ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.27 by root, Wed Nov 14 05:02:07 2007 UTC vs.
Revision 1.199 by root, Thu Oct 23 07:18:21 2008 UTC

…		…
2		2
3	libev - a high performance full-featured event loop written in C	3	libev - a high performance full-featured event loop written in C
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	#include <ev.h>	7	#include <ev.h>
		8
		9	=head2 EXAMPLE PROGRAM
		10
		11	// a single header file is required
		12	#include <ev.h>
		13
		14	// every watcher type has its own typedef'd struct
		15	// with the name ev_<type>
		16	ev_io stdin_watcher;
		17	ev_timer timeout_watcher;
		18
		19	// all watcher callbacks have a similar signature
		20	// this callback is called when data is readable on stdin
		21	static void
		22	stdin_cb (EV_P_ ev_io *w, int revents)
		23	{
		24	puts ("stdin ready");
		25	// for one-shot events, one must manually stop the watcher
		26	// with its corresponding stop function.
		27	ev_io_stop (EV_A_ w);
		28
		29	// this causes all nested ev_loop's to stop iterating
		30	ev_unloop (EV_A_ EVUNLOOP_ALL);
		31	}
		32
		33	// another callback, this time for a time-out
		34	static void
		35	timeout_cb (EV_P_ ev_timer *w, int revents)
		36	{
		37	puts ("timeout");
		38	// this causes the innermost ev_loop to stop iterating
		39	ev_unloop (EV_A_ EVUNLOOP_ONE);
		40	}
		41
		42	int
		43	main (void)
		44	{
		45	// use the default event loop unless you have special needs
		46	ev_loop *loop = ev_default_loop (0);
		47
		48	// initialise an io watcher, then start it
		49	// this one will watch for stdin to become readable
		50	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		51	ev_io_start (loop, &stdin_watcher);
		52
		53	// initialise a timer watcher, then start it
		54	// simple non-repeating 5.5 second timeout
		55	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		56	ev_timer_start (loop, &timeout_watcher);
		57
		58	// now wait for events to arrive
		59	ev_loop (loop, 0);
		60
		61	// unloop was called, so exit
		62	return 0;
		63	}
8		64
9	=head1 DESCRIPTION	65	=head1 DESCRIPTION
10		66
		67	The newest version of this document is also available as an html-formatted
		68	web page you might find easier to navigate when reading it for the first
		69	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		70
11	Libev is an event loop: you register interest in certain events (such as a	71	Libev is an event loop: you register interest in certain events (such as a
12	file descriptor being readable or a timeout occuring), and it will manage	72	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	73	these event sources and provide your program with events.
14		74
15	To do this, it must take more or less complete control over your process	75	To do this, it must take more or less complete control over your process
16	(or thread) by executing the I<event loop> handler, and will then	76	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	77	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	79	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	80	watchers>, which are relatively small C structures you initialise with the
21	details of the event, and then hand it over to libev by I<starting> the	81	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	82	watcher.
23		83
24	=head1 FEATURES	84	=head2 FEATURES
25		85
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	86	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	87	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	88	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
29	events (related to SIGCHLD), and event watchers dealing with the event	89	(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	90	with customised rescheduling (C<ev_periodic>), synchronous signals
		91	(C<ev_signal>), process status change events (C<ev_child>), and event
		92	watchers dealing with the event loop mechanism itself (C<ev_idle>,
		93	C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as
		94	file watchers (C<ev_stat>) and even limited support for fork events
		95	(C<ev_fork>).
		96
		97	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	98	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	99	for example).
33		100
34	=head1 CONVENTIONS	101	=head2 CONVENTIONS
35		102
36	Libev is very configurable. In this manual the default configuration	103	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	104	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	105	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	106	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	107	for multiple event loops, then all functions taking an initial argument of
41	argument of name C<loop> (which is always of type C<struct ev_loop *>)	108	name C<loop> (which is always of type C<ev_loop *>) will not have
42	will not have this argument.	109	this argument.
43		110
44	=head1 TIME REPRESENTATION	111	=head2 TIME REPRESENTATION
45		112
46	Libev represents time as a single floating point number, representing the	113	Libev represents time as a single floating point number, representing the
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	114	(fractional) number of seconds since the (POSIX) epoch (somewhere near
48	the beginning of 1970, details are complicated, don't ask). This type is	115	the beginning of 1970, details are complicated, don't ask). This type is
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	116	called C<ev_tstamp>, which is what you should use too. It usually aliases
50	to the double type in C.	117	to the C<double> type in C, and when you need to do any calculations on
		118	it, you should treat it as some floating point value. Unlike the name
		119	component C<stamp> might indicate, it is also used for time differences
		120	throughout libev.
		121
		122	=head1 ERROR HANDLING
		123
		124	Libev knows three classes of errors: operating system errors, usage errors
		125	and internal errors (bugs).
		126
		127	When libev catches an operating system error it cannot handle (for example
		128	a system call indicating a condition libev cannot fix), it calls the callback
		129	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		130	abort. The default is to print a diagnostic message and to call C<abort
		131	()>.
		132
		133	When libev detects a usage error such as a negative timer interval, then
		134	it will print a diagnostic message and abort (via the C<assert> mechanism,
		135	so C<NDEBUG> will disable this checking): these are programming errors in
		136	the libev caller and need to be fixed there.
		137
		138	Libev also has a few internal error-checking C<assert>ions, and also has
		139	extensive consistency checking code. These do not trigger under normal
		140	circumstances, as they indicate either a bug in libev or worse.
		141
51		142
52	=head1 GLOBAL FUNCTIONS	143	=head1 GLOBAL FUNCTIONS
53		144
54	These functions can be called anytime, even before initialising the	145	These functions can be called anytime, even before initialising the
55	library in any way.	146	library in any way.
…		…
60		151
61	Returns the current time as libev would use it. Please note that the	152	Returns the current time as libev would use it. Please note that the
62	C<ev_now> function is usually faster and also often returns the timestamp	153	C<ev_now> function is usually faster and also often returns the timestamp
63	you actually want to know.	154	you actually want to know.
64		155
		156	=item ev_sleep (ev_tstamp interval)
		157
		158	Sleep for the given interval: The current thread will be blocked until
		159	either it is interrupted or the given time interval has passed. Basically
		160	this is a sub-second-resolution C<sleep ()>.
		161
65	=item int ev_version_major ()	162	=item int ev_version_major ()
66		163
67	=item int ev_version_minor ()	164	=item int ev_version_minor ()
68		165
69	You can find out the major and minor version numbers of the library	166	You can find out the major and minor ABI version numbers of the library
70	you linked against by calling the functions C<ev_version_major> and	167	you linked against by calling the functions C<ev_version_major> and
71	C<ev_version_minor>. If you want, you can compare against the global	168	C<ev_version_minor>. If you want, you can compare against the global
72	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	169	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
73	version of the library your program was compiled against.	170	version of the library your program was compiled against.
74		171
		172	These version numbers refer to the ABI version of the library, not the
		173	release version.
		174
75	Usually, it's a good idea to terminate if the major versions mismatch,	175	Usually, it's a good idea to terminate if the major versions mismatch,
76	as this indicates an incompatible change. Minor versions are usually	176	as this indicates an incompatible change. Minor versions are usually
77	compatible to older versions, so a larger minor version alone is usually	177	compatible to older versions, so a larger minor version alone is usually
78	not a problem.	178	not a problem.
79		179
		180	Example: Make sure we haven't accidentally been linked against the wrong
		181	version.
		182
		183	assert (("libev version mismatch",
		184	ev_version_major () == EV_VERSION_MAJOR
		185	&& ev_version_minor () >= EV_VERSION_MINOR));
		186
		187	=item unsigned int ev_supported_backends ()
		188
		189	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		190	value) compiled into this binary of libev (independent of their
		191	availability on the system you are running on). See C<ev_default_loop> for
		192	a description of the set values.
		193
		194	Example: make sure we have the epoll method, because yeah this is cool and
		195	a must have and can we have a torrent of it please!!!11
		196
		197	assert (("sorry, no epoll, no sex",
		198	ev_supported_backends () & EVBACKEND_EPOLL));
		199
		200	=item unsigned int ev_recommended_backends ()
		201
		202	Return the set of all backends compiled into this binary of libev and also
		203	recommended for this platform. This set is often smaller than the one
		204	returned by C<ev_supported_backends>, as for example kqueue is broken on
		205	most BSDs and will not be auto-detected unless you explicitly request it
		206	(assuming you know what you are doing). This is the set of backends that
		207	libev will probe for if you specify no backends explicitly.
		208
		209	=item unsigned int ev_embeddable_backends ()
		210
		211	Returns the set of backends that are embeddable in other event loops. This
		212	is the theoretical, all-platform, value. To find which backends
		213	might be supported on the current system, you would need to look at
		214	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
		215	recommended ones.
		216
		217	See the description of C<ev_embed> watchers for more info.
		218
80	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	219	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
81		220
82	Sets the allocation function to use (the prototype is similar to the	221	Sets the allocation function to use (the prototype is similar - the
83	realloc C function, the semantics are identical). It is used to allocate	222	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
84	and free memory (no surprises here). If it returns zero when memory	223	used to allocate and free memory (no surprises here). If it returns zero
85	needs to be allocated, the library might abort or take some potentially	224	when memory needs to be allocated (C<size != 0>), the library might abort
86	destructive action. The default is your system realloc function.	225	or take some potentially destructive action.
		226
		227	Since some systems (at least OpenBSD and Darwin) fail to implement
		228	correct C<realloc> semantics, libev will use a wrapper around the system
		229	C<realloc> and C<free> functions by default.
87		230
88	You could override this function in high-availability programs to, say,	231	You could override this function in high-availability programs to, say,
89	free some memory if it cannot allocate memory, to use a special allocator,	232	free some memory if it cannot allocate memory, to use a special allocator,
90	or even to sleep a while and retry until some memory is available.	233	or even to sleep a while and retry until some memory is available.
91		234
		235	Example: Replace the libev allocator with one that waits a bit and then
		236	retries (example requires a standards-compliant C<realloc>).
		237
		238	static void *
		239	persistent_realloc (void *ptr, size_t size)
		240	{
		241	for (;;)
		242	{
		243	void *newptr = realloc (ptr, size);
		244
		245	if (newptr)
		246	return newptr;
		247
		248	sleep (60);
		249	}
		250	}
		251
		252	...
		253	ev_set_allocator (persistent_realloc);
		254
92	=item ev_set_syserr_cb (void (*cb)(const char *msg));	255	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
93		256
94	Set the callback function to call on a retryable syscall error (such	257	Set the callback function to call on a retryable system call error (such
95	as failed select, poll, epoll_wait). The message is a printable string	258	as failed select, poll, epoll_wait). The message is a printable string
96	indicating the system call or subsystem causing the problem. If this	259	indicating the system call or subsystem causing the problem. If this
97	callback is set, then libev will expect it to remedy the sitution, no	260	callback is set, then libev will expect it to remedy the situation, no
98	matter what, when it returns. That is, libev will generally retry the	261	matter what, when it returns. That is, libev will generally retry the
99	requested operation, or, if the condition doesn't go away, do bad stuff	262	requested operation, or, if the condition doesn't go away, do bad stuff
100	(such as abort).	263	(such as abort).
101		264
		265	Example: This is basically the same thing that libev does internally, too.
		266
		267	static void
		268	fatal_error (const char *msg)
		269	{
		270	perror (msg);
		271	abort ();
		272	}
		273
		274	...
		275	ev_set_syserr_cb (fatal_error);
		276
102	=back	277	=back
103		278
104	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	279	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
105		280
106	An event loop is described by a C<struct ev_loop *>. The library knows two	281	An event loop is described by a C<ev_loop *>. The library knows two
107	types of such loops, the I<default> loop, which supports signals and child	282	types of such loops, the I<default> loop, which supports signals and child
108	events, and dynamically created loops which do not.	283	events, and dynamically created loops which do not.
109
110	If you use threads, a common model is to run the default event loop
111	in your main thread (or in a separate thread) and for each thread you
112	create, you also create another event loop. Libev itself does no locking
113	whatsoever, so if you mix calls to the same event loop in different
114	threads, make sure you lock (this is usually a bad idea, though, even if
115	done correctly, because it's hideous and inefficient).
116		284
117	=over 4	285	=over 4
118		286
119	=item struct ev_loop *ev_default_loop (unsigned int flags)	287	=item struct ev_loop *ev_default_loop (unsigned int flags)
120		288
121	This will initialise the default event loop if it hasn't been initialised	289	This will initialise the default event loop if it hasn't been initialised
122	yet and return it. If the default loop could not be initialised, returns	290	yet and return it. If the default loop could not be initialised, returns
123	false. If it already was initialised it simply returns it (and ignores the	291	false. If it already was initialised it simply returns it (and ignores the
124	flags).	292	flags. If that is troubling you, check C<ev_backend ()> afterwards).
125		293
126	If you don't know what event loop to use, use the one returned from this	294	If you don't know what event loop to use, use the one returned from this
127	function.	295	function.
128		296
		297	Note that this function is I<not> thread-safe, so if you want to use it
		298	from multiple threads, you have to lock (note also that this is unlikely,
		299	as loops cannot bes hared easily between threads anyway).
		300
		301	The default loop is the only loop that can handle C<ev_signal> and
		302	C<ev_child> watchers, and to do this, it always registers a handler
		303	for C<SIGCHLD>. If this is a problem for your application you can either
		304	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
		305	can simply overwrite the C<SIGCHLD> signal handler I<after> calling
		306	C<ev_default_init>.
		307
129	The flags argument can be used to specify special behaviour or specific	308	The flags argument can be used to specify special behaviour or specific
130	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	309	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
131		310
132	It supports the following flags:	311	The following flags are supported:
133		312
134	=over 4	313	=over 4
135		314
136	=item C<EVFLAG_AUTO>	315	=item C<EVFLAG_AUTO>
137		316
138	The default flags value. Use this if you have no clue (it's the right	317	The default flags value. Use this if you have no clue (it's the right
139	thing, believe me).	318	thing, believe me).
140		319
141	=item C<EVFLAG_NOENV>	320	=item C<EVFLAG_NOENV>
142		321
143	If this flag bit is ored into the flag value (or the program runs setuid	322	If this flag bit is or'ed into the flag value (or the program runs setuid
144	or setgid) then libev will I<not> look at the environment variable	323	or setgid) then libev will I<not> look at the environment variable
145	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	324	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
146	override the flags completely if it is found in the environment. This is	325	override the flags completely if it is found in the environment. This is
147	useful to try out specific backends to test their performance, or to work	326	useful to try out specific backends to test their performance, or to work
148	around bugs.	327	around bugs.
149		328
		329	=item C<EVFLAG_FORKCHECK>
		330
		331	Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after
		332	a fork, you can also make libev check for a fork in each iteration by
		333	enabling this flag.
		334
		335	This works by calling C<getpid ()> on every iteration of the loop,
		336	and thus this might slow down your event loop if you do a lot of loop
		337	iterations and little real work, but is usually not noticeable (on my
		338	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		339	without a system call and thus I<very> fast, but my GNU/Linux system also has
		340	C<pthread_atfork> which is even faster).
		341
		342	The big advantage of this flag is that you can forget about fork (and
		343	forget about forgetting to tell libev about forking) when you use this
		344	flag.
		345
		346	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		347	environment variable.
		348
150	=item C<EVMETHOD_SELECT> (portable select backend)	349	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
151		350
		351	This is your standard select(2) backend. Not I<completely> standard, as
		352	libev tries to roll its own fd_set with no limits on the number of fds,
		353	but if that fails, expect a fairly low limit on the number of fds when
		354	using this backend. It doesn't scale too well (O(highest_fd)), but its
		355	usually the fastest backend for a low number of (low-numbered :) fds.
		356
		357	To get good performance out of this backend you need a high amount of
		358	parallelism (most of the file descriptors should be busy). If you are
		359	writing a server, you should C<accept ()> in a loop to accept as many
		360	connections as possible during one iteration. You might also want to have
		361	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		362	readiness notifications you get per iteration.
		363
		364	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		365	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		366	C<exceptfds> set on that platform).
		367
152	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	368	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
153		369
154	=item C<EVMETHOD_EPOLL> (linux only)	370	And this is your standard poll(2) backend. It's more complicated
		371	than select, but handles sparse fds better and has no artificial
		372	limit on the number of fds you can use (except it will slow down
		373	considerably with a lot of inactive fds). It scales similarly to select,
		374	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		375	performance tips.
155		376
156	=item C<EVMETHOD_KQUEUE> (some bsds only)	377	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		378	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
157		379
158	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	380	=item C<EVBACKEND_EPOLL> (value 4, Linux)
159		381
160	=item C<EVMETHOD_PORT> (solaris 10 only)	382	For few fds, this backend is a bit little slower than poll and select,
		383	but it scales phenomenally better. While poll and select usually scale
		384	like O(total_fds) where n is the total number of fds (or the highest fd),
		385	epoll scales either O(1) or O(active_fds). The epoll design has a number
		386	of shortcomings, such as silently dropping events in some hard-to-detect
		387	cases and requiring a system call per fd change, no fork support and bad
		388	support for dup.
161		389
162	If one or more of these are ored into the flags value, then only these	390	While stopping, setting and starting an I/O watcher in the same iteration
163	backends will be tried (in the reverse order as given here). If one are	391	will result in some caching, there is still a system call per such incident
164	specified, any backend will do.	392	(because the fd could point to a different file description now), so its
		393	best to avoid that. Also, C<dup ()>'ed file descriptors might not work
		394	very well if you register events for both fds.
		395
		396	Please note that epoll sometimes generates spurious notifications, so you
		397	need to use non-blocking I/O or other means to avoid blocking when no data
		398	(or space) is available.
		399
		400	Best performance from this backend is achieved by not unregistering all
		401	watchers for a file descriptor until it has been closed, if possible,
		402	i.e. keep at least one watcher active per fd at all times. Stopping and
		403	starting a watcher (without re-setting it) also usually doesn't cause
		404	extra overhead.
		405
		406	While nominally embeddable in other event loops, this feature is broken in
		407	all kernel versions tested so far.
		408
		409	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		410	C<EVBACKEND_POLL>.
		411
		412	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		413
		414	Kqueue deserves special mention, as at the time of this writing, it was
		415	broken on all BSDs except NetBSD (usually it doesn't work reliably with
		416	anything but sockets and pipes, except on Darwin, where of course it's
		417	completely useless). For this reason it's not being "auto-detected" unless
		418	you explicitly specify it in the flags (i.e. using C<EVBACKEND_KQUEUE>) or
		419	libev was compiled on a known-to-be-good (-enough) system like NetBSD.
		420
		421	You still can embed kqueue into a normal poll or select backend and use it
		422	only for sockets (after having made sure that sockets work with kqueue on
		423	the target platform). See C<ev_embed> watchers for more info.
		424
		425	It scales in the same way as the epoll backend, but the interface to the
		426	kernel is more efficient (which says nothing about its actual speed, of
		427	course). While stopping, setting and starting an I/O watcher does never
		428	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		429	two event changes per incident. Support for C<fork ()> is very bad and it
		430	drops fds silently in similarly hard-to-detect cases.
		431
		432	This backend usually performs well under most conditions.
		433
		434	While nominally embeddable in other event loops, this doesn't work
		435	everywhere, so you might need to test for this. And since it is broken
		436	almost everywhere, you should only use it when you have a lot of sockets
		437	(for which it usually works), by embedding it into another event loop
		438	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and, did I mention it,
		439	using it only for sockets.
		440
		441	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		442	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		443	C<NOTE_EOF>.
		444
		445	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		446
		447	This is not implemented yet (and might never be, unless you send me an
		448	implementation). According to reports, C</dev/poll> only supports sockets
		449	and is not embeddable, which would limit the usefulness of this backend
		450	immensely.
		451
		452	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		453
		454	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		455	it's really slow, but it still scales very well (O(active_fds)).
		456
		457	Please note that Solaris event ports can deliver a lot of spurious
		458	notifications, so you need to use non-blocking I/O or other means to avoid
		459	blocking when no data (or space) is available.
		460
		461	While this backend scales well, it requires one system call per active
		462	file descriptor per loop iteration. For small and medium numbers of file
		463	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		464	might perform better.
		465
		466	On the positive side, with the exception of the spurious readiness
		467	notifications, this backend actually performed fully to specification
		468	in all tests and is fully embeddable, which is a rare feat among the
		469	OS-specific backends.
		470
		471	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		472	C<EVBACKEND_POLL>.
		473
		474	=item C<EVBACKEND_ALL>
		475
		476	Try all backends (even potentially broken ones that wouldn't be tried
		477	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		478	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		479
		480	It is definitely not recommended to use this flag.
165		481
166	=back	482	=back
		483
		484	If one or more of these are or'ed into the flags value, then only these
		485	backends will be tried (in the reverse order as listed here). If none are
		486	specified, all backends in C<ev_recommended_backends ()> will be tried.
		487
		488	Example: This is the most typical usage.
		489
		490	if (!ev_default_loop (0))
		491	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		492
		493	Example: Restrict libev to the select and poll backends, and do not allow
		494	environment settings to be taken into account:
		495
		496	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		497
		498	Example: Use whatever libev has to offer, but make sure that kqueue is
		499	used if available (warning, breaks stuff, best use only with your own
		500	private event loop and only if you know the OS supports your types of
		501	fds):
		502
		503	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
167		504
168	=item struct ev_loop *ev_loop_new (unsigned int flags)	505	=item struct ev_loop *ev_loop_new (unsigned int flags)
169		506
170	Similar to C<ev_default_loop>, but always creates a new event loop that is	507	Similar to C<ev_default_loop>, but always creates a new event loop that is
171	always distinct from the default loop. Unlike the default loop, it cannot	508	always distinct from the default loop. Unlike the default loop, it cannot
172	handle signal and child watchers, and attempts to do so will be greeted by	509	handle signal and child watchers, and attempts to do so will be greeted by
173	undefined behaviour (or a failed assertion if assertions are enabled).	510	undefined behaviour (or a failed assertion if assertions are enabled).
174		511
		512	Note that this function I<is> thread-safe, and the recommended way to use
		513	libev with threads is indeed to create one loop per thread, and using the
		514	default loop in the "main" or "initial" thread.
		515
		516	Example: Try to create a event loop that uses epoll and nothing else.
		517
		518	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		519	if (!epoller)
		520	fatal ("no epoll found here, maybe it hides under your chair");
		521
175	=item ev_default_destroy ()	522	=item ev_default_destroy ()
176		523
177	Destroys the default loop again (frees all memory and kernel state	524	Destroys the default loop again (frees all memory and kernel state
178	etc.). This stops all registered event watchers (by not touching them in	525	etc.). None of the active event watchers will be stopped in the normal
179	any way whatsoever, although you cannot rely on this :).	526	sense, so e.g. C<ev_is_active> might still return true. It is your
		527	responsibility to either stop all watchers cleanly yourself I<before>
		528	calling this function, or cope with the fact afterwards (which is usually
		529	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		530	for example).
		531
		532	Note that certain global state, such as signal state, will not be freed by
		533	this function, and related watchers (such as signal and child watchers)
		534	would need to be stopped manually.
		535
		536	In general it is not advisable to call this function except in the
		537	rare occasion where you really need to free e.g. the signal handling
		538	pipe fds. If you need dynamically allocated loops it is better to use
		539	C<ev_loop_new> and C<ev_loop_destroy>).
180		540
181	=item ev_loop_destroy (loop)	541	=item ev_loop_destroy (loop)
182		542
183	Like C<ev_default_destroy>, but destroys an event loop created by an	543	Like C<ev_default_destroy>, but destroys an event loop created by an
184	earlier call to C<ev_loop_new>.	544	earlier call to C<ev_loop_new>.
185		545
186	=item ev_default_fork ()	546	=item ev_default_fork ()
187		547
		548	This function sets a flag that causes subsequent C<ev_loop> iterations
188	This function reinitialises the kernel state for backends that have	549	to reinitialise the kernel state for backends that have one. Despite the
189	one. Despite the name, you can call it anytime, but it makes most sense	550	name, you can call it anytime, but it makes most sense after forking, in
190	after forking, in either the parent or child process (or both, but that	551	the child process (or both child and parent, but that again makes little
191	again makes little sense).	552	sense). You I<must> call it in the child before using any of the libev
		553	functions, and it will only take effect at the next C<ev_loop> iteration.
192		554
193	You I<must> call this function after forking if and only if you want to	555	On the other hand, you only need to call this function in the child
194	use the event library in both processes. If you just fork+exec, you don't	556	process if and only if you want to use the event library in the child. If
195	have to call it.	557	you just fork+exec, you don't have to call it at all.
196		558
197	The function itself is quite fast and it's usually not a problem to call	559	The function itself is quite fast and it's usually not a problem to call
198	it just in case after a fork. To make this easy, the function will fit in	560	it just in case after a fork. To make this easy, the function will fit in
199	quite nicely into a call to C<pthread_atfork>:	561	quite nicely into a call to C<pthread_atfork>:
200		562
…		…
202		564
203	=item ev_loop_fork (loop)	565	=item ev_loop_fork (loop)
204		566
205	Like C<ev_default_fork>, but acts on an event loop created by	567	Like C<ev_default_fork>, but acts on an event loop created by
206	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	568	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
207	after fork, and how you do this is entirely your own problem.	569	after fork that you want to re-use in the child, and how you do this is
		570	entirely your own problem.
208		571
		572	=item int ev_is_default_loop (loop)
		573
		574	Returns true when the given loop is, in fact, the default loop, and false
		575	otherwise.
		576
		577	=item unsigned int ev_loop_count (loop)
		578
		579	Returns the count of loop iterations for the loop, which is identical to
		580	the number of times libev did poll for new events. It starts at C<0> and
		581	happily wraps around with enough iterations.
		582
		583	This value can sometimes be useful as a generation counter of sorts (it
		584	"ticks" the number of loop iterations), as it roughly corresponds with
		585	C<ev_prepare> and C<ev_check> calls.
		586
209	=item unsigned int ev_method (loop)	587	=item unsigned int ev_backend (loop)
210		588
211	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	589	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
212	use.	590	use.
213		591
214	=item ev_tstamp ev_now (loop)	592	=item ev_tstamp ev_now (loop)
215		593
216	Returns the current "event loop time", which is the time the event loop	594	Returns the current "event loop time", which is the time the event loop
217	got events and started processing them. This timestamp does not change	595	received events and started processing them. This timestamp does not
218	as long as callbacks are being processed, and this is also the base time	596	change as long as callbacks are being processed, and this is also the base
219	used for relative timers. You can treat it as the timestamp of the event	597	time used for relative timers. You can treat it as the timestamp of the
220	occuring (or more correctly, the mainloop finding out about it).	598	event occurring (or more correctly, libev finding out about it).
		599
		600	=item ev_now_update (loop)
		601
		602	Establishes the current time by querying the kernel, updating the time
		603	returned by C<ev_now ()> in the progress. This is a costly operation and
		604	is usually done automatically within C<ev_loop ()>.
		605
		606	This function is rarely useful, but when some event callback runs for a
		607	very long time without entering the event loop, updating libev's idea of
		608	the current time is a good idea.
		609
		610	See also "The special problem of time updates" in the C<ev_timer> section.
221		611
222	=item ev_loop (loop, int flags)	612	=item ev_loop (loop, int flags)
223		613
224	Finally, this is it, the event handler. This function usually is called	614	Finally, this is it, the event handler. This function usually is called
225	after you initialised all your watchers and you want to start handling	615	after you initialised all your watchers and you want to start handling
226	events.	616	events.
227		617
228	If the flags argument is specified as 0, it will not return until either	618	If the flags argument is specified as C<0>, it will not return until
229	no event watchers are active anymore or C<ev_unloop> was called.	619	either no event watchers are active anymore or C<ev_unloop> was called.
		620
		621	Please note that an explicit C<ev_unloop> is usually better than
		622	relying on all watchers to be stopped when deciding when a program has
		623	finished (especially in interactive programs), but having a program
		624	that automatically loops as long as it has to and no longer by virtue
		625	of relying on its watchers stopping correctly, that is truly a thing of
		626	beauty.
230		627
231	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	628	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
232	those events and any outstanding ones, but will not block your process in	629	those events and any already outstanding ones, but will not block your
233	case there are no events and will return after one iteration of the loop.	630	process in case there are no events and will return after one iteration of
		631	the loop.
234		632
235	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	633	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
236	neccessary) and will handle those and any outstanding ones. It will block	634	necessary) and will handle those and any already outstanding ones. It
237	your process until at least one new event arrives, and will return after	635	will block your process until at least one new event arrives (which could
		636	be an event internal to libev itself, so there is no guarentee that a
		637	user-registered callback will be called), and will return after one
238	one iteration of the loop.	638	iteration of the loop.
239		639
240	This flags value could be used to implement alternative looping	640	This is useful if you are waiting for some external event in conjunction
241	constructs, but the C<prepare> and C<check> watchers provide a better and	641	with something not expressible using other libev watchers (i.e. "roll your
242	more generic mechanism.	642	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		643	usually a better approach for this kind of thing.
243		644
244	Here are the gory details of what ev_loop does:	645	Here are the gory details of what C<ev_loop> does:
245		646
246	1. If there are no active watchers (reference count is zero), return.	647	- Before the first iteration, call any pending watchers.
		648	* If EVFLAG_FORKCHECK was used, check for a fork.
		649	- If a fork was detected (by any means), queue and call all fork watchers.
247	2. Queue and immediately call all prepare watchers.	650	- Queue and call all prepare watchers.
248	3. If we have been forked, recreate the kernel state.	651	- If we have been forked, detach and recreate the kernel state
		652	as to not disturb the other process.
249	4. Update the kernel state with all outstanding changes.	653	- Update the kernel state with all outstanding changes.
250	5. Update the "event loop time".	654	- Update the "event loop time" (ev_now ()).
251	6. Calculate for how long to block.	655	- Calculate for how long to sleep or block, if at all
		656	(active idle watchers, EVLOOP_NONBLOCK or not having
		657	any active watchers at all will result in not sleeping).
		658	- Sleep if the I/O and timer collect interval say so.
252	7. Block the process, waiting for events.	659	- Block the process, waiting for any events.
		660	- Queue all outstanding I/O (fd) events.
253	8. Update the "event loop time" and do time jump handling.	661	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
254	9. Queue all outstanding timers.	662	- Queue all expired timers.
255	10. Queue all outstanding periodics.	663	- Queue all expired periodics.
256	11. If no events are pending now, queue all idle watchers.	664	- Unless any events are pending now, queue all idle watchers.
257	12. Queue all check watchers.	665	- Queue all check watchers.
258	13. Call all queued watchers in reverse order (i.e. check watchers first).	666	- Call all queued watchers in reverse order (i.e. check watchers first).
		667	Signals and child watchers are implemented as I/O watchers, and will
		668	be handled here by queueing them when their watcher gets executed.
259	14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	669	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
260	was used, return, otherwise continue with step #1.	670	were used, or there are no active watchers, return, otherwise
		671	continue with step *.
		672
		673	Example: Queue some jobs and then loop until no events are outstanding
		674	anymore.
		675
		676	... queue jobs here, make sure they register event watchers as long
		677	... as they still have work to do (even an idle watcher will do..)
		678	ev_loop (my_loop, 0);
		679	... jobs done or somebody called unloop. yeah!
261		680
262	=item ev_unloop (loop, how)	681	=item ev_unloop (loop, how)
263		682
264	Can be used to make a call to C<ev_loop> return early (but only after it	683	Can be used to make a call to C<ev_loop> return early (but only after it
265	has processed all outstanding events). The C<how> argument must be either	684	has processed all outstanding events). The C<how> argument must be either
266	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	685	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
267	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	686	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
268		687
		688	This "unloop state" will be cleared when entering C<ev_loop> again.
		689
		690	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
		691
269	=item ev_ref (loop)	692	=item ev_ref (loop)
270		693
271	=item ev_unref (loop)	694	=item ev_unref (loop)
272		695
273	Ref/unref can be used to add or remove a reference count on the event	696	Ref/unref can be used to add or remove a reference count on the event
274	loop: Every watcher keeps one reference, and as long as the reference	697	loop: Every watcher keeps one reference, and as long as the reference
275	count is nonzero, C<ev_loop> will not return on its own. If you have	698	count is nonzero, C<ev_loop> will not return on its own.
		699
276	a watcher you never unregister that should not keep C<ev_loop> from	700	If you have a watcher you never unregister that should not keep C<ev_loop>
277	returning, ev_unref() after starting, and ev_ref() before stopping it. For	701	from returning, call ev_unref() after starting, and ev_ref() before
		702	stopping it.
		703
278	example, libev itself uses this for its internal signal pipe: It is not	704	As an example, libev itself uses this for its internal signal pipe: It is
279	visible to the libev user and should not keep C<ev_loop> from exiting if	705	not visible to the libev user and should not keep C<ev_loop> from exiting
280	no event watchers registered by it are active. It is also an excellent	706	if no event watchers registered by it are active. It is also an excellent
281	way to do this for generic recurring timers or from within third-party	707	way to do this for generic recurring timers or from within third-party
282	libraries. Just remember to I<unref after start> and I<ref before stop>.	708	libraries. Just remember to I<unref after start> and I<ref before stop>
		709	(but only if the watcher wasn't active before, or was active before,
		710	respectively).
		711
		712	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		713	running when nothing else is active.
		714
		715	ev_signal exitsig;
		716	ev_signal_init (&exitsig, sig_cb, SIGINT);
		717	ev_signal_start (loop, &exitsig);
		718	evf_unref (loop);
		719
		720	Example: For some weird reason, unregister the above signal handler again.
		721
		722	ev_ref (loop);
		723	ev_signal_stop (loop, &exitsig);
		724
		725	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		726
		727	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		728
		729	These advanced functions influence the time that libev will spend waiting
		730	for events. Both time intervals are by default C<0>, meaning that libev
		731	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		732	latency.
		733
		734	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		735	allows libev to delay invocation of I/O and timer/periodic callbacks
		736	to increase efficiency of loop iterations (or to increase power-saving
		737	opportunities).
		738
		739	The idea is that sometimes your program runs just fast enough to handle
		740	one (or very few) event(s) per loop iteration. While this makes the
		741	program responsive, it also wastes a lot of CPU time to poll for new
		742	events, especially with backends like C<select ()> which have a high
		743	overhead for the actual polling but can deliver many events at once.
		744
		745	By setting a higher I<io collect interval> you allow libev to spend more
		746	time collecting I/O events, so you can handle more events per iteration,
		747	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		748	C<ev_timer>) will be not affected. Setting this to a non-null value will
		749	introduce an additional C<ev_sleep ()> call into most loop iterations.
		750
		751	Likewise, by setting a higher I<timeout collect interval> you allow libev
		752	to spend more time collecting timeouts, at the expense of increased
		753	latency/jitter/inexactness (the watcher callback will be called
		754	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		755	value will not introduce any overhead in libev.
		756
		757	Many (busy) programs can usually benefit by setting the I/O collect
		758	interval to a value near C<0.1> or so, which is often enough for
		759	interactive servers (of course not for games), likewise for timeouts. It
		760	usually doesn't make much sense to set it to a lower value than C<0.01>,
		761	as this approaches the timing granularity of most systems.
		762
		763	Setting the I<timeout collect interval> can improve the opportunity for
		764	saving power, as the program will "bundle" timer callback invocations that
		765	are "near" in time together, by delaying some, thus reducing the number of
		766	times the process sleeps and wakes up again. Another useful technique to
		767	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		768	they fire on, say, one-second boundaries only.
		769
		770	=item ev_loop_verify (loop)
		771
		772	This function only does something when C<EV_VERIFY> support has been
		773	compiled in. which is the default for non-minimal builds. It tries to go
		774	through all internal structures and checks them for validity. If anything
		775	is found to be inconsistent, it will print an error message to standard
		776	error and call C<abort ()>.
		777
		778	This can be used to catch bugs inside libev itself: under normal
		779	circumstances, this function will never abort as of course libev keeps its
		780	data structures consistent.
283		781
284	=back	782	=back
		783
285		784
286	=head1 ANATOMY OF A WATCHER	785	=head1 ANATOMY OF A WATCHER
287		786
288	A watcher is a structure that you create and register to record your	787	A watcher is a structure that you create and register to record your
289	interest in some event. For instance, if you want to wait for STDIN to	788	interest in some event. For instance, if you want to wait for STDIN to
290	become readable, you would create an C<ev_io> watcher for that:	789	become readable, you would create an C<ev_io> watcher for that:
291		790
292	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	791	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
293	{	792	{
294	ev_io_stop (w);	793	ev_io_stop (w);
295	ev_unloop (loop, EVUNLOOP_ALL);	794	ev_unloop (loop, EVUNLOOP_ALL);
296	}	795	}
297		796
298	struct ev_loop *loop = ev_default_loop (0);	797	struct ev_loop *loop = ev_default_loop (0);
299	struct ev_io stdin_watcher;	798	ev_io stdin_watcher;
300	ev_init (&stdin_watcher, my_cb);	799	ev_init (&stdin_watcher, my_cb);
301	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	800	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
302	ev_io_start (loop, &stdin_watcher);	801	ev_io_start (loop, &stdin_watcher);
303	ev_loop (loop, 0);	802	ev_loop (loop, 0);
304		803
305	As you can see, you are responsible for allocating the memory for your	804	As you can see, you are responsible for allocating the memory for your
306	watcher structures (and it is usually a bad idea to do this on the stack,	805	watcher structures (and it is usually a bad idea to do this on the stack,
307	although this can sometimes be quite valid).	806	although this can sometimes be quite valid).
308		807
309	Each watcher structure must be initialised by a call to C<ev_init	808	Each watcher structure must be initialised by a call to C<ev_init
310	(watcher *, callback)>, which expects a callback to be provided. This	809	(watcher *, callback)>, which expects a callback to be provided. This
311	callback gets invoked each time the event occurs (or, in the case of io	810	callback gets invoked each time the event occurs (or, in the case of I/O
312	watchers, each time the event loop detects that the file descriptor given	811	watchers, each time the event loop detects that the file descriptor given
313	is readable and/or writable).	812	is readable and/or writable).
314		813
315	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	814	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro
316	with arguments specific to this watcher type. There is also a macro	815	with arguments specific to this watcher type. There is also a macro
…		…
322	*) >>), and you can stop watching for events at any time by calling the	821	*) >>), and you can stop watching for events at any time by calling the
323	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	822	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.
324		823
325	As long as your watcher is active (has been started but not stopped) you	824	As long as your watcher is active (has been started but not stopped) you
326	must not touch the values stored in it. Most specifically you must never	825	must not touch the values stored in it. Most specifically you must never
327	reinitialise it or call its set method.	826	reinitialise it or call its C<set> macro.
328
329	You can check whether an event is active by calling the C<ev_is_active
330	(watcher *)> macro. To see whether an event is outstanding (but the
331	callback for it has not been called yet) you can use the C<ev_is_pending
332	(watcher *)> macro.
333		827
334	Each and every callback receives the event loop pointer as first, the	828	Each and every callback receives the event loop pointer as first, the
335	registered watcher structure as second, and a bitset of received events as	829	registered watcher structure as second, and a bitset of received events as
336	third argument.	830	third argument.
337		831
…		…
361	The signal specified in the C<ev_signal> watcher has been received by a thread.	855	The signal specified in the C<ev_signal> watcher has been received by a thread.
362		856
363	=item C<EV_CHILD>	857	=item C<EV_CHILD>
364		858
365	The pid specified in the C<ev_child> watcher has received a status change.	859	The pid specified in the C<ev_child> watcher has received a status change.
		860
		861	=item C<EV_STAT>
		862
		863	The path specified in the C<ev_stat> watcher changed its attributes somehow.
366		864
367	=item C<EV_IDLE>	865	=item C<EV_IDLE>
368		866
369	The C<ev_idle> watcher has determined that you have nothing better to do.	867	The C<ev_idle> watcher has determined that you have nothing better to do.
370		868
…		…
378	received events. Callbacks of both watcher types can start and stop as	876	received events. Callbacks of both watcher types can start and stop as
379	many watchers as they want, and all of them will be taken into account	877	many watchers as they want, and all of them will be taken into account
380	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	878	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
381	C<ev_loop> from blocking).	879	C<ev_loop> from blocking).
382		880
		881	=item C<EV_EMBED>
		882
		883	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		884
		885	=item C<EV_FORK>
		886
		887	The event loop has been resumed in the child process after fork (see
		888	C<ev_fork>).
		889
		890	=item C<EV_ASYNC>
		891
		892	The given async watcher has been asynchronously notified (see C<ev_async>).
		893
383	=item C<EV_ERROR>	894	=item C<EV_ERROR>
384		895
385	An unspecified error has occured, the watcher has been stopped. This might	896	An unspecified error has occurred, the watcher has been stopped. This might
386	happen because the watcher could not be properly started because libev	897	happen because the watcher could not be properly started because libev
387	ran out of memory, a file descriptor was found to be closed or any other	898	ran out of memory, a file descriptor was found to be closed or any other
		899	problem. Libev considers these application bugs.
		900
388	problem. You best act on it by reporting the problem and somehow coping	901	You best act on it by reporting the problem and somehow coping with the
389	with the watcher being stopped.	902	watcher being stopped. Note that well-written programs should not receive
		903	an error ever, so when your watcher receives it, this usually indicates a
		904	bug in your program.
390		905
391	Libev will usually signal a few "dummy" events together with an error,	906	Libev will usually signal a few "dummy" events together with an error, for
392	for example it might indicate that a fd is readable or writable, and if	907	example it might indicate that a fd is readable or writable, and if your
393	your callbacks is well-written it can just attempt the operation and cope	908	callbacks is well-written it can just attempt the operation and cope with
394	with the error from read() or write(). This will not work in multithreaded	909	the error from read() or write(). This will not work in multi-threaded
395	programs, though, so beware.	910	programs, though, as the fd could already be closed and reused for another
		911	thing, so beware.
396		912
397	=back	913	=back
398		914
		915	=head2 GENERIC WATCHER FUNCTIONS
		916
		917	In the following description, C<TYPE> stands for the watcher type,
		918	e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
		919
		920	=over 4
		921
		922	=item C<ev_init> (ev_TYPE *watcher, callback)
		923
		924	This macro initialises the generic portion of a watcher. The contents
		925	of the watcher object can be arbitrary (so C<malloc> will do). Only
		926	the generic parts of the watcher are initialised, you I<need> to call
		927	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		928	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		929	which rolls both calls into one.
		930
		931	You can reinitialise a watcher at any time as long as it has been stopped
		932	(or never started) and there are no pending events outstanding.
		933
		934	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		935	int revents)>.
		936
		937	Example: Initialise an C<ev_io> watcher in two steps.
		938
		939	ev_io w;
		940	ev_init (&w, my_cb);
		941	ev_io_set (&w, STDIN_FILENO, EV_READ);
		942
		943	=item C<ev_TYPE_set> (ev_TYPE *, [args])
		944
		945	This macro initialises the type-specific parts of a watcher. You need to
		946	call C<ev_init> at least once before you call this macro, but you can
		947	call C<ev_TYPE_set> any number of times. You must not, however, call this
		948	macro on a watcher that is active (it can be pending, however, which is a
		949	difference to the C<ev_init> macro).
		950
		951	Although some watcher types do not have type-specific arguments
		952	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		953
		954	See C<ev_init>, above, for an example.
		955
		956	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		957
		958	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		959	calls into a single call. This is the most convenient method to initialise
		960	a watcher. The same limitations apply, of course.
		961
		962	Example: Initialise and set an C<ev_io> watcher in one step.
		963
		964	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		965
		966	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
		967
		968	Starts (activates) the given watcher. Only active watchers will receive
		969	events. If the watcher is already active nothing will happen.
		970
		971	Example: Start the C<ev_io> watcher that is being abused as example in this
		972	whole section.
		973
		974	ev_io_start (EV_DEFAULT_UC, &w);
		975
		976	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
		977
		978	Stops the given watcher if active, and clears the pending status (whether
		979	the watcher was active or not).
		980
		981	It is possible that stopped watchers are pending - for example,
		982	non-repeating timers are being stopped when they become pending - but
		983	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		984	pending. If you want to free or reuse the memory used by the watcher it is
		985	therefore a good idea to always call its C<ev_TYPE_stop> function.
		986
		987	=item bool ev_is_active (ev_TYPE *watcher)
		988
		989	Returns a true value iff the watcher is active (i.e. it has been started
		990	and not yet been stopped). As long as a watcher is active you must not modify
		991	it.
		992
		993	=item bool ev_is_pending (ev_TYPE *watcher)
		994
		995	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		996	events but its callback has not yet been invoked). As long as a watcher
		997	is pending (but not active) you must not call an init function on it (but
		998	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		999	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1000	it).
		1001
		1002	=item callback ev_cb (ev_TYPE *watcher)
		1003
		1004	Returns the callback currently set on the watcher.
		1005
		1006	=item ev_cb_set (ev_TYPE *watcher, callback)
		1007
		1008	Change the callback. You can change the callback at virtually any time
		1009	(modulo threads).
		1010
		1011	=item ev_set_priority (ev_TYPE *watcher, priority)
		1012
		1013	=item int ev_priority (ev_TYPE *watcher)
		1014
		1015	Set and query the priority of the watcher. The priority is a small
		1016	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1017	(default: C<-2>). Pending watchers with higher priority will be invoked
		1018	before watchers with lower priority, but priority will not keep watchers
		1019	from being executed (except for C<ev_idle> watchers).
		1020
		1021	This means that priorities are I<only> used for ordering callback
		1022	invocation after new events have been received. This is useful, for
		1023	example, to reduce latency after idling, or more often, to bind two
		1024	watchers on the same event and make sure one is called first.
		1025
		1026	If you need to suppress invocation when higher priority events are pending
		1027	you need to look at C<ev_idle> watchers, which provide this functionality.
		1028
		1029	You I<must not> change the priority of a watcher as long as it is active or
		1030	pending.
		1031
		1032	The default priority used by watchers when no priority has been set is
		1033	always C<0>, which is supposed to not be too high and not be too low :).
		1034
		1035	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1036	fine, as long as you do not mind that the priority value you query might
		1037	or might not have been adjusted to be within valid range.
		1038
		1039	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1040
		1041	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1042	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1043	can deal with that fact, as both are simply passed through to the
		1044	callback.
		1045
		1046	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1047
		1048	If the watcher is pending, this function clears its pending status and
		1049	returns its C<revents> bitset (as if its callback was invoked). If the
		1050	watcher isn't pending it does nothing and returns C<0>.
		1051
		1052	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1053	callback to be invoked, which can be accomplished with this function.
		1054
		1055	=back
		1056
		1057
399	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1058	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
400		1059
401	Each watcher has, by default, a member C<void *data> that you can change	1060	Each watcher has, by default, a member C<void *data> that you can change
402	and read at any time, libev will completely ignore it. This can be used	1061	and read at any time: libev will completely ignore it. This can be used
403	to associate arbitrary data with your watcher. If you need more data and	1062	to associate arbitrary data with your watcher. If you need more data and
404	don't want to allocate memory and store a pointer to it in that data	1063	don't want to allocate memory and store a pointer to it in that data
405	member, you can also "subclass" the watcher type and provide your own	1064	member, you can also "subclass" the watcher type and provide your own
406	data:	1065	data:
407		1066
408	struct my_io	1067	struct my_io
409	{	1068	{
410	struct ev_io io;	1069	ev_io io;
411	int otherfd;	1070	int otherfd;
412	void *somedata;	1071	void *somedata;
413	struct whatever *mostinteresting;	1072	struct whatever *mostinteresting;
414	}	1073	};
		1074
		1075	...
		1076	struct my_io w;
		1077	ev_io_init (&w.io, my_cb, fd, EV_READ);
415		1078
416	And since your callback will be called with a pointer to the watcher, you	1079	And since your callback will be called with a pointer to the watcher, you
417	can cast it back to your own type:	1080	can cast it back to your own type:
418		1081
419	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1082	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
420	{	1083	{
421	struct my_io *w = (struct my_io *)w_;	1084	struct my_io *w = (struct my_io *)w_;
422	...	1085	...
423	}	1086	}
424		1087
425	More interesting and less C-conformant ways of catsing your callback type	1088	More interesting and less C-conformant ways of casting your callback type
426	have been omitted....	1089	instead have been omitted.
		1090
		1091	Another common scenario is to use some data structure with multiple
		1092	embedded watchers:
		1093
		1094	struct my_biggy
		1095	{
		1096	int some_data;
		1097	ev_timer t1;
		1098	ev_timer t2;
		1099	}
		1100
		1101	In this case getting the pointer to C<my_biggy> is a bit more
		1102	complicated: Either you store the address of your C<my_biggy> struct
		1103	in the C<data> member of the watcher (for woozies), or you need to use
		1104	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1105	programmers):
		1106
		1107	#include <stddef.h>
		1108
		1109	static void
		1110	t1_cb (EV_P_ ev_timer *w, int revents)
		1111	{
		1112	struct my_biggy big = (struct my_biggy *
		1113	(((char *)w) - offsetof (struct my_biggy, t1));
		1114	}
		1115
		1116	static void
		1117	t2_cb (EV_P_ ev_timer *w, int revents)
		1118	{
		1119	struct my_biggy big = (struct my_biggy *
		1120	(((char *)w) - offsetof (struct my_biggy, t2));
		1121	}
427		1122
428		1123
429	=head1 WATCHER TYPES	1124	=head1 WATCHER TYPES
430		1125
431	This section describes each watcher in detail, but will not repeat	1126	This section describes each watcher in detail, but will not repeat
432	information given in the last section.	1127	information given in the last section. Any initialisation/set macros,
		1128	functions and members specific to the watcher type are explained.
433		1129
		1130	Members are additionally marked with either I<[read-only]>, meaning that,
		1131	while the watcher is active, you can look at the member and expect some
		1132	sensible content, but you must not modify it (you can modify it while the
		1133	watcher is stopped to your hearts content), or I<[read-write]>, which
		1134	means you can expect it to have some sensible content while the watcher
		1135	is active, but you can also modify it. Modifying it may not do something
		1136	sensible or take immediate effect (or do anything at all), but libev will
		1137	not crash or malfunction in any way.
		1138
		1139
434	=head2 C<ev_io> - is this file descriptor readable or writable	1140	=head2 C<ev_io> - is this file descriptor readable or writable?
435		1141
436	I/O watchers check whether a file descriptor is readable or writable	1142	I/O watchers check whether a file descriptor is readable or writable
437	in each iteration of the event loop (This behaviour is called	1143	in each iteration of the event loop, or, more precisely, when reading
438	level-triggering because you keep receiving events as long as the	1144	would not block the process and writing would at least be able to write
439	condition persists. Remember you can stop the watcher if you don't want to	1145	some data. This behaviour is called level-triggering because you keep
440	act on the event and neither want to receive future events).	1146	receiving events as long as the condition persists. Remember you can stop
		1147	the watcher if you don't want to act on the event and neither want to
		1148	receive future events.
441		1149
442	In general you can register as many read and/or write event watchers per	1150	In general you can register as many read and/or write event watchers per
443	fd as you want (as long as you don't confuse yourself). Setting all file	1151	fd as you want (as long as you don't confuse yourself). Setting all file
444	descriptors to non-blocking mode is also usually a good idea (but not	1152	descriptors to non-blocking mode is also usually a good idea (but not
445	required if you know what you are doing).	1153	required if you know what you are doing).
446		1154
447	You have to be careful with dup'ed file descriptors, though. Some backends	1155	If you cannot use non-blocking mode, then force the use of a
448	(the linux epoll backend is a notable example) cannot handle dup'ed file	1156	known-to-be-good backend (at the time of this writing, this includes only
449	descriptors correctly if you register interest in two or more fds pointing	1157	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).
450	to the same underlying file/socket etc. description (that is, they share
451	the same underlying "file open").
452		1158
453	If you must do this, then force the use of a known-to-be-good backend	1159	Another thing you have to watch out for is that it is quite easy to
454	(at the time of this writing, this includes only EVMETHOD_SELECT and	1160	receive "spurious" readiness notifications, that is your callback might
455	EVMETHOD_POLL).	1161	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1162	because there is no data. Not only are some backends known to create a
		1163	lot of those (for example Solaris ports), it is very easy to get into
		1164	this situation even with a relatively standard program structure. Thus
		1165	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1166	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1167
		1168	If you cannot run the fd in non-blocking mode (for example you should
		1169	not play around with an Xlib connection), then you have to separately
		1170	re-test whether a file descriptor is really ready with a known-to-be good
		1171	interface such as poll (fortunately in our Xlib example, Xlib already
		1172	does this on its own, so its quite safe to use). Some people additionally
		1173	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1174	indefinitely.
		1175
		1176	But really, best use non-blocking mode.
		1177
		1178	=head3 The special problem of disappearing file descriptors
		1179
		1180	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1181	descriptor (either due to calling C<close> explicitly or any other means,
		1182	such as C<dup2>). The reason is that you register interest in some file
		1183	descriptor, but when it goes away, the operating system will silently drop
		1184	this interest. If another file descriptor with the same number then is
		1185	registered with libev, there is no efficient way to see that this is, in
		1186	fact, a different file descriptor.
		1187
		1188	To avoid having to explicitly tell libev about such cases, libev follows
		1189	the following policy: Each time C<ev_io_set> is being called, libev
		1190	will assume that this is potentially a new file descriptor, otherwise
		1191	it is assumed that the file descriptor stays the same. That means that
		1192	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1193	descriptor even if the file descriptor number itself did not change.
		1194
		1195	This is how one would do it normally anyway, the important point is that
		1196	the libev application should not optimise around libev but should leave
		1197	optimisations to libev.
		1198
		1199	=head3 The special problem of dup'ed file descriptors
		1200
		1201	Some backends (e.g. epoll), cannot register events for file descriptors,
		1202	but only events for the underlying file descriptions. That means when you
		1203	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1204	events for them, only one file descriptor might actually receive events.
		1205
		1206	There is no workaround possible except not registering events
		1207	for potentially C<dup ()>'ed file descriptors, or to resort to
		1208	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1209
		1210	=head3 The special problem of fork
		1211
		1212	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1213	useless behaviour. Libev fully supports fork, but needs to be told about
		1214	it in the child.
		1215
		1216	To support fork in your programs, you either have to call
		1217	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1218	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1219	C<EVBACKEND_POLL>.
		1220
		1221	=head3 The special problem of SIGPIPE
		1222
		1223	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1224	when writing to a pipe whose other end has been closed, your program gets
		1225	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1226	this is sensible behaviour, for daemons, this is usually undesirable.
		1227
		1228	So when you encounter spurious, unexplained daemon exits, make sure you
		1229	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1230	somewhere, as that would have given you a big clue).
		1231
		1232
		1233	=head3 Watcher-Specific Functions
456		1234
457	=over 4	1235	=over 4
458		1236
459	=item ev_io_init (ev_io *, callback, int fd, int events)	1237	=item ev_io_init (ev_io *, callback, int fd, int events)
460		1238
461	=item ev_io_set (ev_io *, int fd, int events)	1239	=item ev_io_set (ev_io *, int fd, int events)
462		1240
463	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1241	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
464	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1242	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
465	EV_WRITE> to receive the given events.	1243	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1244
		1245	=item int fd [read-only]
		1246
		1247	The file descriptor being watched.
		1248
		1249	=item int events [read-only]
		1250
		1251	The events being watched.
466		1252
467	=back	1253	=back
468		1254
		1255	=head3 Examples
		1256
		1257	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1258	readable, but only once. Since it is likely line-buffered, you could
		1259	attempt to read a whole line in the callback.
		1260
		1261	static void
		1262	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1263	{
		1264	ev_io_stop (loop, w);
		1265	.. read from stdin here (or from w->fd) and handle any I/O errors
		1266	}
		1267
		1268	...
		1269	struct ev_loop *loop = ev_default_init (0);
		1270	ev_io stdin_readable;
		1271	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1272	ev_io_start (loop, &stdin_readable);
		1273	ev_loop (loop, 0);
		1274
		1275
469	=head2 C<ev_timer> - relative and optionally recurring timeouts	1276	=head2 C<ev_timer> - relative and optionally repeating timeouts
470		1277
471	Timer watchers are simple relative timers that generate an event after a	1278	Timer watchers are simple relative timers that generate an event after a
472	given time, and optionally repeating in regular intervals after that.	1279	given time, and optionally repeating in regular intervals after that.
473		1280
474	The timers are based on real time, that is, if you register an event that	1281	The timers are based on real time, that is, if you register an event that
475	times out after an hour and you reset your system clock to last years	1282	times out after an hour and you reset your system clock to January last
476	time, it will still time out after (roughly) and hour. "Roughly" because	1283	year, it will still time out after (roughly) one hour. "Roughly" because
477	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1284	detecting time jumps is hard, and some inaccuracies are unavoidable (the
478	monotonic clock option helps a lot here).	1285	monotonic clock option helps a lot here).
		1286
		1287	The callback is guaranteed to be invoked only I<after> its timeout has
		1288	passed, but if multiple timers become ready during the same loop iteration
		1289	then order of execution is undefined.
		1290
		1291	=head3 Be smart about timeouts
		1292
		1293	Many real-world problems involve some kind of timeout, usually for error
		1294	recovery. A typical example is an HTTP request - if the other side hangs,
		1295	you want to raise some error after a while.
		1296
		1297	What follows are some ways to handle this problem, from obvious and
		1298	inefficient to smart and efficient.
		1299
		1300	In the following, a 60 second activity timeout is assumed - a timeout that
		1301	gets reset to 60 seconds each time there is activity (e.g. each time some
		1302	data or other life sign was received).
		1303
		1304	=over 4
		1305
		1306	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1307
		1308	This is the most obvious, but not the most simple way: In the beginning,
		1309	start the watcher:
		1310
		1311	ev_timer_init (timer, callback, 60., 0.);
		1312	ev_timer_start (loop, timer);
		1313
		1314	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1315	and start it again:
		1316
		1317	ev_timer_stop (loop, timer);
		1318	ev_timer_set (timer, 60., 0.);
		1319	ev_timer_start (loop, timer);
		1320
		1321	This is relatively simple to implement, but means that each time there is
		1322	some activity, libev will first have to remove the timer from its internal
		1323	data structure and then add it again. Libev tries to be fast, but it's
		1324	still not a constant-time operation.
		1325
		1326	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1327
		1328	This is the easiest way, and involves using C<ev_timer_again> instead of
		1329	C<ev_timer_start>.
		1330
		1331	To implement this, configure an C<ev_timer> with a C<repeat> value
		1332	of C<60> and then call C<ev_timer_again> at start and each time you
		1333	successfully read or write some data. If you go into an idle state where
		1334	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1335	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1336
		1337	That means you can ignore both the C<ev_timer_start> function and the
		1338	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1339	member and C<ev_timer_again>.
		1340
		1341	At start:
		1342
		1343	ev_timer_init (timer, callback);
		1344	timer->repeat = 60.;
		1345	ev_timer_again (loop, timer);
		1346
		1347	Each time there is some activity:
		1348
		1349	ev_timer_again (loop, timer);
		1350
		1351	It is even possible to change the time-out on the fly, regardless of
		1352	whether the watcher is active or not:
		1353
		1354	timer->repeat = 30.;
		1355	ev_timer_again (loop, timer);
		1356
		1357	This is slightly more efficient then stopping/starting the timer each time
		1358	you want to modify its timeout value, as libev does not have to completely
		1359	remove and re-insert the timer from/into its internal data structure.
		1360
		1361	It is, however, even simpler than the "obvious" way to do it.
		1362
		1363	=item 3. Let the timer time out, but then re-arm it as required.
		1364
		1365	This method is more tricky, but usually most efficient: Most timeouts are
		1366	relatively long compared to the intervals between other activity - in
		1367	our example, within 60 seconds, there are usually many I/O events with
		1368	associated activity resets.
		1369
		1370	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1371	but remember the time of last activity, and check for a real timeout only
		1372	within the callback:
		1373
		1374	ev_tstamp last_activity; // time of last activity
		1375
		1376	static void
		1377	callback (EV_P_ ev_timer *w, int revents)
		1378	{
		1379	ev_tstamp now = ev_now (EV_A);
		1380	ev_tstamp timeout = last_activity + 60.;
		1381
		1382	// if last_activity + 60. is older than now, we did time out
		1383	if (timeout < now)
		1384	{
		1385	// timeout occured, take action
		1386	}
		1387	else
		1388	{
		1389	// callback was invoked, but there was some activity, re-arm
		1390	// the watcher to fire in last_activity + 60, which is
		1391	// guaranteed to be in the future, so "again" is positive:
		1392	w->again = timeout - now;
		1393	ev_timer_again (EV_A_ w);
		1394	}
		1395	}
		1396
		1397	To summarise the callback: first calculate the real timeout (defined
		1398	as "60 seconds after the last activity"), then check if that time has
		1399	been reached, which means something I<did>, in fact, time out. Otherwise
		1400	the callback was invoked too early (C<timeout> is in the future), so
		1401	re-schedule the timer to fire at that future time, to see if maybe we have
		1402	a timeout then.
		1403
		1404	Note how C<ev_timer_again> is used, taking advantage of the
		1405	C<ev_timer_again> optimisation when the timer is already running.
		1406
		1407	This scheme causes more callback invocations (about one every 60 seconds
		1408	minus half the average time between activity), but virtually no calls to
		1409	libev to change the timeout.
		1410
		1411	To start the timer, simply initialise the watcher and set C<last_activity>
		1412	to the current time (meaning we just have some activity :), then call the
		1413	callback, which will "do the right thing" and start the timer:
		1414
		1415	ev_timer_init (timer, callback);
		1416	last_activity = ev_now (loop);
		1417	callback (loop, timer, EV_TIMEOUT);
		1418
		1419	And when there is some activity, simply store the current time in
		1420	C<last_activity>, no libev calls at all:
		1421
		1422	last_actiivty = ev_now (loop);
		1423
		1424	This technique is slightly more complex, but in most cases where the
		1425	time-out is unlikely to be triggered, much more efficient.
		1426
		1427	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1428	callback :) - just change the timeout and invoke the callback, which will
		1429	fix things for you.
		1430
		1431	=item 4. Whee, use a double-linked list for your timeouts.
		1432
		1433	If there is not one request, but many thousands, all employing some kind
		1434	of timeout with the same timeout value, then one can do even better:
		1435
		1436	When starting the timeout, calculate the timeout value and put the timeout
		1437	at the I<end> of the list.
		1438
		1439	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1440	the list is expected to fire (for example, using the technique #3).
		1441
		1442	When there is some activity, remove the timer from the list, recalculate
		1443	the timeout, append it to the end of the list again, and make sure to
		1444	update the C<ev_timer> if it was taken from the beginning of the list.
		1445
		1446	This way, one can manage an unlimited number of timeouts in O(1) time for
		1447	starting, stopping and updating the timers, at the expense of a major
		1448	complication, and having to use a constant timeout. The constant timeout
		1449	ensures that the list stays sorted.
		1450
		1451	=back
		1452
		1453	So what method is the best?
		1454
		1455	The method #2 is a simple no-brain-required solution that is adequate in
		1456	most situations. Method #3 requires a bit more thinking, but handles many
		1457	cases better, and isn't very complicated either. In most case, choosing
		1458	either one is fine.
		1459
		1460	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1461	rather complicated, but extremely efficient, something that really pays
		1462	off after the first or so million of active timers, i.e. it's usually
		1463	overkill :)
		1464
		1465	=head3 The special problem of time updates
		1466
		1467	Establishing the current time is a costly operation (it usually takes at
		1468	least two system calls): EV therefore updates its idea of the current
		1469	time only before and after C<ev_loop> collects new events, which causes a
		1470	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1471	lots of events in one iteration.
479		1472
480	The relative timeouts are calculated relative to the C<ev_now ()>	1473	The relative timeouts are calculated relative to the C<ev_now ()>
481	time. This is usually the right thing as this timestamp refers to the time	1474	time. This is usually the right thing as this timestamp refers to the time
482	of the event triggering whatever timeout you are modifying/starting. If	1475	of the event triggering whatever timeout you are modifying/starting. If
483	you suspect event processing to be delayed and you *need* to base the timeout	1476	you suspect event processing to be delayed and you I<need> to base the
484	on the current time, use something like this to adjust for this:	1477	timeout on the current time, use something like this to adjust for this:
485		1478
486	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1479	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
487		1480
		1481	If the event loop is suspended for a long time, you can also force an
		1482	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1483	()>.
		1484
		1485	=head3 Watcher-Specific Functions and Data Members
		1486
488	=over 4	1487	=over 4
489		1488
490	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1489	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
491		1490
492	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1491	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
493		1492
494	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1493	Configure the timer to trigger after C<after> seconds. If C<repeat>
495	C<0.>, then it will automatically be stopped. If it is positive, then the	1494	is C<0.>, then it will automatically be stopped once the timeout is
496	timer will automatically be configured to trigger again C<repeat> seconds	1495	reached. If it is positive, then the timer will automatically be
497	later, again, and again, until stopped manually.	1496	configured to trigger again C<repeat> seconds later, again, and again,
		1497	until stopped manually.
498		1498
499	The timer itself will do a best-effort at avoiding drift, that is, if you	1499	The timer itself will do a best-effort at avoiding drift, that is, if
500	configure a timer to trigger every 10 seconds, then it will trigger at	1500	you configure a timer to trigger every 10 seconds, then it will normally
501	exactly 10 second intervals. If, however, your program cannot keep up with	1501	trigger at exactly 10 second intervals. If, however, your program cannot
502	the timer (because it takes longer than those 10 seconds to do stuff) the	1502	keep up with the timer (because it takes longer than those 10 seconds to
503	timer will not fire more than once per event loop iteration.	1503	do stuff) the timer will not fire more than once per event loop iteration.
504		1504
505	=item ev_timer_again (loop)	1505	=item ev_timer_again (loop, ev_timer *)
506		1506
507	This will act as if the timer timed out and restart it again if it is	1507	This will act as if the timer timed out and restart it again if it is
508	repeating. The exact semantics are:	1508	repeating. The exact semantics are:
509		1509
		1510	If the timer is pending, its pending status is cleared.
		1511
510	If the timer is started but nonrepeating, stop it.	1512	If the timer is started but non-repeating, stop it (as if it timed out).
511		1513
512	If the timer is repeating, either start it if necessary (with the repeat	1514	If the timer is repeating, either start it if necessary (with the
513	value), or reset the running timer to the repeat value.	1515	C<repeat> value), or reset the running timer to the C<repeat> value.
514		1516
515	This sounds a bit complicated, but here is a useful and typical	1517	This sounds a bit complicated, see "Be smart about timeouts", above, for a
516	example: Imagine you have a tcp connection and you want a so-called idle	1518	usage example.
517	timeout, that is, you want to be called when there have been, say, 60	1519
518	seconds of inactivity on the socket. The easiest way to do this is to	1520	=item ev_tstamp repeat [read-write]
519	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	1521
520	time you successfully read or write some data. If you go into an idle	1522	The current C<repeat> value. Will be used each time the watcher times out
521	state where you do not expect data to travel on the socket, you can stop	1523	or C<ev_timer_again> is called, and determines the next timeout (if any),
522	the timer, and again will automatically restart it if need be.	1524	which is also when any modifications are taken into account.
523		1525
524	=back	1526	=back
525		1527
		1528	=head3 Examples
		1529
		1530	Example: Create a timer that fires after 60 seconds.
		1531
		1532	static void
		1533	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1534	{
		1535	.. one minute over, w is actually stopped right here
		1536	}
		1537
		1538	ev_timer mytimer;
		1539	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1540	ev_timer_start (loop, &mytimer);
		1541
		1542	Example: Create a timeout timer that times out after 10 seconds of
		1543	inactivity.
		1544
		1545	static void
		1546	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1547	{
		1548	.. ten seconds without any activity
		1549	}
		1550
		1551	ev_timer mytimer;
		1552	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1553	ev_timer_again (&mytimer); /* start timer */
		1554	ev_loop (loop, 0);
		1555
		1556	// and in some piece of code that gets executed on any "activity":
		1557	// reset the timeout to start ticking again at 10 seconds
		1558	ev_timer_again (&mytimer);
		1559
		1560
526	=head2 C<ev_periodic> - to cron or not to cron	1561	=head2 C<ev_periodic> - to cron or not to cron?
527		1562
528	Periodic watchers are also timers of a kind, but they are very versatile	1563	Periodic watchers are also timers of a kind, but they are very versatile
529	(and unfortunately a bit complex).	1564	(and unfortunately a bit complex).
530		1565
531	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1566	Unlike C<ev_timer>'s, they are not based on real time (or relative time)
532	but on wallclock time (absolute time). You can tell a periodic watcher	1567	but on wall clock time (absolute time). You can tell a periodic watcher
533	to trigger "at" some specific point in time. For example, if you tell a	1568	to trigger after some specific point in time. For example, if you tell a
534	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1569	periodic watcher to trigger in 10 seconds (by specifying e.g. C<ev_now ()
535	+ 10.>) and then reset your system clock to the last year, then it will	1570	+ 10.>, that is, an absolute time not a delay) and then reset your system
		1571	clock to January of the previous year, then it will take more than year
536	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	1572	to trigger the event (unlike an C<ev_timer>, which would still trigger
537	roughly 10 seconds later and of course not if you reset your system time	1573	roughly 10 seconds later as it uses a relative timeout).
538	again).
539		1574
540	They can also be used to implement vastly more complex timers, such as	1575	C<ev_periodic>s can also be used to implement vastly more complex timers,
541	triggering an event on eahc midnight, local time.	1576	such as triggering an event on each "midnight, local time", or other
		1577	complicated rules.
		1578
		1579	As with timers, the callback is guaranteed to be invoked only when the
		1580	time (C<at>) has passed, but if multiple periodic timers become ready
		1581	during the same loop iteration, then order of execution is undefined.
		1582
		1583	=head3 Watcher-Specific Functions and Data Members
542		1584
543	=over 4	1585	=over 4
544		1586
545	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1587	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
546		1588
547	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1589	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
548		1590
549	Lots of arguments, lets sort it out... There are basically three modes of	1591	Lots of arguments, lets sort it out... There are basically three modes of
550	operation, and we will explain them from simplest to complex:	1592	operation, and we will explain them from simplest to most complex:
551
552		1593
553	=over 4	1594	=over 4
554		1595
555	=item * absolute timer (interval = reschedule_cb = 0)	1596	=item * absolute timer (at = time, interval = reschedule_cb = 0)
556		1597
557	In this configuration the watcher triggers an event at the wallclock time	1598	In this configuration the watcher triggers an event after the wall clock
558	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1599	time C<at> has passed. It will not repeat and will not adjust when a time
559	that is, if it is to be run at January 1st 2011 then it will run when the	1600	jump occurs, that is, if it is to be run at January 1st 2011 then it will
560	system time reaches or surpasses this time.	1601	only run when the system clock reaches or surpasses this time.
561		1602
562	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	1603	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
563		1604
564	In this mode the watcher will always be scheduled to time out at the next	1605	In this mode the watcher will always be scheduled to time out at the next
565	C<at + N * interval> time (for some integer N) and then repeat, regardless	1606	C<at + N * interval> time (for some integer N, which can also be negative)
566	of any time jumps.	1607	and then repeat, regardless of any time jumps.
567		1608
568	This can be used to create timers that do not drift with respect to system	1609	This can be used to create timers that do not drift with respect to the
569	time:	1610	system clock, for example, here is a C<ev_periodic> that triggers each
		1611	hour, on the hour:
570		1612
571	ev_periodic_set (&periodic, 0., 3600., 0);	1613	ev_periodic_set (&periodic, 0., 3600., 0);
572		1614
573	This doesn't mean there will always be 3600 seconds in between triggers,	1615	This doesn't mean there will always be 3600 seconds in between triggers,
574	but only that the the callback will be called when the system time shows a	1616	but only that the callback will be called when the system time shows a
575	full hour (UTC), or more correctly, when the system time is evenly divisible	1617	full hour (UTC), or more correctly, when the system time is evenly divisible
576	by 3600.	1618	by 3600.
577		1619
578	Another way to think about it (for the mathematically inclined) is that	1620	Another way to think about it (for the mathematically inclined) is that
579	C<ev_periodic> will try to run the callback in this mode at the next possible	1621	C<ev_periodic> will try to run the callback in this mode at the next possible
580	time where C<time = at (mod interval)>, regardless of any time jumps.	1622	time where C<time = at (mod interval)>, regardless of any time jumps.
581		1623
		1624	For numerical stability it is preferable that the C<at> value is near
		1625	C<ev_now ()> (the current time), but there is no range requirement for
		1626	this value, and in fact is often specified as zero.
		1627
		1628	Note also that there is an upper limit to how often a timer can fire (CPU
		1629	speed for example), so if C<interval> is very small then timing stability
		1630	will of course deteriorate. Libev itself tries to be exact to be about one
		1631	millisecond (if the OS supports it and the machine is fast enough).
		1632
582	=item * manual reschedule mode (reschedule_cb = callback)	1633	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
583		1634
584	In this mode the values for C<interval> and C<at> are both being	1635	In this mode the values for C<interval> and C<at> are both being
585	ignored. Instead, each time the periodic watcher gets scheduled, the	1636	ignored. Instead, each time the periodic watcher gets scheduled, the
586	reschedule callback will be called with the watcher as first, and the	1637	reschedule callback will be called with the watcher as first, and the
587	current time as second argument.	1638	current time as second argument.
588		1639
589	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	1640	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
590	ever, or make any event loop modifications>. If you need to stop it,	1641	ever, or make ANY event loop modifications whatsoever>.
591	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
592	starting a prepare watcher).
593		1642
		1643	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		1644	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		1645	only event loop modification you are allowed to do).
		1646
594	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	1647	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
595	ev_tstamp now)>, e.g.:	1648	*w, ev_tstamp now)>, e.g.:
596		1649
		1650	static ev_tstamp
597	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	1651	my_rescheduler (ev_periodic *w, ev_tstamp now)
598	{	1652	{
599	return now + 60.;	1653	return now + 60.;
600	}	1654	}
601		1655
602	It must return the next time to trigger, based on the passed time value	1656	It must return the next time to trigger, based on the passed time value
603	(that is, the lowest time value larger than to the second argument). It	1657	(that is, the lowest time value larger than to the second argument). It
604	will usually be called just before the callback will be triggered, but	1658	will usually be called just before the callback will be triggered, but
605	might be called at other times, too.	1659	might be called at other times, too.
606		1660
607	NOTE: I<< This callback must always return a time that is later than the	1661	NOTE: I<< This callback must always return a time that is higher than or
608	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	1662	equal to the passed C<now> value >>.
609		1663
610	This can be used to create very complex timers, such as a timer that	1664	This can be used to create very complex timers, such as a timer that
611	triggers on each midnight, local time. To do this, you would calculate the	1665	triggers on "next midnight, local time". To do this, you would calculate the
612	next midnight after C<now> and return the timestamp value for this. How	1666	next midnight after C<now> and return the timestamp value for this. How
613	you do this is, again, up to you (but it is not trivial, which is the main	1667	you do this is, again, up to you (but it is not trivial, which is the main
614	reason I omitted it as an example).	1668	reason I omitted it as an example).
615		1669
616	=back	1670	=back
…		…
620	Simply stops and restarts the periodic watcher again. This is only useful	1674	Simply stops and restarts the periodic watcher again. This is only useful
621	when you changed some parameters or the reschedule callback would return	1675	when you changed some parameters or the reschedule callback would return
622	a different time than the last time it was called (e.g. in a crond like	1676	a different time than the last time it was called (e.g. in a crond like
623	program when the crontabs have changed).	1677	program when the crontabs have changed).
624		1678
		1679	=item ev_tstamp ev_periodic_at (ev_periodic *)
		1680
		1681	When active, returns the absolute time that the watcher is supposed to
		1682	trigger next.
		1683
		1684	=item ev_tstamp offset [read-write]
		1685
		1686	When repeating, this contains the offset value, otherwise this is the
		1687	absolute point in time (the C<at> value passed to C<ev_periodic_set>).
		1688
		1689	Can be modified any time, but changes only take effect when the periodic
		1690	timer fires or C<ev_periodic_again> is being called.
		1691
		1692	=item ev_tstamp interval [read-write]
		1693
		1694	The current interval value. Can be modified any time, but changes only
		1695	take effect when the periodic timer fires or C<ev_periodic_again> is being
		1696	called.
		1697
		1698	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		1699
		1700	The current reschedule callback, or C<0>, if this functionality is
		1701	switched off. Can be changed any time, but changes only take effect when
		1702	the periodic timer fires or C<ev_periodic_again> is being called.
		1703
625	=back	1704	=back
626		1705
		1706	=head3 Examples
		1707
		1708	Example: Call a callback every hour, or, more precisely, whenever the
		1709	system time is divisible by 3600. The callback invocation times have
		1710	potentially a lot of jitter, but good long-term stability.
		1711
		1712	static void
		1713	clock_cb (struct ev_loop *loop, ev_io *w, int revents)
		1714	{
		1715	... its now a full hour (UTC, or TAI or whatever your clock follows)
		1716	}
		1717
		1718	ev_periodic hourly_tick;
		1719	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		1720	ev_periodic_start (loop, &hourly_tick);
		1721
		1722	Example: The same as above, but use a reschedule callback to do it:
		1723
		1724	#include <math.h>
		1725
		1726	static ev_tstamp
		1727	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		1728	{
		1729	return now + (3600. - fmod (now, 3600.));
		1730	}
		1731
		1732	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		1733
		1734	Example: Call a callback every hour, starting now:
		1735
		1736	ev_periodic hourly_tick;
		1737	ev_periodic_init (&hourly_tick, clock_cb,
		1738	fmod (ev_now (loop), 3600.), 3600., 0);
		1739	ev_periodic_start (loop, &hourly_tick);
		1740
		1741
627	=head2 C<ev_signal> - signal me when a signal gets signalled	1742	=head2 C<ev_signal> - signal me when a signal gets signalled!
628		1743
629	Signal watchers will trigger an event when the process receives a specific	1744	Signal watchers will trigger an event when the process receives a specific
630	signal one or more times. Even though signals are very asynchronous, libev	1745	signal one or more times. Even though signals are very asynchronous, libev
631	will try it's best to deliver signals synchronously, i.e. as part of the	1746	will try it's best to deliver signals synchronously, i.e. as part of the
632	normal event processing, like any other event.	1747	normal event processing, like any other event.
633		1748
		1749	If you want signals asynchronously, just use C<sigaction> as you would
		1750	do without libev and forget about sharing the signal. You can even use
		1751	C<ev_async> from a signal handler to synchronously wake up an event loop.
		1752
634	You can configure as many watchers as you like per signal. Only when the	1753	You can configure as many watchers as you like per signal. Only when the
635	first watcher gets started will libev actually register a signal watcher	1754	first watcher gets started will libev actually register a signal handler
636	with the kernel (thus it coexists with your own signal handlers as long	1755	with the kernel (thus it coexists with your own signal handlers as long as
637	as you don't register any with libev). Similarly, when the last signal	1756	you don't register any with libev for the same signal). Similarly, when
638	watcher for a signal is stopped libev will reset the signal handler to	1757	the last signal watcher for a signal is stopped, libev will reset the
639	SIG_DFL (regardless of what it was set to before).	1758	signal handler to SIG_DFL (regardless of what it was set to before).
		1759
		1760	If possible and supported, libev will install its handlers with
		1761	C<SA_RESTART> behaviour enabled, so system calls should not be unduly
		1762	interrupted. If you have a problem with system calls getting interrupted by
		1763	signals you can block all signals in an C<ev_check> watcher and unblock
		1764	them in an C<ev_prepare> watcher.
		1765
		1766	=head3 Watcher-Specific Functions and Data Members
640		1767
641	=over 4	1768	=over 4
642		1769
643	=item ev_signal_init (ev_signal *, callback, int signum)	1770	=item ev_signal_init (ev_signal *, callback, int signum)
644		1771
645	=item ev_signal_set (ev_signal *, int signum)	1772	=item ev_signal_set (ev_signal *, int signum)
646		1773
647	Configures the watcher to trigger on the given signal number (usually one	1774	Configures the watcher to trigger on the given signal number (usually one
648	of the C<SIGxxx> constants).	1775	of the C<SIGxxx> constants).
649		1776
		1777	=item int signum [read-only]
		1778
		1779	The signal the watcher watches out for.
		1780
650	=back	1781	=back
651		1782
		1783	=head3 Examples
		1784
		1785	Example: Try to exit cleanly on SIGINT.
		1786
		1787	static void
		1788	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		1789	{
		1790	ev_unloop (loop, EVUNLOOP_ALL);
		1791	}
		1792
		1793	ev_signal signal_watcher;
		1794	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		1795	ev_signal_start (loop, &signal_watcher);
		1796
		1797
652	=head2 C<ev_child> - wait for pid status changes	1798	=head2 C<ev_child> - watch out for process status changes
653		1799
654	Child watchers trigger when your process receives a SIGCHLD in response to	1800	Child watchers trigger when your process receives a SIGCHLD in response to
655	some child status changes (most typically when a child of yours dies).	1801	some child status changes (most typically when a child of yours dies or
		1802	exits). It is permissible to install a child watcher I<after> the child
		1803	has been forked (which implies it might have already exited), as long
		1804	as the event loop isn't entered (or is continued from a watcher), i.e.,
		1805	forking and then immediately registering a watcher for the child is fine,
		1806	but forking and registering a watcher a few event loop iterations later is
		1807	not.
		1808
		1809	Only the default event loop is capable of handling signals, and therefore
		1810	you can only register child watchers in the default event loop.
		1811
		1812	=head3 Process Interaction
		1813
		1814	Libev grabs C<SIGCHLD> as soon as the default event loop is
		1815	initialised. This is necessary to guarantee proper behaviour even if
		1816	the first child watcher is started after the child exits. The occurrence
		1817	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		1818	synchronously as part of the event loop processing. Libev always reaps all
		1819	children, even ones not watched.
		1820
		1821	=head3 Overriding the Built-In Processing
		1822
		1823	Libev offers no special support for overriding the built-in child
		1824	processing, but if your application collides with libev's default child
		1825	handler, you can override it easily by installing your own handler for
		1826	C<SIGCHLD> after initialising the default loop, and making sure the
		1827	default loop never gets destroyed. You are encouraged, however, to use an
		1828	event-based approach to child reaping and thus use libev's support for
		1829	that, so other libev users can use C<ev_child> watchers freely.
		1830
		1831	=head3 Stopping the Child Watcher
		1832
		1833	Currently, the child watcher never gets stopped, even when the
		1834	child terminates, so normally one needs to stop the watcher in the
		1835	callback. Future versions of libev might stop the watcher automatically
		1836	when a child exit is detected.
		1837
		1838	=head3 Watcher-Specific Functions and Data Members
656		1839
657	=over 4	1840	=over 4
658		1841
659	=item ev_child_init (ev_child *, callback, int pid)	1842	=item ev_child_init (ev_child *, callback, int pid, int trace)
660		1843
661	=item ev_child_set (ev_child *, int pid)	1844	=item ev_child_set (ev_child *, int pid, int trace)
662		1845
663	Configures the watcher to wait for status changes of process C<pid> (or	1846	Configures the watcher to wait for status changes of process C<pid> (or
664	I<any> process if C<pid> is specified as C<0>). The callback can look	1847	I<any> process if C<pid> is specified as C<0>). The callback can look
665	at the C<rstatus> member of the C<ev_child> watcher structure to see	1848	at the C<rstatus> member of the C<ev_child> watcher structure to see
666	the status word (use the macros from C<sys/wait.h> and see your systems	1849	the status word (use the macros from C<sys/wait.h> and see your systems
667	C<waitpid> documentation). The C<rpid> member contains the pid of the	1850	C<waitpid> documentation). The C<rpid> member contains the pid of the
668	process causing the status change.	1851	process causing the status change. C<trace> must be either C<0> (only
		1852	activate the watcher when the process terminates) or C<1> (additionally
		1853	activate the watcher when the process is stopped or continued).
		1854
		1855	=item int pid [read-only]
		1856
		1857	The process id this watcher watches out for, or C<0>, meaning any process id.
		1858
		1859	=item int rpid [read-write]
		1860
		1861	The process id that detected a status change.
		1862
		1863	=item int rstatus [read-write]
		1864
		1865	The process exit/trace status caused by C<rpid> (see your systems
		1866	C<waitpid> and C<sys/wait.h> documentation for details).
669		1867
670	=back	1868	=back
671		1869
		1870	=head3 Examples
		1871
		1872	Example: C<fork()> a new process and install a child handler to wait for
		1873	its completion.
		1874
		1875	ev_child cw;
		1876
		1877	static void
		1878	child_cb (EV_P_ ev_child *w, int revents)
		1879	{
		1880	ev_child_stop (EV_A_ w);
		1881	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		1882	}
		1883
		1884	pid_t pid = fork ();
		1885
		1886	if (pid < 0)
		1887	// error
		1888	else if (pid == 0)
		1889	{
		1890	// the forked child executes here
		1891	exit (1);
		1892	}
		1893	else
		1894	{
		1895	ev_child_init (&cw, child_cb, pid, 0);
		1896	ev_child_start (EV_DEFAULT_ &cw);
		1897	}
		1898
		1899
		1900	=head2 C<ev_stat> - did the file attributes just change?
		1901
		1902	This watches a file system path for attribute changes. That is, it calls
		1903	C<stat> regularly (or when the OS says it changed) and sees if it changed
		1904	compared to the last time, invoking the callback if it did.
		1905
		1906	The path does not need to exist: changing from "path exists" to "path does
		1907	not exist" is a status change like any other. The condition "path does
		1908	not exist" is signified by the C<st_nlink> field being zero (which is
		1909	otherwise always forced to be at least one) and all the other fields of
		1910	the stat buffer having unspecified contents.
		1911
		1912	The path I<should> be absolute and I<must not> end in a slash. If it is
		1913	relative and your working directory changes, the behaviour is undefined.
		1914
		1915	Since there is no standard kernel interface to do this, the portable
		1916	implementation simply calls C<stat (2)> regularly on the path to see if
		1917	it changed somehow. You can specify a recommended polling interval for
		1918	this case. If you specify a polling interval of C<0> (highly recommended!)
		1919	then a I<suitable, unspecified default> value will be used (which
		1920	you can expect to be around five seconds, although this might change
		1921	dynamically). Libev will also impose a minimum interval which is currently
		1922	around C<0.1>, but thats usually overkill.
		1923
		1924	This watcher type is not meant for massive numbers of stat watchers,
		1925	as even with OS-supported change notifications, this can be
		1926	resource-intensive.
		1927
		1928	At the time of this writing, the only OS-specific interface implemented
		1929	is the Linux inotify interface (implementing kqueue support is left as
		1930	an exercise for the reader. Note, however, that the author sees no way
		1931	of implementing C<ev_stat> semantics with kqueue).
		1932
		1933	=head3 ABI Issues (Largefile Support)
		1934
		1935	Libev by default (unless the user overrides this) uses the default
		1936	compilation environment, which means that on systems with large file
		1937	support disabled by default, you get the 32 bit version of the stat
		1938	structure. When using the library from programs that change the ABI to
		1939	use 64 bit file offsets the programs will fail. In that case you have to
		1940	compile libev with the same flags to get binary compatibility. This is
		1941	obviously the case with any flags that change the ABI, but the problem is
		1942	most noticeably disabled with ev_stat and large file support.
		1943
		1944	The solution for this is to lobby your distribution maker to make large
		1945	file interfaces available by default (as e.g. FreeBSD does) and not
		1946	optional. Libev cannot simply switch on large file support because it has
		1947	to exchange stat structures with application programs compiled using the
		1948	default compilation environment.
		1949
		1950	=head3 Inotify and Kqueue
		1951
		1952	When C<inotify (7)> support has been compiled into libev (generally
		1953	only available with Linux 2.6.25 or above due to bugs in earlier
		1954	implementations) and present at runtime, it will be used to speed up
		1955	change detection where possible. The inotify descriptor will be created
		1956	lazily when the first C<ev_stat> watcher is being started.
		1957
		1958	Inotify presence does not change the semantics of C<ev_stat> watchers
		1959	except that changes might be detected earlier, and in some cases, to avoid
		1960	making regular C<stat> calls. Even in the presence of inotify support
		1961	there are many cases where libev has to resort to regular C<stat> polling,
		1962	but as long as the path exists, libev usually gets away without polling.
		1963
		1964	There is no support for kqueue, as apparently it cannot be used to
		1965	implement this functionality, due to the requirement of having a file
		1966	descriptor open on the object at all times, and detecting renames, unlinks
		1967	etc. is difficult.
		1968
		1969	=head3 The special problem of stat time resolution
		1970
		1971	The C<stat ()> system call only supports full-second resolution portably, and
		1972	even on systems where the resolution is higher, most file systems still
		1973	only support whole seconds.
		1974
		1975	That means that, if the time is the only thing that changes, you can
		1976	easily miss updates: on the first update, C<ev_stat> detects a change and
		1977	calls your callback, which does something. When there is another update
		1978	within the same second, C<ev_stat> will be unable to detect unless the
		1979	stat data does change in other ways (e.g. file size).
		1980
		1981	The solution to this is to delay acting on a change for slightly more
		1982	than a second (or till slightly after the next full second boundary), using
		1983	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		1984	ev_timer_again (loop, w)>).
		1985
		1986	The C<.02> offset is added to work around small timing inconsistencies
		1987	of some operating systems (where the second counter of the current time
		1988	might be be delayed. One such system is the Linux kernel, where a call to
		1989	C<gettimeofday> might return a timestamp with a full second later than
		1990	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		1991	update file times then there will be a small window where the kernel uses
		1992	the previous second to update file times but libev might already execute
		1993	the timer callback).
		1994
		1995	=head3 Watcher-Specific Functions and Data Members
		1996
		1997	=over 4
		1998
		1999	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2000
		2001	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2002
		2003	Configures the watcher to wait for status changes of the given
		2004	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2005	be detected and should normally be specified as C<0> to let libev choose
		2006	a suitable value. The memory pointed to by C<path> must point to the same
		2007	path for as long as the watcher is active.
		2008
		2009	The callback will receive an C<EV_STAT> event when a change was detected,
		2010	relative to the attributes at the time the watcher was started (or the
		2011	last change was detected).
		2012
		2013	=item ev_stat_stat (loop, ev_stat *)
		2014
		2015	Updates the stat buffer immediately with new values. If you change the
		2016	watched path in your callback, you could call this function to avoid
		2017	detecting this change (while introducing a race condition if you are not
		2018	the only one changing the path). Can also be useful simply to find out the
		2019	new values.
		2020
		2021	=item ev_statdata attr [read-only]
		2022
		2023	The most-recently detected attributes of the file. Although the type is
		2024	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2025	suitable for your system, but you can only rely on the POSIX-standardised
		2026	members to be present. If the C<st_nlink> member is C<0>, then there was
		2027	some error while C<stat>ing the file.
		2028
		2029	=item ev_statdata prev [read-only]
		2030
		2031	The previous attributes of the file. The callback gets invoked whenever
		2032	C<prev> != C<attr>, or, more precisely, one or more of these members
		2033	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2034	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2035
		2036	=item ev_tstamp interval [read-only]
		2037
		2038	The specified interval.
		2039
		2040	=item const char *path [read-only]
		2041
		2042	The file system path that is being watched.
		2043
		2044	=back
		2045
		2046	=head3 Examples
		2047
		2048	Example: Watch C</etc/passwd> for attribute changes.
		2049
		2050	static void
		2051	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2052	{
		2053	/* /etc/passwd changed in some way */
		2054	if (w->attr.st_nlink)
		2055	{
		2056	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2057	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2058	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2059	}
		2060	else
		2061	/* you shalt not abuse printf for puts */
		2062	puts ("wow, /etc/passwd is not there, expect problems. "
		2063	"if this is windows, they already arrived\n");
		2064	}
		2065
		2066	...
		2067	ev_stat passwd;
		2068
		2069	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2070	ev_stat_start (loop, &passwd);
		2071
		2072	Example: Like above, but additionally use a one-second delay so we do not
		2073	miss updates (however, frequent updates will delay processing, too, so
		2074	one might do the work both on C<ev_stat> callback invocation I<and> on
		2075	C<ev_timer> callback invocation).
		2076
		2077	static ev_stat passwd;
		2078	static ev_timer timer;
		2079
		2080	static void
		2081	timer_cb (EV_P_ ev_timer *w, int revents)
		2082	{
		2083	ev_timer_stop (EV_A_ w);
		2084
		2085	/* now it's one second after the most recent passwd change */
		2086	}
		2087
		2088	static void
		2089	stat_cb (EV_P_ ev_stat *w, int revents)
		2090	{
		2091	/* reset the one-second timer */
		2092	ev_timer_again (EV_A_ &timer);
		2093	}
		2094
		2095	...
		2096	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2097	ev_stat_start (loop, &passwd);
		2098	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2099
		2100
672	=head2 C<ev_idle> - when you've got nothing better to do	2101	=head2 C<ev_idle> - when you've got nothing better to do...
673		2102
674	Idle watchers trigger events when there are no other events are pending	2103	Idle watchers trigger events when no other events of the same or higher
675	(prepare, check and other idle watchers do not count). That is, as long	2104	priority are pending (prepare, check and other idle watchers do not count
676	as your process is busy handling sockets or timeouts (or even signals,	2105	as receiving "events").
677	imagine) it will not be triggered. But when your process is idle all idle	2106
678	watchers are being called again and again, once per event loop iteration -	2107	That is, as long as your process is busy handling sockets or timeouts
		2108	(or even signals, imagine) of the same or higher priority it will not be
		2109	triggered. But when your process is idle (or only lower-priority watchers
		2110	are pending), the idle watchers are being called once per event loop
679	until stopped, that is, or your process receives more events and becomes	2111	iteration - until stopped, that is, or your process receives more events
680	busy.	2112	and becomes busy again with higher priority stuff.
681		2113
682	The most noteworthy effect is that as long as any idle watchers are	2114	The most noteworthy effect is that as long as any idle watchers are
683	active, the process will not block when waiting for new events.	2115	active, the process will not block when waiting for new events.
684		2116
685	Apart from keeping your process non-blocking (which is a useful	2117	Apart from keeping your process non-blocking (which is a useful
686	effect on its own sometimes), idle watchers are a good place to do	2118	effect on its own sometimes), idle watchers are a good place to do
687	"pseudo-background processing", or delay processing stuff to after the	2119	"pseudo-background processing", or delay processing stuff to after the
688	event loop has handled all outstanding events.	2120	event loop has handled all outstanding events.
689		2121
		2122	=head3 Watcher-Specific Functions and Data Members
		2123
690	=over 4	2124	=over 4
691		2125
692	=item ev_idle_init (ev_signal *, callback)	2126	=item ev_idle_init (ev_signal *, callback)
693		2127
694	Initialises and configures the idle watcher - it has no parameters of any	2128	Initialises and configures the idle watcher - it has no parameters of any
695	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2129	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
696	believe me.	2130	believe me.
697		2131
698	=back	2132	=back
699		2133
		2134	=head3 Examples
		2135
		2136	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2137	callback, free it. Also, use no error checking, as usual.
		2138
		2139	static void
		2140	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2141	{
		2142	free (w);
		2143	// now do something you wanted to do when the program has
		2144	// no longer anything immediate to do.
		2145	}
		2146
		2147	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2148	ev_idle_init (idle_watcher, idle_cb);
		2149	ev_idle_start (loop, idle_cb);
		2150
		2151
700	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2152	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
701		2153
702	Prepare and check watchers are usually (but not always) used in tandem:	2154	Prepare and check watchers are usually (but not always) used in pairs:
703	prepare watchers get invoked before the process blocks and check watchers	2155	prepare watchers get invoked before the process blocks and check watchers
704	afterwards.	2156	afterwards.
705		2157
		2158	You I<must not> call C<ev_loop> or similar functions that enter
		2159	the current event loop from either C<ev_prepare> or C<ev_check>
		2160	watchers. Other loops than the current one are fine, however. The
		2161	rationale behind this is that you do not need to check for recursion in
		2162	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2163	C<ev_check> so if you have one watcher of each kind they will always be
		2164	called in pairs bracketing the blocking call.
		2165
706	Their main purpose is to integrate other event mechanisms into libev. This	2166	Their main purpose is to integrate other event mechanisms into libev and
707	could be used, for example, to track variable changes, implement your own	2167	their use is somewhat advanced. They could be used, for example, to track
708	watchers, integrate net-snmp or a coroutine library and lots more.	2168	variable changes, implement your own watchers, integrate net-snmp or a
		2169	coroutine library and lots more. They are also occasionally useful if
		2170	you cache some data and want to flush it before blocking (for example,
		2171	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2172	watcher).
709		2173
710	This is done by examining in each prepare call which file descriptors need	2174	This is done by examining in each prepare call which file descriptors
711	to be watched by the other library, registering C<ev_io> watchers for	2175	need to be watched by the other library, registering C<ev_io> watchers
712	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2176	for them and starting an C<ev_timer> watcher for any timeouts (many
713	provide just this functionality). Then, in the check watcher you check for	2177	libraries provide exactly this functionality). Then, in the check watcher,
714	any events that occured (by checking the pending status of all watchers	2178	you check for any events that occurred (by checking the pending status
715	and stopping them) and call back into the library. The I/O and timer	2179	of all watchers and stopping them) and call back into the library. The
716	callbacks will never actually be called (but must be valid nevertheless,	2180	I/O and timer callbacks will never actually be called (but must be valid
717	because you never know, you know?).	2181	nevertheless, because you never know, you know?).
718		2182
719	As another example, the Perl Coro module uses these hooks to integrate	2183	As another example, the Perl Coro module uses these hooks to integrate
720	coroutines into libev programs, by yielding to other active coroutines	2184	coroutines into libev programs, by yielding to other active coroutines
721	during each prepare and only letting the process block if no coroutines	2185	during each prepare and only letting the process block if no coroutines
722	are ready to run (it's actually more complicated: it only runs coroutines	2186	are ready to run (it's actually more complicated: it only runs coroutines
723	with priority higher than or equal to the event loop and one coroutine	2187	with priority higher than or equal to the event loop and one coroutine
724	of lower priority, but only once, using idle watchers to keep the event	2188	of lower priority, but only once, using idle watchers to keep the event
725	loop from blocking if lower-priority coroutines are active, thus mapping	2189	loop from blocking if lower-priority coroutines are active, thus mapping
726	low-priority coroutines to idle/background tasks).	2190	low-priority coroutines to idle/background tasks).
727		2191
		2192	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2193	priority, to ensure that they are being run before any other watchers
		2194	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2195
		2196	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2197	activate ("feed") events into libev. While libev fully supports this, they
		2198	might get executed before other C<ev_check> watchers did their job. As
		2199	C<ev_check> watchers are often used to embed other (non-libev) event
		2200	loops those other event loops might be in an unusable state until their
		2201	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2202	others).
		2203
		2204	=head3 Watcher-Specific Functions and Data Members
		2205
728	=over 4	2206	=over 4
729		2207
730	=item ev_prepare_init (ev_prepare *, callback)	2208	=item ev_prepare_init (ev_prepare *, callback)
731		2209
732	=item ev_check_init (ev_check *, callback)	2210	=item ev_check_init (ev_check *, callback)
733		2211
734	Initialises and configures the prepare or check watcher - they have no	2212	Initialises and configures the prepare or check watcher - they have no
735	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2213	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
736	macros, but using them is utterly, utterly and completely pointless.	2214	macros, but using them is utterly, utterly, utterly and completely
		2215	pointless.
737		2216
738	=back	2217	=back
739		2218
		2219	=head3 Examples
		2220
		2221	There are a number of principal ways to embed other event loops or modules
		2222	into libev. Here are some ideas on how to include libadns into libev
		2223	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2224	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2225	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2226	Glib event loop).
		2227
		2228	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2229	and in a check watcher, destroy them and call into libadns. What follows
		2230	is pseudo-code only of course. This requires you to either use a low
		2231	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2232	the callbacks for the IO/timeout watchers might not have been called yet.
		2233
		2234	static ev_io iow [nfd];
		2235	static ev_timer tw;
		2236
		2237	static void
		2238	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2239	{
		2240	}
		2241
		2242	// create io watchers for each fd and a timer before blocking
		2243	static void
		2244	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2245	{
		2246	int timeout = 3600000;
		2247	struct pollfd fds [nfd];
		2248	// actual code will need to loop here and realloc etc.
		2249	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2250
		2251	/* the callback is illegal, but won't be called as we stop during check */
		2252	ev_timer_init (&tw, 0, timeout * 1e-3);
		2253	ev_timer_start (loop, &tw);
		2254
		2255	// create one ev_io per pollfd
		2256	for (int i = 0; i < nfd; ++i)
		2257	{
		2258	ev_io_init (iow + i, io_cb, fds [i].fd,
		2259	((fds [i].events & POLLIN ? EV_READ : 0)
		2260	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2261
		2262	fds [i].revents = 0;
		2263	ev_io_start (loop, iow + i);
		2264	}
		2265	}
		2266
		2267	// stop all watchers after blocking
		2268	static void
		2269	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2270	{
		2271	ev_timer_stop (loop, &tw);
		2272
		2273	for (int i = 0; i < nfd; ++i)
		2274	{
		2275	// set the relevant poll flags
		2276	// could also call adns_processreadable etc. here
		2277	struct pollfd *fd = fds + i;
		2278	int revents = ev_clear_pending (iow + i);
		2279	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2280	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2281
		2282	// now stop the watcher
		2283	ev_io_stop (loop, iow + i);
		2284	}
		2285
		2286	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2287	}
		2288
		2289	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2290	in the prepare watcher and would dispose of the check watcher.
		2291
		2292	Method 3: If the module to be embedded supports explicit event
		2293	notification (libadns does), you can also make use of the actual watcher
		2294	callbacks, and only destroy/create the watchers in the prepare watcher.
		2295
		2296	static void
		2297	timer_cb (EV_P_ ev_timer *w, int revents)
		2298	{
		2299	adns_state ads = (adns_state)w->data;
		2300	update_now (EV_A);
		2301
		2302	adns_processtimeouts (ads, &tv_now);
		2303	}
		2304
		2305	static void
		2306	io_cb (EV_P_ ev_io *w, int revents)
		2307	{
		2308	adns_state ads = (adns_state)w->data;
		2309	update_now (EV_A);
		2310
		2311	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2312	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2313	}
		2314
		2315	// do not ever call adns_afterpoll
		2316
		2317	Method 4: Do not use a prepare or check watcher because the module you
		2318	want to embed is not flexible enough to support it. Instead, you can
		2319	override their poll function. The drawback with this solution is that the
		2320	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2321	this approach, effectively embedding EV as a client into the horrible
		2322	libglib event loop.
		2323
		2324	static gint
		2325	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2326	{
		2327	int got_events = 0;
		2328
		2329	for (n = 0; n < nfds; ++n)
		2330	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2331
		2332	if (timeout >= 0)
		2333	// create/start timer
		2334
		2335	// poll
		2336	ev_loop (EV_A_ 0);
		2337
		2338	// stop timer again
		2339	if (timeout >= 0)
		2340	ev_timer_stop (EV_A_ &to);
		2341
		2342	// stop io watchers again - their callbacks should have set
		2343	for (n = 0; n < nfds; ++n)
		2344	ev_io_stop (EV_A_ iow [n]);
		2345
		2346	return got_events;
		2347	}
		2348
		2349
		2350	=head2 C<ev_embed> - when one backend isn't enough...
		2351
		2352	This is a rather advanced watcher type that lets you embed one event loop
		2353	into another (currently only C<ev_io> events are supported in the embedded
		2354	loop, other types of watchers might be handled in a delayed or incorrect
		2355	fashion and must not be used).
		2356
		2357	There are primarily two reasons you would want that: work around bugs and
		2358	prioritise I/O.
		2359
		2360	As an example for a bug workaround, the kqueue backend might only support
		2361	sockets on some platform, so it is unusable as generic backend, but you
		2362	still want to make use of it because you have many sockets and it scales
		2363	so nicely. In this case, you would create a kqueue-based loop and embed
		2364	it into your default loop (which might use e.g. poll). Overall operation
		2365	will be a bit slower because first libev has to call C<poll> and then
		2366	C<kevent>, but at least you can use both mechanisms for what they are
		2367	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2368
		2369	As for prioritising I/O: under rare circumstances you have the case where
		2370	some fds have to be watched and handled very quickly (with low latency),
		2371	and even priorities and idle watchers might have too much overhead. In
		2372	this case you would put all the high priority stuff in one loop and all
		2373	the rest in a second one, and embed the second one in the first.
		2374
		2375	As long as the watcher is active, the callback will be invoked every time
		2376	there might be events pending in the embedded loop. The callback must then
		2377	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
		2378	their callbacks (you could also start an idle watcher to give the embedded
		2379	loop strictly lower priority for example). You can also set the callback
		2380	to C<0>, in which case the embed watcher will automatically execute the
		2381	embedded loop sweep.
		2382
		2383	As long as the watcher is started it will automatically handle events. The
		2384	callback will be invoked whenever some events have been handled. You can
		2385	set the callback to C<0> to avoid having to specify one if you are not
		2386	interested in that.
		2387
		2388	Also, there have not currently been made special provisions for forking:
		2389	when you fork, you not only have to call C<ev_loop_fork> on both loops,
		2390	but you will also have to stop and restart any C<ev_embed> watchers
		2391	yourself - but you can use a fork watcher to handle this automatically,
		2392	and future versions of libev might do just that.
		2393
		2394	Unfortunately, not all backends are embeddable: only the ones returned by
		2395	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2396	portable one.
		2397
		2398	So when you want to use this feature you will always have to be prepared
		2399	that you cannot get an embeddable loop. The recommended way to get around
		2400	this is to have a separate variables for your embeddable loop, try to
		2401	create it, and if that fails, use the normal loop for everything.
		2402
		2403	=head3 C<ev_embed> and fork
		2404
		2405	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2406	automatically be applied to the embedded loop as well, so no special
		2407	fork handling is required in that case. When the watcher is not running,
		2408	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2409	as applicable.
		2410
		2411	=head3 Watcher-Specific Functions and Data Members
		2412
		2413	=over 4
		2414
		2415	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2416
		2417	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2418
		2419	Configures the watcher to embed the given loop, which must be
		2420	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2421	invoked automatically, otherwise it is the responsibility of the callback
		2422	to invoke it (it will continue to be called until the sweep has been done,
		2423	if you do not want that, you need to temporarily stop the embed watcher).
		2424
		2425	=item ev_embed_sweep (loop, ev_embed *)
		2426
		2427	Make a single, non-blocking sweep over the embedded loop. This works
		2428	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2429	appropriate way for embedded loops.
		2430
		2431	=item struct ev_loop *other [read-only]
		2432
		2433	The embedded event loop.
		2434
		2435	=back
		2436
		2437	=head3 Examples
		2438
		2439	Example: Try to get an embeddable event loop and embed it into the default
		2440	event loop. If that is not possible, use the default loop. The default
		2441	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2442	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2443	used).
		2444
		2445	struct ev_loop *loop_hi = ev_default_init (0);
		2446	struct ev_loop *loop_lo = 0;
		2447	ev_embed embed;
		2448
		2449	// see if there is a chance of getting one that works
		2450	// (remember that a flags value of 0 means autodetection)
		2451	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2452	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2453	: 0;
		2454
		2455	// if we got one, then embed it, otherwise default to loop_hi
		2456	if (loop_lo)
		2457	{
		2458	ev_embed_init (&embed, 0, loop_lo);
		2459	ev_embed_start (loop_hi, &embed);
		2460	}
		2461	else
		2462	loop_lo = loop_hi;
		2463
		2464	Example: Check if kqueue is available but not recommended and create
		2465	a kqueue backend for use with sockets (which usually work with any
		2466	kqueue implementation). Store the kqueue/socket-only event loop in
		2467	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2468
		2469	struct ev_loop *loop = ev_default_init (0);
		2470	struct ev_loop *loop_socket = 0;
		2471	ev_embed embed;
		2472
		2473	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2474	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2475	{
		2476	ev_embed_init (&embed, 0, loop_socket);
		2477	ev_embed_start (loop, &embed);
		2478	}
		2479
		2480	if (!loop_socket)
		2481	loop_socket = loop;
		2482
		2483	// now use loop_socket for all sockets, and loop for everything else
		2484
		2485
		2486	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2487
		2488	Fork watchers are called when a C<fork ()> was detected (usually because
		2489	whoever is a good citizen cared to tell libev about it by calling
		2490	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2491	event loop blocks next and before C<ev_check> watchers are being called,
		2492	and only in the child after the fork. If whoever good citizen calling
		2493	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2494	handlers will be invoked, too, of course.
		2495
		2496	=head3 Watcher-Specific Functions and Data Members
		2497
		2498	=over 4
		2499
		2500	=item ev_fork_init (ev_signal *, callback)
		2501
		2502	Initialises and configures the fork watcher - it has no parameters of any
		2503	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		2504	believe me.
		2505
		2506	=back
		2507
		2508
		2509	=head2 C<ev_async> - how to wake up another event loop
		2510
		2511	In general, you cannot use an C<ev_loop> from multiple threads or other
		2512	asynchronous sources such as signal handlers (as opposed to multiple event
		2513	loops - those are of course safe to use in different threads).
		2514
		2515	Sometimes, however, you need to wake up another event loop you do not
		2516	control, for example because it belongs to another thread. This is what
		2517	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2518	can signal it by calling C<ev_async_send>, which is thread- and signal
		2519	safe.
		2520
		2521	This functionality is very similar to C<ev_signal> watchers, as signals,
		2522	too, are asynchronous in nature, and signals, too, will be compressed
		2523	(i.e. the number of callback invocations may be less than the number of
		2524	C<ev_async_sent> calls).
		2525
		2526	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2527	just the default loop.
		2528
		2529	=head3 Queueing
		2530
		2531	C<ev_async> does not support queueing of data in any way. The reason
		2532	is that the author does not know of a simple (or any) algorithm for a
		2533	multiple-writer-single-reader queue that works in all cases and doesn't
		2534	need elaborate support such as pthreads.
		2535
		2536	That means that if you want to queue data, you have to provide your own
		2537	queue. But at least I can tell you how to implement locking around your
		2538	queue:
		2539
		2540	=over 4
		2541
		2542	=item queueing from a signal handler context
		2543
		2544	To implement race-free queueing, you simply add to the queue in the signal
		2545	handler but you block the signal handler in the watcher callback. Here is
		2546	an example that does that for some fictitious SIGUSR1 handler:
		2547
		2548	static ev_async mysig;
		2549
		2550	static void
		2551	sigusr1_handler (void)
		2552	{
		2553	sometype data;
		2554
		2555	// no locking etc.
		2556	queue_put (data);
		2557	ev_async_send (EV_DEFAULT_ &mysig);
		2558	}
		2559
		2560	static void
		2561	mysig_cb (EV_P_ ev_async *w, int revents)
		2562	{
		2563	sometype data;
		2564	sigset_t block, prev;
		2565
		2566	sigemptyset (&block);
		2567	sigaddset (&block, SIGUSR1);
		2568	sigprocmask (SIG_BLOCK, &block, &prev);
		2569
		2570	while (queue_get (&data))
		2571	process (data);
		2572
		2573	if (sigismember (&prev, SIGUSR1)
		2574	sigprocmask (SIG_UNBLOCK, &block, 0);
		2575	}
		2576
		2577	(Note: pthreads in theory requires you to use C<pthread_setmask>
		2578	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		2579	either...).
		2580
		2581	=item queueing from a thread context
		2582
		2583	The strategy for threads is different, as you cannot (easily) block
		2584	threads but you can easily preempt them, so to queue safely you need to
		2585	employ a traditional mutex lock, such as in this pthread example:
		2586
		2587	static ev_async mysig;
		2588	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		2589
		2590	static void
		2591	otherthread (void)
		2592	{
		2593	// only need to lock the actual queueing operation
		2594	pthread_mutex_lock (&mymutex);
		2595	queue_put (data);
		2596	pthread_mutex_unlock (&mymutex);
		2597
		2598	ev_async_send (EV_DEFAULT_ &mysig);
		2599	}
		2600
		2601	static void
		2602	mysig_cb (EV_P_ ev_async *w, int revents)
		2603	{
		2604	pthread_mutex_lock (&mymutex);
		2605
		2606	while (queue_get (&data))
		2607	process (data);
		2608
		2609	pthread_mutex_unlock (&mymutex);
		2610	}
		2611
		2612	=back
		2613
		2614
		2615	=head3 Watcher-Specific Functions and Data Members
		2616
		2617	=over 4
		2618
		2619	=item ev_async_init (ev_async *, callback)
		2620
		2621	Initialises and configures the async watcher - it has no parameters of any
		2622	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
		2623	trust me.
		2624
		2625	=item ev_async_send (loop, ev_async *)
		2626
		2627	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		2628	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		2629	C<ev_feed_event>, this call is safe to do from other threads, signal or
		2630	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		2631	section below on what exactly this means).
		2632
		2633	This call incurs the overhead of a system call only once per loop iteration,
		2634	so while the overhead might be noticeable, it doesn't apply to repeated
		2635	calls to C<ev_async_send>.
		2636
		2637	=item bool = ev_async_pending (ev_async *)
		2638
		2639	Returns a non-zero value when C<ev_async_send> has been called on the
		2640	watcher but the event has not yet been processed (or even noted) by the
		2641	event loop.
		2642
		2643	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		2644	the loop iterates next and checks for the watcher to have become active,
		2645	it will reset the flag again. C<ev_async_pending> can be used to very
		2646	quickly check whether invoking the loop might be a good idea.
		2647
		2648	Not that this does I<not> check whether the watcher itself is pending, only
		2649	whether it has been requested to make this watcher pending.
		2650
		2651	=back
		2652
		2653
740	=head1 OTHER FUNCTIONS	2654	=head1 OTHER FUNCTIONS
741		2655
742	There are some other functions of possible interest. Described. Here. Now.	2656	There are some other functions of possible interest. Described. Here. Now.
743		2657
744	=over 4	2658	=over 4
745		2659
746	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	2660	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
747		2661
748	This function combines a simple timer and an I/O watcher, calls your	2662	This function combines a simple timer and an I/O watcher, calls your
749	callback on whichever event happens first and automatically stop both	2663	callback on whichever event happens first and automatically stops both
750	watchers. This is useful if you want to wait for a single event on an fd	2664	watchers. This is useful if you want to wait for a single event on an fd
751	or timeout without having to allocate/configure/start/stop/free one or	2665	or timeout without having to allocate/configure/start/stop/free one or
752	more watchers yourself.	2666	more watchers yourself.
753		2667
754	If C<fd> is less than 0, then no I/O watcher will be started and events	2668	If C<fd> is less than 0, then no I/O watcher will be started and the
755	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	2669	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
756	C<events> set will be craeted and started.	2670	the given C<fd> and C<events> set will be created and started.
757		2671
758	If C<timeout> is less than 0, then no timeout watcher will be	2672	If C<timeout> is less than 0, then no timeout watcher will be
759	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	2673	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
760	repeat = 0) will be started. While C<0> is a valid timeout, it is of	2674	repeat = 0) will be started. C<0> is a valid timeout.
761	dubious value.
762		2675
763	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	2676	The callback has the type C<void (*cb)(int revents, void *arg)> and gets
764	passed an C<revents> set like normal event callbacks (a combination of	2677	passed an C<revents> set like normal event callbacks (a combination of
765	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	2678	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
766	value passed to C<ev_once>:	2679	value passed to C<ev_once>. Note that it is possible to receive I<both>
		2680	a timeout and an io event at the same time - you probably should give io
		2681	events precedence.
767		2682
		2683	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		2684
768	static void stdin_ready (int revents, void *arg)	2685	static void stdin_ready (int revents, void *arg)
769	{	2686	{
770	if (revents & EV_TIMEOUT)
771	/* doh, nothing entered */;
772	else if (revents & EV_READ)	2687	if (revents & EV_READ)
773	/* stdin might have data for us, joy! */;	2688	/* stdin might have data for us, joy! */;
		2689	else if (revents & EV_TIMEOUT)
		2690	/* doh, nothing entered */;
774	}	2691	}
775		2692
776	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	2693	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
777		2694
778	=item ev_feed_event (loop, watcher, int events)	2695	=item ev_feed_event (struct ev_loop *, watcher *, int revents)
779		2696
780	Feeds the given event set into the event loop, as if the specified event	2697	Feeds the given event set into the event loop, as if the specified event
781	had happened for the specified watcher (which must be a pointer to an	2698	had happened for the specified watcher (which must be a pointer to an
782	initialised but not necessarily started event watcher).	2699	initialised but not necessarily started event watcher).
783		2700
784	=item ev_feed_fd_event (loop, int fd, int revents)	2701	=item ev_feed_fd_event (struct ev_loop *, int fd, int revents)
785		2702
786	Feed an event on the given fd, as if a file descriptor backend detected	2703	Feed an event on the given fd, as if a file descriptor backend detected
787	the given events it.	2704	the given events it.
788		2705
789	=item ev_feed_signal_event (loop, int signum)	2706	=item ev_feed_signal_event (struct ev_loop *loop, int signum)
790		2707
791	Feed an event as if the given signal occured (loop must be the default loop!).	2708	Feed an event as if the given signal occurred (C<loop> must be the default
		2709	loop!).
792		2710
793	=back	2711	=back
		2712
794		2713
795	=head1 LIBEVENT EMULATION	2714	=head1 LIBEVENT EMULATION
796		2715
797	Libev offers a compatibility emulation layer for libevent. It cannot	2716	Libev offers a compatibility emulation layer for libevent. It cannot
798	emulate the internals of libevent, so here are some usage hints:	2717	emulate the internals of libevent, so here are some usage hints:
…		…
810		2729
811	=item * Priorities are not currently supported. Initialising priorities	2730	=item * Priorities are not currently supported. Initialising priorities
812	will fail and all watchers will have the same priority, even though there	2731	will fail and all watchers will have the same priority, even though there
813	is an ev_pri field.	2732	is an ev_pri field.
814		2733
		2734	=item * In libevent, the last base created gets the signals, in libev, the
		2735	first base created (== the default loop) gets the signals.
		2736
815	=item * Other members are not supported.	2737	=item * Other members are not supported.
816		2738
817	=item * The libev emulation is I<not> ABI compatible to libevent, you need	2739	=item * The libev emulation is I<not> ABI compatible to libevent, you need
818	to use the libev header file and library.	2740	to use the libev header file and library.
819		2741
820	=back	2742	=back
821		2743
822	=head1 C++ SUPPORT	2744	=head1 C++ SUPPORT
823		2745
824	TBD.	2746	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		2747	you to use some convenience methods to start/stop watchers and also change
		2748	the callback model to a model using method callbacks on objects.
		2749
		2750	To use it,
		2751
		2752	#include <ev++.h>
		2753
		2754	This automatically includes F<ev.h> and puts all of its definitions (many
		2755	of them macros) into the global namespace. All C++ specific things are
		2756	put into the C<ev> namespace. It should support all the same embedding
		2757	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		2758
		2759	Care has been taken to keep the overhead low. The only data member the C++
		2760	classes add (compared to plain C-style watchers) is the event loop pointer
		2761	that the watcher is associated with (or no additional members at all if
		2762	you disable C<EV_MULTIPLICITY> when embedding libev).
		2763
		2764	Currently, functions, and static and non-static member functions can be
		2765	used as callbacks. Other types should be easy to add as long as they only
		2766	need one additional pointer for context. If you need support for other
		2767	types of functors please contact the author (preferably after implementing
		2768	it).
		2769
		2770	Here is a list of things available in the C<ev> namespace:
		2771
		2772	=over 4
		2773
		2774	=item C<ev::READ>, C<ev::WRITE> etc.
		2775
		2776	These are just enum values with the same values as the C<EV_READ> etc.
		2777	macros from F<ev.h>.
		2778
		2779	=item C<ev::tstamp>, C<ev::now>
		2780
		2781	Aliases to the same types/functions as with the C<ev_> prefix.
		2782
		2783	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		2784
		2785	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		2786	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		2787	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		2788	defines by many implementations.
		2789
		2790	All of those classes have these methods:
		2791
		2792	=over 4
		2793
		2794	=item ev::TYPE::TYPE ()
		2795
		2796	=item ev::TYPE::TYPE (struct ev_loop *)
		2797
		2798	=item ev::TYPE::~TYPE
		2799
		2800	The constructor (optionally) takes an event loop to associate the watcher
		2801	with. If it is omitted, it will use C<EV_DEFAULT>.
		2802
		2803	The constructor calls C<ev_init> for you, which means you have to call the
		2804	C<set> method before starting it.
		2805
		2806	It will not set a callback, however: You have to call the templated C<set>
		2807	method to set a callback before you can start the watcher.
		2808
		2809	(The reason why you have to use a method is a limitation in C++ which does
		2810	not allow explicit template arguments for constructors).
		2811
		2812	The destructor automatically stops the watcher if it is active.
		2813
		2814	=item w->set<class, &class::method> (object *)
		2815
		2816	This method sets the callback method to call. The method has to have a
		2817	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		2818	first argument and the C<revents> as second. The object must be given as
		2819	parameter and is stored in the C<data> member of the watcher.
		2820
		2821	This method synthesizes efficient thunking code to call your method from
		2822	the C callback that libev requires. If your compiler can inline your
		2823	callback (i.e. it is visible to it at the place of the C<set> call and
		2824	your compiler is good :), then the method will be fully inlined into the
		2825	thunking function, making it as fast as a direct C callback.
		2826
		2827	Example: simple class declaration and watcher initialisation
		2828
		2829	struct myclass
		2830	{
		2831	void io_cb (ev::io &w, int revents) { }
		2832	}
		2833
		2834	myclass obj;
		2835	ev::io iow;
		2836	iow.set <myclass, &myclass::io_cb> (&obj);
		2837
		2838	=item w->set<function> (void *data = 0)
		2839
		2840	Also sets a callback, but uses a static method or plain function as
		2841	callback. The optional C<data> argument will be stored in the watcher's
		2842	C<data> member and is free for you to use.
		2843
		2844	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		2845
		2846	See the method-C<set> above for more details.
		2847
		2848	Example: Use a plain function as callback.
		2849
		2850	static void io_cb (ev::io &w, int revents) { }
		2851	iow.set <io_cb> ();
		2852
		2853	=item w->set (struct ev_loop *)
		2854
		2855	Associates a different C<struct ev_loop> with this watcher. You can only
		2856	do this when the watcher is inactive (and not pending either).
		2857
		2858	=item w->set ([arguments])
		2859
		2860	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		2861	called at least once. Unlike the C counterpart, an active watcher gets
		2862	automatically stopped and restarted when reconfiguring it with this
		2863	method.
		2864
		2865	=item w->start ()
		2866
		2867	Starts the watcher. Note that there is no C<loop> argument, as the
		2868	constructor already stores the event loop.
		2869
		2870	=item w->stop ()
		2871
		2872	Stops the watcher if it is active. Again, no C<loop> argument.
		2873
		2874	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		2875
		2876	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		2877	C<ev_TYPE_again> function.
		2878
		2879	=item w->sweep () (C<ev::embed> only)
		2880
		2881	Invokes C<ev_embed_sweep>.
		2882
		2883	=item w->update () (C<ev::stat> only)
		2884
		2885	Invokes C<ev_stat_stat>.
		2886
		2887	=back
		2888
		2889	=back
		2890
		2891	Example: Define a class with an IO and idle watcher, start one of them in
		2892	the constructor.
		2893
		2894	class myclass
		2895	{
		2896	ev::io io ; void io_cb (ev::io &w, int revents);
		2897	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		2898
		2899	myclass (int fd)
		2900	{
		2901	io .set <myclass, &myclass::io_cb > (this);
		2902	idle.set <myclass, &myclass::idle_cb> (this);
		2903
		2904	io.start (fd, ev::READ);
		2905	}
		2906	};
		2907
		2908
		2909	=head1 OTHER LANGUAGE BINDINGS
		2910
		2911	Libev does not offer other language bindings itself, but bindings for a
		2912	number of languages exist in the form of third-party packages. If you know
		2913	any interesting language binding in addition to the ones listed here, drop
		2914	me a note.
		2915
		2916	=over 4
		2917
		2918	=item Perl
		2919
		2920	The EV module implements the full libev API and is actually used to test
		2921	libev. EV is developed together with libev. Apart from the EV core module,
		2922	there are additional modules that implement libev-compatible interfaces
		2923	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		2924	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		2925	and C<EV::Glib>).
		2926
		2927	It can be found and installed via CPAN, its homepage is at
		2928	L<http://software.schmorp.de/pkg/EV>.
		2929
		2930	=item Python
		2931
		2932	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		2933	seems to be quite complete and well-documented. Note, however, that the
		2934	patch they require for libev is outright dangerous as it breaks the ABI
		2935	for everybody else, and therefore, should never be applied in an installed
		2936	libev (if python requires an incompatible ABI then it needs to embed
		2937	libev).
		2938
		2939	=item Ruby
		2940
		2941	Tony Arcieri has written a ruby extension that offers access to a subset
		2942	of the libev API and adds file handle abstractions, asynchronous DNS and
		2943	more on top of it. It can be found via gem servers. Its homepage is at
		2944	L<http://rev.rubyforge.org/>.
		2945
		2946	=item D
		2947
		2948	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		2949	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		2950
		2951	=back
		2952
		2953
		2954	=head1 MACRO MAGIC
		2955
		2956	Libev can be compiled with a variety of options, the most fundamental
		2957	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		2958	functions and callbacks have an initial C<struct ev_loop *> argument.
		2959
		2960	To make it easier to write programs that cope with either variant, the
		2961	following macros are defined:
		2962
		2963	=over 4
		2964
		2965	=item C<EV_A>, C<EV_A_>
		2966
		2967	This provides the loop I<argument> for functions, if one is required ("ev
		2968	loop argument"). The C<EV_A> form is used when this is the sole argument,
		2969	C<EV_A_> is used when other arguments are following. Example:
		2970
		2971	ev_unref (EV_A);
		2972	ev_timer_add (EV_A_ watcher);
		2973	ev_loop (EV_A_ 0);
		2974
		2975	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		2976	which is often provided by the following macro.
		2977
		2978	=item C<EV_P>, C<EV_P_>
		2979
		2980	This provides the loop I<parameter> for functions, if one is required ("ev
		2981	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		2982	C<EV_P_> is used when other parameters are following. Example:
		2983
		2984	// this is how ev_unref is being declared
		2985	static void ev_unref (EV_P);
		2986
		2987	// this is how you can declare your typical callback
		2988	static void cb (EV_P_ ev_timer *w, int revents)
		2989
		2990	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		2991	suitable for use with C<EV_A>.
		2992
		2993	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		2994
		2995	Similar to the other two macros, this gives you the value of the default
		2996	loop, if multiple loops are supported ("ev loop default").
		2997
		2998	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		2999
		3000	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3001	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3002	is undefined when the default loop has not been initialised by a previous
		3003	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3004
		3005	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3006	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3007
		3008	=back
		3009
		3010	Example: Declare and initialise a check watcher, utilising the above
		3011	macros so it will work regardless of whether multiple loops are supported
		3012	or not.
		3013
		3014	static void
		3015	check_cb (EV_P_ ev_timer *w, int revents)
		3016	{
		3017	ev_check_stop (EV_A_ w);
		3018	}
		3019
		3020	ev_check check;
		3021	ev_check_init (&check, check_cb);
		3022	ev_check_start (EV_DEFAULT_ &check);
		3023	ev_loop (EV_DEFAULT_ 0);
		3024
		3025	=head1 EMBEDDING
		3026
		3027	Libev can (and often is) directly embedded into host
		3028	applications. Examples of applications that embed it include the Deliantra
		3029	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3030	and rxvt-unicode.
		3031
		3032	The goal is to enable you to just copy the necessary files into your
		3033	source directory without having to change even a single line in them, so
		3034	you can easily upgrade by simply copying (or having a checked-out copy of
		3035	libev somewhere in your source tree).
		3036
		3037	=head2 FILESETS
		3038
		3039	Depending on what features you need you need to include one or more sets of files
		3040	in your application.
		3041
		3042	=head3 CORE EVENT LOOP
		3043
		3044	To include only the libev core (all the C<ev_*> functions), with manual
		3045	configuration (no autoconf):
		3046
		3047	#define EV_STANDALONE 1
		3048	#include "ev.c"
		3049
		3050	This will automatically include F<ev.h>, too, and should be done in a
		3051	single C source file only to provide the function implementations. To use
		3052	it, do the same for F<ev.h> in all files wishing to use this API (best
		3053	done by writing a wrapper around F<ev.h> that you can include instead and
		3054	where you can put other configuration options):
		3055
		3056	#define EV_STANDALONE 1
		3057	#include "ev.h"
		3058
		3059	Both header files and implementation files can be compiled with a C++
		3060	compiler (at least, thats a stated goal, and breakage will be treated
		3061	as a bug).
		3062
		3063	You need the following files in your source tree, or in a directory
		3064	in your include path (e.g. in libev/ when using -Ilibev):
		3065
		3066	ev.h
		3067	ev.c
		3068	ev_vars.h
		3069	ev_wrap.h
		3070
		3071	ev_win32.c required on win32 platforms only
		3072
		3073	ev_select.c only when select backend is enabled (which is enabled by default)
		3074	ev_poll.c only when poll backend is enabled (disabled by default)
		3075	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3076	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3077	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3078
		3079	F<ev.c> includes the backend files directly when enabled, so you only need
		3080	to compile this single file.
		3081
		3082	=head3 LIBEVENT COMPATIBILITY API
		3083
		3084	To include the libevent compatibility API, also include:
		3085
		3086	#include "event.c"
		3087
		3088	in the file including F<ev.c>, and:
		3089
		3090	#include "event.h"
		3091
		3092	in the files that want to use the libevent API. This also includes F<ev.h>.
		3093
		3094	You need the following additional files for this:
		3095
		3096	event.h
		3097	event.c
		3098
		3099	=head3 AUTOCONF SUPPORT
		3100
		3101	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3102	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3103	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3104	include F<config.h> and configure itself accordingly.
		3105
		3106	For this of course you need the m4 file:
		3107
		3108	libev.m4
		3109
		3110	=head2 PREPROCESSOR SYMBOLS/MACROS
		3111
		3112	Libev can be configured via a variety of preprocessor symbols you have to
		3113	define before including any of its files. The default in the absence of
		3114	autoconf is documented for every option.
		3115
		3116	=over 4
		3117
		3118	=item EV_STANDALONE
		3119
		3120	Must always be C<1> if you do not use autoconf configuration, which
		3121	keeps libev from including F<config.h>, and it also defines dummy
		3122	implementations for some libevent functions (such as logging, which is not
		3123	supported). It will also not define any of the structs usually found in
		3124	F<event.h> that are not directly supported by the libev core alone.
		3125
		3126	=item EV_USE_MONOTONIC
		3127
		3128	If defined to be C<1>, libev will try to detect the availability of the
		3129	monotonic clock option at both compile time and runtime. Otherwise no use
		3130	of the monotonic clock option will be attempted. If you enable this, you
		3131	usually have to link against librt or something similar. Enabling it when
		3132	the functionality isn't available is safe, though, although you have
		3133	to make sure you link against any libraries where the C<clock_gettime>
		3134	function is hiding in (often F<-lrt>).
		3135
		3136	=item EV_USE_REALTIME
		3137
		3138	If defined to be C<1>, libev will try to detect the availability of the
		3139	real-time clock option at compile time (and assume its availability at
		3140	runtime if successful). Otherwise no use of the real-time clock option will
		3141	be attempted. This effectively replaces C<gettimeofday> by C<clock_get
		3142	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
		3143	note about libraries in the description of C<EV_USE_MONOTONIC>, though.
		3144
		3145	=item EV_USE_NANOSLEEP
		3146
		3147	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3148	and will use it for delays. Otherwise it will use C<select ()>.
		3149
		3150	=item EV_USE_EVENTFD
		3151
		3152	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3153	available and will probe for kernel support at runtime. This will improve
		3154	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3155	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3156	2.7 or newer, otherwise disabled.
		3157
		3158	=item EV_USE_SELECT
		3159
		3160	If undefined or defined to be C<1>, libev will compile in support for the
		3161	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3162	other method takes over, select will be it. Otherwise the select backend
		3163	will not be compiled in.
		3164
		3165	=item EV_SELECT_USE_FD_SET
		3166
		3167	If defined to C<1>, then the select backend will use the system C<fd_set>
		3168	structure. This is useful if libev doesn't compile due to a missing
		3169	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout on
		3170	exotic systems. This usually limits the range of file descriptors to some
		3171	low limit such as 1024 or might have other limitations (winsocket only
		3172	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
		3173	influence the size of the C<fd_set> used.
		3174
		3175	=item EV_SELECT_IS_WINSOCKET
		3176
		3177	When defined to C<1>, the select backend will assume that
		3178	select/socket/connect etc. don't understand file descriptors but
		3179	wants osf handles on win32 (this is the case when the select to
		3180	be used is the winsock select). This means that it will call
		3181	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3182	it is assumed that all these functions actually work on fds, even
		3183	on win32. Should not be defined on non-win32 platforms.
		3184
		3185	=item EV_FD_TO_WIN32_HANDLE
		3186
		3187	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3188	file descriptors to socket handles. When not defining this symbol (the
		3189	default), then libev will call C<_get_osfhandle>, which is usually
		3190	correct. In some cases, programs use their own file descriptor management,
		3191	in which case they can provide this function to map fds to socket handles.
		3192
		3193	=item EV_USE_POLL
		3194
		3195	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3196	backend. Otherwise it will be enabled on non-win32 platforms. It
		3197	takes precedence over select.
		3198
		3199	=item EV_USE_EPOLL
		3200
		3201	If defined to be C<1>, libev will compile in support for the Linux
		3202	C<epoll>(7) backend. Its availability will be detected at runtime,
		3203	otherwise another method will be used as fallback. This is the preferred
		3204	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3205	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3206
		3207	=item EV_USE_KQUEUE
		3208
		3209	If defined to be C<1>, libev will compile in support for the BSD style
		3210	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3211	otherwise another method will be used as fallback. This is the preferred
		3212	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3213	supports some types of fds correctly (the only platform we found that
		3214	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3215	not be used unless explicitly requested. The best way to use it is to find
		3216	out whether kqueue supports your type of fd properly and use an embedded
		3217	kqueue loop.
		3218
		3219	=item EV_USE_PORT
		3220
		3221	If defined to be C<1>, libev will compile in support for the Solaris
		3222	10 port style backend. Its availability will be detected at runtime,
		3223	otherwise another method will be used as fallback. This is the preferred
		3224	backend for Solaris 10 systems.
		3225
		3226	=item EV_USE_DEVPOLL
		3227
		3228	Reserved for future expansion, works like the USE symbols above.
		3229
		3230	=item EV_USE_INOTIFY
		3231
		3232	If defined to be C<1>, libev will compile in support for the Linux inotify
		3233	interface to speed up C<ev_stat> watchers. Its actual availability will
		3234	be detected at runtime. If undefined, it will be enabled if the headers
		3235	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3236
		3237	=item EV_ATOMIC_T
		3238
		3239	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3240	access is atomic with respect to other threads or signal contexts. No such
		3241	type is easily found in the C language, so you can provide your own type
		3242	that you know is safe for your purposes. It is used both for signal handler "locking"
		3243	as well as for signal and thread safety in C<ev_async> watchers.
		3244
		3245	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3246	(from F<signal.h>), which is usually good enough on most platforms.
		3247
		3248	=item EV_H
		3249
		3250	The name of the F<ev.h> header file used to include it. The default if
		3251	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3252	used to virtually rename the F<ev.h> header file in case of conflicts.
		3253
		3254	=item EV_CONFIG_H
		3255
		3256	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3257	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3258	C<EV_H>, above.
		3259
		3260	=item EV_EVENT_H
		3261
		3262	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3263	of how the F<event.h> header can be found, the default is C<"event.h">.
		3264
		3265	=item EV_PROTOTYPES
		3266
		3267	If defined to be C<0>, then F<ev.h> will not define any function
		3268	prototypes, but still define all the structs and other symbols. This is
		3269	occasionally useful if you want to provide your own wrapper functions
		3270	around libev functions.
		3271
		3272	=item EV_MULTIPLICITY
		3273
		3274	If undefined or defined to C<1>, then all event-loop-specific functions
		3275	will have the C<struct ev_loop *> as first argument, and you can create
		3276	additional independent event loops. Otherwise there will be no support
		3277	for multiple event loops and there is no first event loop pointer
		3278	argument. Instead, all functions act on the single default loop.
		3279
		3280	=item EV_MINPRI
		3281
		3282	=item EV_MAXPRI
		3283
		3284	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3285	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3286	provide for more priorities by overriding those symbols (usually defined
		3287	to be C<-2> and C<2>, respectively).
		3288
		3289	When doing priority-based operations, libev usually has to linearly search
		3290	all the priorities, so having many of them (hundreds) uses a lot of space
		3291	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3292	fine.
		3293
		3294	If your embedding application does not need any priorities, defining these
		3295	both to C<0> will save some memory and CPU.
		3296
		3297	=item EV_PERIODIC_ENABLE
		3298
		3299	If undefined or defined to be C<1>, then periodic timers are supported. If
		3300	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3301	code.
		3302
		3303	=item EV_IDLE_ENABLE
		3304
		3305	If undefined or defined to be C<1>, then idle watchers are supported. If
		3306	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3307	code.
		3308
		3309	=item EV_EMBED_ENABLE
		3310
		3311	If undefined or defined to be C<1>, then embed watchers are supported. If
		3312	defined to be C<0>, then they are not. Embed watchers rely on most other
		3313	watcher types, which therefore must not be disabled.
		3314
		3315	=item EV_STAT_ENABLE
		3316
		3317	If undefined or defined to be C<1>, then stat watchers are supported. If
		3318	defined to be C<0>, then they are not.
		3319
		3320	=item EV_FORK_ENABLE
		3321
		3322	If undefined or defined to be C<1>, then fork watchers are supported. If
		3323	defined to be C<0>, then they are not.
		3324
		3325	=item EV_ASYNC_ENABLE
		3326
		3327	If undefined or defined to be C<1>, then async watchers are supported. If
		3328	defined to be C<0>, then they are not.
		3329
		3330	=item EV_MINIMAL
		3331
		3332	If you need to shave off some kilobytes of code at the expense of some
		3333	speed, define this symbol to C<1>. Currently this is used to override some
		3334	inlining decisions, saves roughly 30% code size on amd64. It also selects a
		3335	much smaller 2-heap for timer management over the default 4-heap.
		3336
		3337	=item EV_PID_HASHSIZE
		3338
		3339	C<ev_child> watchers use a small hash table to distribute workload by
		3340	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
		3341	than enough. If you need to manage thousands of children you might want to
		3342	increase this value (I<must> be a power of two).
		3343
		3344	=item EV_INOTIFY_HASHSIZE
		3345
		3346	C<ev_stat> watchers use a small hash table to distribute workload by
		3347	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
		3348	usually more than enough. If you need to manage thousands of C<ev_stat>
		3349	watchers you might want to increase this value (I<must> be a power of
		3350	two).
		3351
		3352	=item EV_USE_4HEAP
		3353
		3354	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3355	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		3356	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		3357	faster performance with many (thousands) of watchers.
		3358
		3359	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3360	(disabled).
		3361
		3362	=item EV_HEAP_CACHE_AT
		3363
		3364	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3365	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		3366	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3367	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3368	but avoids random read accesses on heap changes. This improves performance
		3369	noticeably with many (hundreds) of watchers.
		3370
		3371	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3372	(disabled).
		3373
		3374	=item EV_VERIFY
		3375
		3376	Controls how much internal verification (see C<ev_loop_verify ()>) will
		3377	be done: If set to C<0>, no internal verification code will be compiled
		3378	in. If set to C<1>, then verification code will be compiled in, but not
		3379	called. If set to C<2>, then the internal verification code will be
		3380	called once per loop, which can slow down libev. If set to C<3>, then the
		3381	verification code will be called very frequently, which will slow down
		3382	libev considerably.
		3383
		3384	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
		3385	C<0>.
		3386
		3387	=item EV_COMMON
		3388
		3389	By default, all watchers have a C<void *data> member. By redefining
		3390	this macro to a something else you can include more and other types of
		3391	members. You have to define it each time you include one of the files,
		3392	though, and it must be identical each time.
		3393
		3394	For example, the perl EV module uses something like this:
		3395
		3396	#define EV_COMMON \
		3397	SV *self; /* contains this struct */ \
		3398	SV *cb_sv, *fh /* note no trailing ";" */
		3399
		3400	=item EV_CB_DECLARE (type)
		3401
		3402	=item EV_CB_INVOKE (watcher, revents)
		3403
		3404	=item ev_set_cb (ev, cb)
		3405
		3406	Can be used to change the callback member declaration in each watcher,
		3407	and the way callbacks are invoked and set. Must expand to a struct member
		3408	definition and a statement, respectively. See the F<ev.h> header file for
		3409	their default definitions. One possible use for overriding these is to
		3410	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		3411	method calls instead of plain function calls in C++.
		3412
		3413	=back
		3414
		3415	=head2 EXPORTED API SYMBOLS
		3416
		3417	If you need to re-export the API (e.g. via a DLL) and you need a list of
		3418	exported symbols, you can use the provided F<Symbol.*> files which list
		3419	all public symbols, one per line:
		3420
		3421	Symbols.ev for libev proper
		3422	Symbols.event for the libevent emulation
		3423
		3424	This can also be used to rename all public symbols to avoid clashes with
		3425	multiple versions of libev linked together (which is obviously bad in
		3426	itself, but sometimes it is inconvenient to avoid this).
		3427
		3428	A sed command like this will create wrapper C<#define>'s that you need to
		3429	include before including F<ev.h>:
		3430
		3431	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		3432
		3433	This would create a file F<wrap.h> which essentially looks like this:
		3434
		3435	#define ev_backend myprefix_ev_backend
		3436	#define ev_check_start myprefix_ev_check_start
		3437	#define ev_check_stop myprefix_ev_check_stop
		3438	...
		3439
		3440	=head2 EXAMPLES
		3441
		3442	For a real-world example of a program the includes libev
		3443	verbatim, you can have a look at the EV perl module
		3444	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		3445	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		3446	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		3447	will be compiled. It is pretty complex because it provides its own header
		3448	file.
		3449
		3450	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		3451	that everybody includes and which overrides some configure choices:
		3452
		3453	#define EV_MINIMAL 1
		3454	#define EV_USE_POLL 0
		3455	#define EV_MULTIPLICITY 0
		3456	#define EV_PERIODIC_ENABLE 0
		3457	#define EV_STAT_ENABLE 0
		3458	#define EV_FORK_ENABLE 0
		3459	#define EV_CONFIG_H <config.h>
		3460	#define EV_MINPRI 0
		3461	#define EV_MAXPRI 0
		3462
		3463	#include "ev++.h"
		3464
		3465	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		3466
		3467	#include "ev_cpp.h"
		3468	#include "ev.c"
		3469
		3470	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		3471
		3472	=head2 THREADS AND COROUTINES
		3473
		3474	=head3 THREADS
		3475
		3476	All libev functions are reentrant and thread-safe unless explicitly
		3477	documented otherwise, but libev implements no locking itself. This means
		3478	that you can use as many loops as you want in parallel, as long as there
		3479	are no concurrent calls into any libev function with the same loop
		3480	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		3481	of course): libev guarantees that different event loops share no data
		3482	structures that need any locking.
		3483
		3484	Or to put it differently: calls with different loop parameters can be done
		3485	concurrently from multiple threads, calls with the same loop parameter
		3486	must be done serially (but can be done from different threads, as long as
		3487	only one thread ever is inside a call at any point in time, e.g. by using
		3488	a mutex per loop).
		3489
		3490	Specifically to support threads (and signal handlers), libev implements
		3491	so-called C<ev_async> watchers, which allow some limited form of
		3492	concurrency on the same event loop, namely waking it up "from the
		3493	outside".
		3494
		3495	If you want to know which design (one loop, locking, or multiple loops
		3496	without or something else still) is best for your problem, then I cannot
		3497	help you, but here is some generic advice:
		3498
		3499	=over 4
		3500
		3501	=item * most applications have a main thread: use the default libev loop
		3502	in that thread, or create a separate thread running only the default loop.
		3503
		3504	This helps integrating other libraries or software modules that use libev
		3505	themselves and don't care/know about threading.
		3506
		3507	=item * one loop per thread is usually a good model.
		3508
		3509	Doing this is almost never wrong, sometimes a better-performance model
		3510	exists, but it is always a good start.
		3511
		3512	=item * other models exist, such as the leader/follower pattern, where one
		3513	loop is handed through multiple threads in a kind of round-robin fashion.
		3514
		3515	Choosing a model is hard - look around, learn, know that usually you can do
		3516	better than you currently do :-)
		3517
		3518	=item * often you need to talk to some other thread which blocks in the
		3519	event loop.
		3520
		3521	C<ev_async> watchers can be used to wake them up from other threads safely
		3522	(or from signal contexts...).
		3523
		3524	An example use would be to communicate signals or other events that only
		3525	work in the default loop by registering the signal watcher with the
		3526	default loop and triggering an C<ev_async> watcher from the default loop
		3527	watcher callback into the event loop interested in the signal.
		3528
		3529	=back
		3530
		3531	=head3 COROUTINES
		3532
		3533	Libev is very accommodating to coroutines ("cooperative threads"):
		3534	libev fully supports nesting calls to its functions from different
		3535	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		3536	different coroutines, and switch freely between both coroutines running the
		3537	loop, as long as you don't confuse yourself). The only exception is that
		3538	you must not do this from C<ev_periodic> reschedule callbacks.
		3539
		3540	Care has been taken to ensure that libev does not keep local state inside
		3541	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		3542	they do not clal any callbacks.
		3543
		3544	=head2 COMPILER WARNINGS
		3545
		3546	Depending on your compiler and compiler settings, you might get no or a
		3547	lot of warnings when compiling libev code. Some people are apparently
		3548	scared by this.
		3549
		3550	However, these are unavoidable for many reasons. For one, each compiler
		3551	has different warnings, and each user has different tastes regarding
		3552	warning options. "Warn-free" code therefore cannot be a goal except when
		3553	targeting a specific compiler and compiler-version.
		3554
		3555	Another reason is that some compiler warnings require elaborate
		3556	workarounds, or other changes to the code that make it less clear and less
		3557	maintainable.
		3558
		3559	And of course, some compiler warnings are just plain stupid, or simply
		3560	wrong (because they don't actually warn about the condition their message
		3561	seems to warn about). For example, certain older gcc versions had some
		3562	warnings that resulted an extreme number of false positives. These have
		3563	been fixed, but some people still insist on making code warn-free with
		3564	such buggy versions.
		3565
		3566	While libev is written to generate as few warnings as possible,
		3567	"warn-free" code is not a goal, and it is recommended not to build libev
		3568	with any compiler warnings enabled unless you are prepared to cope with
		3569	them (e.g. by ignoring them). Remember that warnings are just that:
		3570	warnings, not errors, or proof of bugs.
		3571
		3572
		3573	=head2 VALGRIND
		3574
		3575	Valgrind has a special section here because it is a popular tool that is
		3576	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		3577
		3578	If you think you found a bug (memory leak, uninitialised data access etc.)
		3579	in libev, then check twice: If valgrind reports something like:
		3580
		3581	==2274== definitely lost: 0 bytes in 0 blocks.
		3582	==2274== possibly lost: 0 bytes in 0 blocks.
		3583	==2274== still reachable: 256 bytes in 1 blocks.
		3584
		3585	Then there is no memory leak, just as memory accounted to global variables
		3586	is not a memleak - the memory is still being refernced, and didn't leak.
		3587
		3588	Similarly, under some circumstances, valgrind might report kernel bugs
		3589	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		3590	although an acceptable workaround has been found here), or it might be
		3591	confused.
		3592
		3593	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		3594	make it into some kind of religion.
		3595
		3596	If you are unsure about something, feel free to contact the mailing list
		3597	with the full valgrind report and an explanation on why you think this
		3598	is a bug in libev (best check the archives, too :). However, don't be
		3599	annoyed when you get a brisk "this is no bug" answer and take the chance
		3600	of learning how to interpret valgrind properly.
		3601
		3602	If you need, for some reason, empty reports from valgrind for your project
		3603	I suggest using suppression lists.
		3604
		3605
		3606	=head1 PORTABILITY NOTES
		3607
		3608	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		3609
		3610	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		3611	requires, and its I/O model is fundamentally incompatible with the POSIX
		3612	model. Libev still offers limited functionality on this platform in
		3613	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		3614	descriptors. This only applies when using Win32 natively, not when using
		3615	e.g. cygwin.
		3616
		3617	Lifting these limitations would basically require the full
		3618	re-implementation of the I/O system. If you are into these kinds of
		3619	things, then note that glib does exactly that for you in a very portable
		3620	way (note also that glib is the slowest event library known to man).
		3621
		3622	There is no supported compilation method available on windows except
		3623	embedding it into other applications.
		3624
		3625	Not a libev limitation but worth mentioning: windows apparently doesn't
		3626	accept large writes: instead of resulting in a partial write, windows will
		3627	either accept everything or return C<ENOBUFS> if the buffer is too large,
		3628	so make sure you only write small amounts into your sockets (less than a
		3629	megabyte seems safe, but this apparently depends on the amount of memory
		3630	available).
		3631
		3632	Due to the many, low, and arbitrary limits on the win32 platform and
		3633	the abysmal performance of winsockets, using a large number of sockets
		3634	is not recommended (and not reasonable). If your program needs to use
		3635	more than a hundred or so sockets, then likely it needs to use a totally
		3636	different implementation for windows, as libev offers the POSIX readiness
		3637	notification model, which cannot be implemented efficiently on windows
		3638	(Microsoft monopoly games).
		3639
		3640	A typical way to use libev under windows is to embed it (see the embedding
		3641	section for details) and use the following F<evwrap.h> header file instead
		3642	of F<ev.h>:
		3643
		3644	#define EV_STANDALONE /* keeps ev from requiring config.h */
		3645	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		3646
		3647	#include "ev.h"
		3648
		3649	And compile the following F<evwrap.c> file into your project (make sure
		3650	you do I<not> compile the F<ev.c> or any other embedded source files!):
		3651
		3652	#include "evwrap.h"
		3653	#include "ev.c"
		3654
		3655	=over 4
		3656
		3657	=item The winsocket select function
		3658
		3659	The winsocket C<select> function doesn't follow POSIX in that it
		3660	requires socket I<handles> and not socket I<file descriptors> (it is
		3661	also extremely buggy). This makes select very inefficient, and also
		3662	requires a mapping from file descriptors to socket handles (the Microsoft
		3663	C runtime provides the function C<_open_osfhandle> for this). See the
		3664	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		3665	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		3666
		3667	The configuration for a "naked" win32 using the Microsoft runtime
		3668	libraries and raw winsocket select is:
		3669
		3670	#define EV_USE_SELECT 1
		3671	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		3672
		3673	Note that winsockets handling of fd sets is O(n), so you can easily get a
		3674	complexity in the O(n²) range when using win32.
		3675
		3676	=item Limited number of file descriptors
		3677
		3678	Windows has numerous arbitrary (and low) limits on things.
		3679
		3680	Early versions of winsocket's select only supported waiting for a maximum
		3681	of C<64> handles (probably owning to the fact that all windows kernels
		3682	can only wait for C<64> things at the same time internally; Microsoft
		3683	recommends spawning a chain of threads and wait for 63 handles and the
		3684	previous thread in each. Great).
		3685
		3686	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		3687	to some high number (e.g. C<2048>) before compiling the winsocket select
		3688	call (which might be in libev or elsewhere, for example, perl does its own
		3689	select emulation on windows).
		3690
		3691	Another limit is the number of file descriptors in the Microsoft runtime
		3692	libraries, which by default is C<64> (there must be a hidden I<64> fetish
		3693	or something like this inside Microsoft). You can increase this by calling
		3694	C<_setmaxstdio>, which can increase this limit to C<2048> (another
		3695	arbitrary limit), but is broken in many versions of the Microsoft runtime
		3696	libraries.
		3697
		3698	This might get you to about C<512> or C<2048> sockets (depending on
		3699	windows version and/or the phase of the moon). To get more, you need to
		3700	wrap all I/O functions and provide your own fd management, but the cost of
		3701	calling select (O(n²)) will likely make this unworkable.
		3702
		3703	=back
		3704
		3705	=head2 PORTABILITY REQUIREMENTS
		3706
		3707	In addition to a working ISO-C implementation and of course the
		3708	backend-specific APIs, libev relies on a few additional extensions:
		3709
		3710	=over 4
		3711
		3712	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		3713	calling conventions regardless of C<ev_watcher_type *>.
		3714
		3715	Libev assumes not only that all watcher pointers have the same internal
		3716	structure (guaranteed by POSIX but not by ISO C for example), but it also
		3717	assumes that the same (machine) code can be used to call any watcher
		3718	callback: The watcher callbacks have different type signatures, but libev
		3719	calls them using an C<ev_watcher *> internally.
		3720
		3721	=item C<sig_atomic_t volatile> must be thread-atomic as well
		3722
		3723	The type C<sig_atomic_t volatile> (or whatever is defined as
		3724	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		3725	threads. This is not part of the specification for C<sig_atomic_t>, but is
		3726	believed to be sufficiently portable.
		3727
		3728	=item C<sigprocmask> must work in a threaded environment
		3729
		3730	Libev uses C<sigprocmask> to temporarily block signals. This is not
		3731	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		3732	pthread implementations will either allow C<sigprocmask> in the "main
		3733	thread" or will block signals process-wide, both behaviours would
		3734	be compatible with libev. Interaction between C<sigprocmask> and
		3735	C<pthread_sigmask> could complicate things, however.
		3736
		3737	The most portable way to handle signals is to block signals in all threads
		3738	except the initial one, and run the default loop in the initial thread as
		3739	well.
		3740
		3741	=item C<long> must be large enough for common memory allocation sizes
		3742
		3743	To improve portability and simplify its API, libev uses C<long> internally
		3744	instead of C<size_t> when allocating its data structures. On non-POSIX
		3745	systems (Microsoft...) this might be unexpectedly low, but is still at
		3746	least 31 bits everywhere, which is enough for hundreds of millions of
		3747	watchers.
		3748
		3749	=item C<double> must hold a time value in seconds with enough accuracy
		3750
		3751	The type C<double> is used to represent timestamps. It is required to
		3752	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		3753	enough for at least into the year 4000. This requirement is fulfilled by
		3754	implementations implementing IEEE 754 (basically all existing ones).
		3755
		3756	=back
		3757
		3758	If you know of other additional requirements drop me a note.
		3759
		3760
		3761	=head1 ALGORITHMIC COMPLEXITIES
		3762
		3763	In this section the complexities of (many of) the algorithms used inside
		3764	libev will be documented. For complexity discussions about backends see
		3765	the documentation for C<ev_default_init>.
		3766
		3767	All of the following are about amortised time: If an array needs to be
		3768	extended, libev needs to realloc and move the whole array, but this
		3769	happens asymptotically rarer with higher number of elements, so O(1) might
		3770	mean that libev does a lengthy realloc operation in rare cases, but on
		3771	average it is much faster and asymptotically approaches constant time.
		3772
		3773	=over 4
		3774
		3775	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		3776
		3777	This means that, when you have a watcher that triggers in one hour and
		3778	there are 100 watchers that would trigger before that, then inserting will
		3779	have to skip roughly seven (C<ld 100>) of these watchers.
		3780
		3781	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		3782
		3783	That means that changing a timer costs less than removing/adding them,
		3784	as only the relative motion in the event queue has to be paid for.
		3785
		3786	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		3787
		3788	These just add the watcher into an array or at the head of a list.
		3789
		3790	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		3791
		3792	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		3793
		3794	These watchers are stored in lists, so they need to be walked to find the
		3795	correct watcher to remove. The lists are usually short (you don't usually
		3796	have many watchers waiting for the same fd or signal: one is typical, two
		3797	is rare).
		3798
		3799	=item Finding the next timer in each loop iteration: O(1)
		3800
		3801	By virtue of using a binary or 4-heap, the next timer is always found at a
		3802	fixed position in the storage array.
		3803
		3804	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		3805
		3806	A change means an I/O watcher gets started or stopped, which requires
		3807	libev to recalculate its status (and possibly tell the kernel, depending
		3808	on backend and whether C<ev_io_set> was used).
		3809
		3810	=item Activating one watcher (putting it into the pending state): O(1)
		3811
		3812	=item Priority handling: O(number_of_priorities)
		3813
		3814	Priorities are implemented by allocating some space for each
		3815	priority. When doing priority-based operations, libev usually has to
		3816	linearly search all the priorities, but starting/stopping and activating
		3817	watchers becomes O(1) with respect to priority handling.
		3818
		3819	=item Sending an ev_async: O(1)
		3820
		3821	=item Processing ev_async_send: O(number_of_async_watchers)
		3822
		3823	=item Processing signals: O(max_signal_number)
		3824
		3825	Sending involves a system call I<iff> there were no other C<ev_async_send>
		3826	calls in the current loop iteration. Checking for async and signal events
		3827	involves iterating over all running async watchers or all signal numbers.
		3828
		3829	=back
		3830
825		3831
826	=head1 AUTHOR	3832	=head1 AUTHOR
827		3833
828	Marc Lehmann <libev@schmorp.de>.	3834	Marc Lehmann <libev@schmorp.de>.
829		3835

Diff Legend

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