3859
		3860	If undefined or defined to C<1>, then all event-loop-specific functions
		3861	will have the C<struct ev_loop *> as first argument, and you can create
		3862	additional independent event loops. Otherwise there will be no support
		3863	for multiple event loops and there is no first event loop pointer
		3864	argument. Instead, all functions act on the single default loop.
		3865
		3866	=item EV_MINPRI
		3867
		3868	=item EV_MAXPRI
		3869
		3870	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3871	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3872	provide for more priorities by overriding those symbols (usually defined
		3873	to be C<-2> and C<2>, respectively).
		3874
		3875	When doing priority-based operations, libev usually has to linearly search
		3876	all the priorities, so having many of them (hundreds) uses a lot of space
		3877	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3878	fine.
		3879
		3880	If your embedding application does not need any priorities, defining these
		3881	both to C<0> will save some memory and CPU.
		3882
		3883	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		3884	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		3885	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		3886
		3887	If undefined or defined to be C<1> (and the platform supports it), then
		3888	the respective watcher type is supported. If defined to be C<0>, then it
		3889	is not. Disabling watcher types mainly saves code size.
		3890
		3891	=item EV_FEATURES
		3892
		3893	If you need to shave off some kilobytes of code at the expense of some
		3894	speed (but with the full API), you can define this symbol to request
		3895	certain subsets of functionality. The default is to enable all features
		3896	that can be enabled on the platform.
		3897
		3898	A typical way to use this symbol is to define it to C<0> (or to a bitset
		3899	with some broad features you want) and then selectively re-enable
		3900	additional parts you want, for example if you want everything minimal,
		3901	but multiple event loop support, async and child watchers and the poll
		3902	backend, use this:
		3903
		3904	#define EV_FEATURES 0
		3905	#define EV_MULTIPLICITY 1
		3906	#define EV_USE_POLL 1
		3907	#define EV_CHILD_ENABLE 1
		3908	#define EV_ASYNC_ENABLE 1
		3909
		3910	The actual value is a bitset, it can be a combination of the following
		3911	values:
		3912
		3913	=over 4
		3914
		3915	=item C<1> - faster/larger code
		3916
		3917	Use larger code to speed up some operations.
		3918
		3919	Currently this is used to override some inlining decisions (enlarging the
		3920	code size by roughly 30% on amd64).
		3921
		3922	When optimising for size, use of compiler flags such as C<-Os> with
		3923	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		3924	assertions.
		3925
		3926	=item C<2> - faster/larger data structures
		3927
		3928	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		3929	hash table sizes and so on. This will usually further increase code size
		3930	and can additionally have an effect on the size of data structures at
		3931	runtime.
		3932
		3933	=item C<4> - full API configuration
		3934
		3935	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		3936	enables multiplicity (C<EV_MULTIPLICITY>=1).
		3937
		3938	=item C<8> - full API
		3939
		3940	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		3941	details on which parts of the API are still available without this
		3942	feature, and do not complain if this subset changes over time.
		3943
		3944	=item C<16> - enable all optional watcher types
		3945
		3946	Enables all optional watcher types. If you want to selectively enable
		3947	only some watcher types other than I/O and timers (e.g. prepare,
		3948	embed, async, child...) you can enable them manually by defining
		3949	C<EV_watchertype_ENABLE> to C<1> instead.
		3950
		3951	=item C<32> - enable all backends
		3952
		3953	This enables all backends - without this feature, you need to enable at
		3954	least one backend manually (C<EV_USE_SELECT> is a good choice).
		3955
		3956	=item C<64> - enable OS-specific "helper" APIs
		3957
		3958	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		3959	default.
		3960
		3961	=back
		3962
		3963	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		3964	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		3965	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		3966	watchers, timers and monotonic clock support.
		3967
		3968	With an intelligent-enough linker (gcc+binutils are intelligent enough
		3969	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		3970	your program might be left out as well - a binary starting a timer and an
		3971	I/O watcher then might come out at only 5Kb.
		3972
		3973	=item EV_AVOID_STDIO
		3974
		3975	If this is set to C<1> at compiletime, then libev will avoid using stdio
		3976	functions (printf, scanf, perror etc.). This will increase the code size
		3977	somewhat, but if your program doesn't otherwise depend on stdio and your
		3978	libc allows it, this avoids linking in the stdio library which is quite
		3979	big.
		3980
		3981	Note that error messages might become less precise when this option is
		3982	enabled.
		3983
		3984	=item EV_NSIG
		3985
		3986	The highest supported signal number, +1 (or, the number of
		3987	signals): Normally, libev tries to deduce the maximum number of signals
		3988	automatically, but sometimes this fails, in which case it can be
		3989	specified. Also, using a lower number than detected (C<32> should be
		3990	good for about any system in existence) can save some memory, as libev
		3991	statically allocates some 12-24 bytes per signal number.
		3992
		3993	=item EV_PID_HASHSIZE
		3994
		3995	C<ev_child> watchers use a small hash table to distribute workload by
		3996	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		3997	usually more than enough. If you need to manage thousands of children you
		3998	might want to increase this value (I<must> be a power of two).
		3999
		4000	=item EV_INOTIFY_HASHSIZE
		4001
		4002	C<ev_stat> watchers use a small hash table to distribute workload by
		4003	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		4004	disabled), usually more than enough. If you need to manage thousands of
		4005	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4006	power of two).
		4007
		4008	=item EV_USE_4HEAP
		4009
		4010	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4011	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4012	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4013	faster performance with many (thousands) of watchers.
		4014
		4015	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4016	will be C<0>.
		4017
		4018	=item EV_HEAP_CACHE_AT
		4019
		4020	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4021	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4022	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4023	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4024	but avoids random read accesses on heap changes. This improves performance
		4025	noticeably with many (hundreds) of watchers.
		4026
		4027	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4028	will be C<0>.
		4029
		4030	=item EV_VERIFY
		4031
		4032	Controls how much internal verification (see C<ev_loop_verify ()>) will
		4033	be done: If set to C<0>, no internal verification code will be compiled
		4034	in. If set to C<1>, then verification code will be compiled in, but not
		4035	called. If set to C<2>, then the internal verification code will be
		4036	called once per loop, which can slow down libev. If set to C<3>, then the
		4037	verification code will be called very frequently, which will slow down
		4038	libev considerably.
		4039
		4040	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4041	will be C<0>.
		4042
		4043	=item EV_COMMON
		4044
		4045	By default, all watchers have a C<void *data> member. By redefining
		4046	this macro to something else you can include more and other types of
		4047	members. You have to define it each time you include one of the files,
		4048	though, and it must be identical each time.
		4049
		4050	For example, the perl EV module uses something like this:
		4051
		4052	#define EV_COMMON \
		4053	SV *self; /* contains this struct */ \
		4054	SV *cb_sv, *fh /* note no trailing ";" */
		4055
		4056	=item EV_CB_DECLARE (type)
		4057
		4058	=item EV_CB_INVOKE (watcher, revents)
		4059
		4060	=item ev_set_cb (ev, cb)
		4061
		4062	Can be used to change the callback member declaration in each watcher,
		4063	and the way callbacks are invoked and set. Must expand to a struct member
		4064	definition and a statement, respectively. See the F<ev.h> header file for
		4065	their default definitions. One possible use for overriding these is to
		4066	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		4067	method calls instead of plain function calls in C++.
		4068
		4069	=back
		4070
		4071	=head2 EXPORTED API SYMBOLS
		4072
		4073	If you need to re-export the API (e.g. via a DLL) and you need a list of
		4074	exported symbols, you can use the provided F<Symbol.*> files which list
		4075	all public symbols, one per line:
		4076
		4077	Symbols.ev for libev proper
		4078	Symbols.event for the libevent emulation
		4079
		4080	This can also be used to rename all public symbols to avoid clashes with
		4081	multiple versions of libev linked together (which is obviously bad in
		4082	itself, but sometimes it is inconvenient to avoid this).
		4083
		4084	A sed command like this will create wrapper C<#define>'s that you need to
		4085	include before including F<ev.h>:
		4086
		4087	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		4088
		4089	This would create a file F<wrap.h> which essentially looks like this:
		4090
		4091	#define ev_backend myprefix_ev_backend
		4092	#define ev_check_start myprefix_ev_check_start
		4093	#define ev_check_stop myprefix_ev_check_stop
		4094	...
		4095
		4096	=head2 EXAMPLES
		4097
		4098	For a real-world example of a program the includes libev
		4099	verbatim, you can have a look at the EV perl module
		4100	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		4101	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		4102	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		4103	will be compiled. It is pretty complex because it provides its own header
		4104	file.
		4105
		4106	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		4107	that everybody includes and which overrides some configure choices:
		4108
		4109	#define EV_FEATURES 8
		4110	#define EV_USE_SELECT 1
		4111	#define EV_PREPARE_ENABLE 1
		4112	#define EV_IDLE_ENABLE 1
		4113	#define EV_SIGNAL_ENABLE 1
		4114	#define EV_CHILD_ENABLE 1
		4115	#define EV_USE_STDEXCEPT 0
		4116	#define EV_CONFIG_H <config.h>
		4117
		4118	#include "ev++.h"
		4119
		4120	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		4121
		4122	#include "ev_cpp.h"
		4123	#include "ev.c"
		4124
		4125	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		4126
		4127	=head2 THREADS AND COROUTINES
		4128
		4129	=head3 THREADS
		4130
		4131	All libev functions are reentrant and thread-safe unless explicitly
		4132	documented otherwise, but libev implements no locking itself. This means
		4133	that you can use as many loops as you want in parallel, as long as there
		4134	are no concurrent calls into any libev function with the same loop
		4135	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		4136	of course): libev guarantees that different event loops share no data
		4137	structures that need any locking.
		4138
		4139	Or to put it differently: calls with different loop parameters can be done
		4140	concurrently from multiple threads, calls with the same loop parameter
		4141	must be done serially (but can be done from different threads, as long as
		4142	only one thread ever is inside a call at any point in time, e.g. by using
		4143	a mutex per loop).
		4144
		4145	Specifically to support threads (and signal handlers), libev implements
		4146	so-called C<ev_async> watchers, which allow some limited form of
		4147	concurrency on the same event loop, namely waking it up "from the
		4148	outside".
		4149
		4150	If you want to know which design (one loop, locking, or multiple loops
		4151	without or something else still) is best for your problem, then I cannot
		4152	help you, but here is some generic advice:
		4153
		4154	=over 4
		4155
		4156	=item * most applications have a main thread: use the default libev loop
		4157	in that thread, or create a separate thread running only the default loop.
		4158
		4159	This helps integrating other libraries or software modules that use libev
		4160	themselves and don't care/know about threading.
		4161
		4162	=item * one loop per thread is usually a good model.
		4163
		4164	Doing this is almost never wrong, sometimes a better-performance model
		4165	exists, but it is always a good start.
		4166
		4167	=item * other models exist, such as the leader/follower pattern, where one
		4168	loop is handed through multiple threads in a kind of round-robin fashion.
		4169
		4170	Choosing a model is hard - look around, learn, know that usually you can do
		4171	better than you currently do :-)
		4172
		4173	=item * often you need to talk to some other thread which blocks in the
		4174	event loop.
		4175
		4176	C<ev_async> watchers can be used to wake them up from other threads safely
		4177	(or from signal contexts...).
		4178
		4179	An example use would be to communicate signals or other events that only
		4180	work in the default loop by registering the signal watcher with the
		4181	default loop and triggering an C<ev_async> watcher from the default loop
		4182	watcher callback into the event loop interested in the signal.
		4183
		4184	=back
		4185
		4186	=head4 THREAD LOCKING EXAMPLE
		4187
		4188	Here is a fictitious example of how to run an event loop in a different
		4189	thread than where callbacks are being invoked and watchers are
		4190	created/added/removed.
		4191
		4192	For a real-world example, see the C<EV::Loop::Async> perl module,
		4193	which uses exactly this technique (which is suited for many high-level
		4194	languages).
		4195
		4196	The example uses a pthread mutex to protect the loop data, a condition
		4197	variable to wait for callback invocations, an async watcher to notify the
		4198	event loop thread and an unspecified mechanism to wake up the main thread.
		4199
		4200	First, you need to associate some data with the event loop:
		4201
		4202	typedef struct {
		4203	mutex_t lock; /* global loop lock */
		4204	ev_async async_w;
		4205	thread_t tid;
		4206	cond_t invoke_cv;
		4207	} userdata;
		4208
		4209	void prepare_loop (EV_P)
		4210	{
		4211	// for simplicity, we use a static userdata struct.
		4212	static userdata u;
		4213
		4214	ev_async_init (&u->async_w, async_cb);
		4215	ev_async_start (EV_A_ &u->async_w);
		4216
		4217	pthread_mutex_init (&u->lock, 0);
		4218	pthread_cond_init (&u->invoke_cv, 0);
		4219
		4220	// now associate this with the loop
		4221	ev_set_userdata (EV_A_ u);
		4222	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4223	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4224
		4225	// then create the thread running ev_loop
		4226	pthread_create (&u->tid, 0, l_run, EV_A);
		4227	}
		4228
		4229	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4230	solely to wake up the event loop so it takes notice of any new watchers
		4231	that might have been added:
		4232
		4233	static void
		4234	async_cb (EV_P_ ev_async *w, int revents)
		4235	{
		4236	// just used for the side effects
		4237	}
		4238
		4239	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4240	protecting the loop data, respectively.
		4241
		4242	static void
		4243	l_release (EV_P)
		4244	{
		4245	userdata *u = ev_userdata (EV_A);
		4246	pthread_mutex_unlock (&u->lock);
		4247	}
		4248
		4249	static void
		4250	l_acquire (EV_P)
		4251	{
		4252	userdata *u = ev_userdata (EV_A);
		4253	pthread_mutex_lock (&u->lock);
		4254	}
		4255
		4256	The event loop thread first acquires the mutex, and then jumps straight
		4257	into C<ev_loop>:
		4258
		4259	void *
		4260	l_run (void *thr_arg)
		4261	{
		4262	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4263
		4264	l_acquire (EV_A);
		4265	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4266	ev_loop (EV_A_ 0);
		4267	l_release (EV_A);
		4268
		4269	return 0;
		4270	}
		4271
		4272	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4273	signal the main thread via some unspecified mechanism (signals? pipe
		4274	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4275	have been called (in a while loop because a) spurious wakeups are possible
		4276	and b) skipping inter-thread-communication when there are no pending
		4277	watchers is very beneficial):
		4278
		4279	static void
		4280	l_invoke (EV_P)
		4281	{
		4282	userdata *u = ev_userdata (EV_A);
		4283
		4284	while (ev_pending_count (EV_A))
		4285	{
		4286	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4287	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4288	}
		4289	}
		4290
		4291	Now, whenever the main thread gets told to invoke pending watchers, it
		4292	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4293	thread to continue:
		4294
		4295	static void
		4296	real_invoke_pending (EV_P)
		4297	{
		4298	userdata *u = ev_userdata (EV_A);
		4299
		4300	pthread_mutex_lock (&u->lock);
		4301	ev_invoke_pending (EV_A);
		4302	pthread_cond_signal (&u->invoke_cv);
		4303	pthread_mutex_unlock (&u->lock);
		4304	}
		4305
		4306	Whenever you want to start/stop a watcher or do other modifications to an
		4307	event loop, you will now have to lock:
		4308
		4309	ev_timer timeout_watcher;
		4310	userdata *u = ev_userdata (EV_A);
		4311
		4312	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4313
		4314	pthread_mutex_lock (&u->lock);
		4315	ev_timer_start (EV_A_ &timeout_watcher);
		4316	ev_async_send (EV_A_ &u->async_w);
		4317	pthread_mutex_unlock (&u->lock);
		4318
		4319	Note that sending the C<ev_async> watcher is required because otherwise
		4320	an event loop currently blocking in the kernel will have no knowledge
		4321	about the newly added timer. By waking up the loop it will pick up any new
		4322	watchers in the next event loop iteration.
		4323
		4324	=head3 COROUTINES
		4325
		4326	Libev is very accommodating to coroutines ("cooperative threads"):
		4327	libev fully supports nesting calls to its functions from different
		4328	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		4329	different coroutines, and switch freely between both coroutines running
		4330	the loop, as long as you don't confuse yourself). The only exception is
		4331	that you must not do this from C<ev_periodic> reschedule callbacks.
		4332
		4333	Care has been taken to ensure that libev does not keep local state inside
		4334	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		4335	they do not call any callbacks.
		4336
		4337	=head2 COMPILER WARNINGS
		4338
		4339	Depending on your compiler and compiler settings, you might get no or a
		4340	lot of warnings when compiling libev code. Some people are apparently
		4341	scared by this.
		4342
		4343	However, these are unavoidable for many reasons. For one, each compiler
		4344	has different warnings, and each user has different tastes regarding
		4345	warning options. "Warn-free" code therefore cannot be a goal except when
		4346	targeting a specific compiler and compiler-version.
		4347
		4348	Another reason is that some compiler warnings require elaborate
		4349	workarounds, or other changes to the code that make it less clear and less
		4350	maintainable.
		4351
		4352	And of course, some compiler warnings are just plain stupid, or simply
		4353	wrong (because they don't actually warn about the condition their message
		4354	seems to warn about). For example, certain older gcc versions had some
		4355	warnings that resulted in an extreme number of false positives. These have
		4356	been fixed, but some people still insist on making code warn-free with
		4357	such buggy versions.
		4358
		4359	While libev is written to generate as few warnings as possible,
		4360	"warn-free" code is not a goal, and it is recommended not to build libev
		4361	with any compiler warnings enabled unless you are prepared to cope with
		4362	them (e.g. by ignoring them). Remember that warnings are just that:
		4363	warnings, not errors, or proof of bugs.
		4364
		4365
		4366	=head2 VALGRIND
		4367
		4368	Valgrind has a special section here because it is a popular tool that is
		4369	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4370
		4371	If you think you found a bug (memory leak, uninitialised data access etc.)
		4372	in libev, then check twice: If valgrind reports something like:
		4373
		4374	==2274== definitely lost: 0 bytes in 0 blocks.
		4375	==2274== possibly lost: 0 bytes in 0 blocks.
		4376	==2274== still reachable: 256 bytes in 1 blocks.
		4377
		4378	Then there is no memory leak, just as memory accounted to global variables
		4379	is not a memleak - the memory is still being referenced, and didn't leak.
		4380
		4381	Similarly, under some circumstances, valgrind might report kernel bugs
		4382	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4383	although an acceptable workaround has been found here), or it might be
		4384	confused.
		4385
		4386	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4387	make it into some kind of religion.
		4388
		4389	If you are unsure about something, feel free to contact the mailing list
		4390	with the full valgrind report and an explanation on why you think this
		4391	is a bug in libev (best check the archives, too :). However, don't be
		4392	annoyed when you get a brisk "this is no bug" answer and take the chance
		4393	of learning how to interpret valgrind properly.
		4394
		4395	If you need, for some reason, empty reports from valgrind for your project
		4396	I suggest using suppression lists.
		4397
		4398
		4399	=head1 PORTABILITY NOTES
		4400
		4401	=head2 GNU/LINUX 32 BIT LIMITATIONS
		4402
		4403	GNU/Linux is the only common platform that supports 64 bit file/large file
		4404	interfaces but I<disables> them by default.
		4405
		4406	That means that libev compiled in the default environment doesn't support
		4407	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		4408
		4409	Unfortunately, many programs try to work around this GNU/Linux issue
		4410	by enabling the large file API, which makes them incompatible with the
		4411	standard libev compiled for their system.
		4412
		4413	Likewise, libev cannot enable the large file API itself as this would
		4414	suddenly make it incompatible to the default compile time environment,
		4415	i.e. all programs not using special compile switches.
		4416
		4417	=head2 OS/X AND DARWIN BUGS
		4418
		4419	The whole thing is a bug if you ask me - basically any system interface
		4420	you touch is broken, whether it is locales, poll, kqueue or even the
		4421	OpenGL drivers.
		4422
		4423	=head3 C<kqueue> is buggy
		4424
		4425	The kqueue syscall is broken in all known versions - most versions support
		4426	only sockets, many support pipes.
		4427
		4428	Libev tries to work around this by not using C<kqueue> by default on
		4429	this rotten platform, but of course you can still ask for it when creating
		4430	a loop.
		4431
		4432	=head3 C<poll> is buggy
		4433
		4434	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		4435	implementation by something calling C<kqueue> internally around the 10.5.6
		4436	release, so now C<kqueue> I<and> C<poll> are broken.
		4437
		4438	Libev tries to work around this by not using C<poll> by default on
		4439	this rotten platform, but of course you can still ask for it when creating
		4440	a loop.
		4441
		4442	=head3 C<select> is buggy
		4443
		4444	All that's left is C<select>, and of course Apple found a way to fuck this
		4445	one up as well: On OS/X, C<select> actively limits the number of file
		4446	descriptors you can pass in to 1024 - your program suddenly crashes when
		4447	you use more.
		4448
		4449	There is an undocumented "workaround" for this - defining
		4450	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		4451	work on OS/X.
		4452
		4453	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		4454
		4455	=head3 C<errno> reentrancy
		4456
		4457	The default compile environment on Solaris is unfortunately so
		4458	thread-unsafe that you can't even use components/libraries compiled
		4459	without C<-D_REENTRANT> (as long as they use C<errno>), which, of course,
		4460	isn't defined by default.
		4461
		4462	If you want to use libev in threaded environments you have to make sure
		4463	it's compiled with C<_REENTRANT> defined.
		4464
		4465	=head3 Event port backend
		4466
		4467	The scalable event interface for Solaris is called "event ports". Unfortunately,
		4468	this mechanism is very buggy. If you run into high CPU usage, your program
		4469	freezes or you get a large number of spurious wakeups, make sure you have
		4470	all the relevant and latest kernel patches applied. No, I don't know which
		4471	ones, but there are multiple ones.
		4472
		4473	If you can't get it to work, you can try running the program by setting
		4474	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		4475	C<select> backends.
		4476
		4477	=head2 AIX POLL BUG
		4478
		4479	AIX unfortunately has a broken C<poll.h> header. Libev works around
		4480	this by trying to avoid the poll backend altogether (i.e. it's not even
		4481	compiled in), which normally isn't a big problem as C<select> works fine
		4482	with large bitsets, and AIX is dead anyway.
		4483
		4484	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4485
		4486	=head3 General issues
		4487
		4488	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4489	requires, and its I/O model is fundamentally incompatible with the POSIX
		4490	model. Libev still offers limited functionality on this platform in
		4491	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4492	descriptors. This only applies when using Win32 natively, not when using
		4493	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		4494	as every compielr comes with a slightly differently broken/incompatible
		4495	environment.
		4496
		4497	Lifting these limitations would basically require the full
		4498	re-implementation of the I/O system. If you are into this kind of thing,
		4499	then note that glib does exactly that for you in a very portable way (note
		4500	also that glib is the slowest event library known to man).
		4501
		4502	There is no supported compilation method available on windows except
		4503	embedding it into other applications.
		4504
		4505	Sensible signal handling is officially unsupported by Microsoft - libev
		4506	tries its best, but under most conditions, signals will simply not work.
		4507
		4508	Not a libev limitation but worth mentioning: windows apparently doesn't
		4509	accept large writes: instead of resulting in a partial write, windows will
		4510	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4511	so make sure you only write small amounts into your sockets (less than a
		4512	megabyte seems safe, but this apparently depends on the amount of memory
		4513	available).
		4514
		4515	Due to the many, low, and arbitrary limits on the win32 platform and
		4516	the abysmal performance of winsockets, using a large number of sockets
		4517	is not recommended (and not reasonable). If your program needs to use
		4518	more than a hundred or so sockets, then likely it needs to use a totally
		4519	different implementation for windows, as libev offers the POSIX readiness
		4520	notification model, which cannot be implemented efficiently on windows
		4521	(due to Microsoft monopoly games).
		4522
		4523	A typical way to use libev under windows is to embed it (see the embedding
		4524	section for details) and use the following F<evwrap.h> header file instead
		4525	of F<ev.h>:
		4526
		4527	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4528	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4529
		4530	#include "ev.h"
		4531
		4532	And compile the following F<evwrap.c> file into your project (make sure
		4533	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4534
		4535	#include "evwrap.h"
		4536	#include "ev.c"
		4537
		4538	=head3 The winsocket C<select> function
		4539
		4540	The winsocket C<select> function doesn't follow POSIX in that it
		4541	requires socket I<handles> and not socket I<file descriptors> (it is
		4542	also extremely buggy). This makes select very inefficient, and also
		4543	requires a mapping from file descriptors to socket handles (the Microsoft
		4544	C runtime provides the function C<_open_osfhandle> for this). See the
		4545	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4546	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4547
		4548	The configuration for a "naked" win32 using the Microsoft runtime
		4549	libraries and raw winsocket select is:
		4550
		4551	#define EV_USE_SELECT 1
		4552	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4553
		4554	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4555	complexity in the O(n²) range when using win32.
		4556
		4557	=head3 Limited number of file descriptors
		4558
		4559	Windows has numerous arbitrary (and low) limits on things.
		4560
		4561	Early versions of winsocket's select only supported waiting for a maximum
		4562	of C<64> handles (probably owning to the fact that all windows kernels
		4563	can only wait for C<64> things at the same time internally; Microsoft
		4564	recommends spawning a chain of threads and wait for 63 handles and the
		4565	previous thread in each. Sounds great!).
		4566
		4567	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4568	to some high number (e.g. C<2048>) before compiling the winsocket select
		4569	call (which might be in libev or elsewhere, for example, perl and many
		4570	other interpreters do their own select emulation on windows).
		4571
		4572	Another limit is the number of file descriptors in the Microsoft runtime
		4573	libraries, which by default is C<64> (there must be a hidden I<64>
		4574	fetish or something like this inside Microsoft). You can increase this
		4575	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4576	(another arbitrary limit), but is broken in many versions of the Microsoft
		4577	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4578	(depending on windows version and/or the phase of the moon). To get more,
		4579	you need to wrap all I/O functions and provide your own fd management, but
		4580	the cost of calling select (O(n²)) will likely make this unworkable.
		4581
		4582	=head2 PORTABILITY REQUIREMENTS
		4583
		4584	In addition to a working ISO-C implementation and of course the
		4585	backend-specific APIs, libev relies on a few additional extensions:
		4586
		4587	=over 4
		4588
		4589	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4590	calling conventions regardless of C<ev_watcher_type *>.
		4591
		4592	Libev assumes not only that all watcher pointers have the same internal
		4593	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4594	assumes that the same (machine) code can be used to call any watcher
		4595	callback: The watcher callbacks have different type signatures, but libev
		4596	calls them using an C<ev_watcher *> internally.
		4597
		4598	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4599
		4600	The type C<sig_atomic_t volatile> (or whatever is defined as
		4601	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4602	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4603	believed to be sufficiently portable.
		4604
		4605	=item C<sigprocmask> must work in a threaded environment
		4606
		4607	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4608	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4609	pthread implementations will either allow C<sigprocmask> in the "main
		4610	thread" or will block signals process-wide, both behaviours would
		4611	be compatible with libev. Interaction between C<sigprocmask> and
		4612	C<pthread_sigmask> could complicate things, however.
		4613
		4614	The most portable way to handle signals is to block signals in all threads
		4615	except the initial one, and run the default loop in the initial thread as
		4616	well.
		4617
		4618	=item C<long> must be large enough for common memory allocation sizes
		4619
		4620	To improve portability and simplify its API, libev uses C<long> internally
		4621	instead of C<size_t> when allocating its data structures. On non-POSIX
		4622	systems (Microsoft...) this might be unexpectedly low, but is still at
		4623	least 31 bits everywhere, which is enough for hundreds of millions of
		4624	watchers.
		4625
		4626	=item C<double> must hold a time value in seconds with enough accuracy
		4627
		4628	The type C<double> is used to represent timestamps. It is required to
		4629	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		4630	enough for at least into the year 4000. This requirement is fulfilled by
		4631	implementations implementing IEEE 754, which is basically all existing
		4632	ones. With IEEE 754 doubles, you get microsecond accuracy until at least
		4633	2200.
		4634
		4635	=back
		4636
		4637	If you know of other additional requirements drop me a note.
		4638
		4639
		4640	=head1 ALGORITHMIC COMPLEXITIES
		4641
		4642	In this section the complexities of (many of) the algorithms used inside
		4643	libev will be documented. For complexity discussions about backends see
		4644	the documentation for C<ev_default_init>.
		4645
		4646	All of the following are about amortised time: If an array needs to be
		4647	extended, libev needs to realloc and move the whole array, but this
		4648	happens asymptotically rarer with higher number of elements, so O(1) might
		4649	mean that libev does a lengthy realloc operation in rare cases, but on
		4650	average it is much faster and asymptotically approaches constant time.
		4651
		4652	=over 4
		4653
		4654	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		4655
		4656	This means that, when you have a watcher that triggers in one hour and
		4657	there are 100 watchers that would trigger before that, then inserting will
		4658	have to skip roughly seven (C<ld 100>) of these watchers.
		4659
		4660	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		4661
		4662	That means that changing a timer costs less than removing/adding them,
		4663	as only the relative motion in the event queue has to be paid for.
		4664
		4665	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		4666
		4667	These just add the watcher into an array or at the head of a list.
		4668
		4669	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		4670
		4671	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		4672
		4673	These watchers are stored in lists, so they need to be walked to find the
		4674	correct watcher to remove. The lists are usually short (you don't usually
		4675	have many watchers waiting for the same fd or signal: one is typical, two
		4676	is rare).
		4677
		4678	=item Finding the next timer in each loop iteration: O(1)
		4679
		4680	By virtue of using a binary or 4-heap, the next timer is always found at a
		4681	fixed position in the storage array.
		4682
		4683	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		4684
		4685	A change means an I/O watcher gets started or stopped, which requires
		4686	libev to recalculate its status (and possibly tell the kernel, depending
		4687	on backend and whether C<ev_io_set> was used).
		4688
		4689	=item Activating one watcher (putting it into the pending state): O(1)
		4690
		4691	=item Priority handling: O(number_of_priorities)
		4692
		4693	Priorities are implemented by allocating some space for each
		4694	priority. When doing priority-based operations, libev usually has to
		4695	linearly search all the priorities, but starting/stopping and activating
		4696	watchers becomes O(1) with respect to priority handling.
		4697
		4698	=item Sending an ev_async: O(1)
		4699
		4700	=item Processing ev_async_send: O(number_of_async_watchers)
		4701
		4702	=item Processing signals: O(max_signal_number)
		4703
		4704	Sending involves a system call I<iff> there were no other C<ev_async_send>
		4705	calls in the current loop iteration. Checking for async and signal events
		4706	involves iterating over all running async watchers or all signal numbers.
		4707
		4708	=back
		4709
		4710
		4711	=head1 PORTING FROM LIBEV 3.X TO 4.X
		4712
		4713	The major version 4 introduced some minor incompatible changes to the API.
		4714
		4715	At the moment, the C<ev.h> header file tries to implement superficial
		4716	compatibility, so most programs should still compile. Those might be
		4717	removed in later versions of libev, so better update early than late.
		4718
		4719	=over 4
		4720
		4721	=item C<ev_loop_count> renamed to C<ev_iteration>
		4722
		4723	=item C<ev_loop_depth> renamed to C<ev_depth>
		4724
		4725	=item C<ev_loop_verify> renamed to C<ev_verify>
		4726
		4727	Most functions working on C<struct ev_loop> objects don't have an
		4728	C<ev_loop_> prefix, so it was removed. Note that C<ev_loop_fork> is
		4729	still called C<ev_loop_fork> because it would otherwise clash with the
		4730	C<ev_fork> typedef.
		4731
		4732	=item C<EV_TIMEOUT> renamed to C<EV_TIMER> in C<revents>
		4733
		4734	This is a simple rename - all other watcher types use their name
		4735	as revents flag, and now C<ev_timer> does, too.
		4736
		4737	Both C<EV_TIMER> and C<EV_TIMEOUT> symbols were present in 3.x versions
		4738	and continue to be present for the foreseeable future, so this is mostly a
		4739	documentation change.
		4740
		4741	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		4742
		4743	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		4744	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		4745	and work, but the library code will of course be larger.
		4746
		4747	=back
		4748
		4749
		4750	=head1 GLOSSARY
		4751
		4752	=over 4
		4753
		4754	=item active
		4755
		4756	A watcher is active as long as it has been started (has been attached to
		4757	an event loop) but not yet stopped (disassociated from the event loop).
		4758
		4759	=item application
		4760
		4761	In this document, an application is whatever is using libev.
		4762
		4763	=item callback
		4764
		4765	The address of a function that is called when some event has been
		4766	detected. Callbacks are being passed the event loop, the watcher that
		4767	received the event, and the actual event bitset.
		4768
		4769	=item callback invocation
		4770
		4771	The act of calling the callback associated with a watcher.
		4772
		4773	=item event
		4774
		4775	A change of state of some external event, such as data now being available
		4776	for reading on a file descriptor, time having passed or simply not having
		4777	any other events happening anymore.
		4778
		4779	In libev, events are represented as single bits (such as C<EV_READ> or
		4780	C<EV_TIMER>).
		4781
		4782	=item event library
		4783
		4784	A software package implementing an event model and loop.
		4785
		4786	=item event loop
		4787
		4788	An entity that handles and processes external events and converts them
		4789	into callback invocations.
		4790
		4791	=item event model
		4792
		4793	The model used to describe how an event loop handles and processes
		4794	watchers and events.
		4795
		4796	=item pending
		4797
		4798	A watcher is pending as soon as the corresponding event has been detected,
		4799	and stops being pending as soon as the watcher will be invoked or its
		4800	pending status is explicitly cleared by the application.
		4801
		4802	A watcher can be pending, but not active. Stopping a watcher also clears
		4803	its pending status.
		4804
		4805	=item real time
		4806
		4807	The physical time that is observed. It is apparently strictly monotonic :)
		4808
		4809	=item wall-clock time
		4810
		4811	The time and date as shown on clocks. Unlike real time, it can actually
		4812	be wrong and jump forwards and backwards, e.g. when the you adjust your
		4813	clock.
		4814
		4815	=item watcher
		4816
		4817	A data structure that describes interest in certain events. Watchers need
		4818	to be started (attached to an event loop) before they can receive events.
		4819
		4820	=item watcher invocation
		4821
		4822	The act of calling the callback associated with a watcher.
		4823
		4824	=back
805		4825
806	=head1 AUTHOR	4826	=head1 AUTHOR
807		4827
808	Marc Lehmann <libev@schmorp.de>.	4828	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
809		4829

Diff Legend

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