[ViewVC] Diff of: cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.350 by sf-exg, Mon Jan 10 08:36:41 2011 UTC vs.
Revision 1.371 by root, Sat Jun 4 05:25:03 2011 UTC

…		…
58	ev_timer_start (loop, &timeout_watcher);	58	ev_timer_start (loop, &timeout_watcher);
59		59
60	// now wait for events to arrive	60	// now wait for events to arrive
61	ev_run (loop, 0);	61	ev_run (loop, 0);
62		62
63	// unloop was called, so exit	63	// break was called, so exit
64	return 0;	64	return 0;
65	}	65	}
66		66
67	=head1 ABOUT THIS DOCUMENT	67	=head1 ABOUT THIS DOCUMENT
68		68
…		…
178	you actually want to know. Also interesting is the combination of	178	you actually want to know. Also interesting is the combination of
179	C<ev_update_now> and C<ev_now>.	179	C<ev_update_now> and C<ev_now>.
180		180
181	=item ev_sleep (ev_tstamp interval)	181	=item ev_sleep (ev_tstamp interval)
182		182
183	Sleep for the given interval: The current thread will be blocked until	183	Sleep for the given interval: The current thread will be blocked
184	either it is interrupted or the given time interval has passed. Basically	184	until either it is interrupted or the given time interval has
		185	passed (approximately - it might return a bit earlier even if not
		186	interrupted). Returns immediately if C<< interval <= 0 >>.
		187
185	this is a sub-second-resolution C<sleep ()>.	188	Basically this is a sub-second-resolution C<sleep ()>.
		189
		190	The range of the C<interval> is limited - libev only guarantees to work
		191	with sleep times of up to one day (C<< interval <= 86400 >>).
186		192
187	=item int ev_version_major ()	193	=item int ev_version_major ()
188		194
189	=item int ev_version_minor ()	195	=item int ev_version_minor ()
190		196
…		…
442		448
443	This behaviour is useful when you want to do your own signal handling, or	449	This behaviour is useful when you want to do your own signal handling, or
444	want to handle signals only in specific threads and want to avoid libev	450	want to handle signals only in specific threads and want to avoid libev
445	unblocking the signals.	451	unblocking the signals.
446		452
		453	It's also required by POSIX in a threaded program, as libev calls
		454	C<sigprocmask>, whose behaviour is officially unspecified.
		455
447	This flag's behaviour will become the default in future versions of libev.	456	This flag's behaviour will become the default in future versions of libev.
448		457
449	=item C<EVBACKEND_SELECT> (value 1, portable select backend)	458	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
450		459
451	This is your standard select(2) backend. Not I<completely> standard, as	460	This is your standard select(2) backend. Not I<completely> standard, as
…		…
480	=item C<EVBACKEND_EPOLL> (value 4, Linux)	489	=item C<EVBACKEND_EPOLL> (value 4, Linux)
481		490
482	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9	491	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
483	kernels).	492	kernels).
484		493
485	For few fds, this backend is a bit little slower than poll and select,	494	For few fds, this backend is a bit little slower than poll and select, but
486	but it scales phenomenally better. While poll and select usually scale	495	it scales phenomenally better. While poll and select usually scale like
487	like O(total_fds) where n is the total number of fds (or the highest fd),	496	O(total_fds) where total_fds is the total number of fds (or the highest
488	epoll scales either O(1) or O(active_fds).	497	fd), epoll scales either O(1) or O(active_fds).
489		498
490	The epoll mechanism deserves honorable mention as the most misdesigned	499	The epoll mechanism deserves honorable mention as the most misdesigned
491	of the more advanced event mechanisms: mere annoyances include silently	500	of the more advanced event mechanisms: mere annoyances include silently
492	dropping file descriptors, requiring a system call per change per file	501	dropping file descriptors, requiring a system call per change per file
493	descriptor (and unnecessary guessing of parameters), problems with dup,	502	descriptor (and unnecessary guessing of parameters), problems with dup,
…		…
496	0.1ms) and so on. The biggest issue is fork races, however - if a program	505	0.1ms) and so on. The biggest issue is fork races, however - if a program
497	forks then I<both> parent and child process have to recreate the epoll	506	forks then I<both> parent and child process have to recreate the epoll
498	set, which can take considerable time (one syscall per file descriptor)	507	set, which can take considerable time (one syscall per file descriptor)
499	and is of course hard to detect.	508	and is of course hard to detect.
500		509
501	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but	510	Epoll is also notoriously buggy - embedding epoll fds I<should> work,
502	of course I<doesn't>, and epoll just loves to report events for totally	511	but of course I<doesn't>, and epoll just loves to report events for
503	I<different> file descriptors (even already closed ones, so one cannot	512	totally I<different> file descriptors (even already closed ones, so
504	even remove them from the set) than registered in the set (especially	513	one cannot even remove them from the set) than registered in the set
505	on SMP systems). Libev tries to counter these spurious notifications by	514	(especially on SMP systems). Libev tries to counter these spurious
506	employing an additional generation counter and comparing that against the	515	notifications by employing an additional generation counter and comparing
507	events to filter out spurious ones, recreating the set when required. Last	516	that against the events to filter out spurious ones, recreating the set
		517	when required. Epoll also errornously rounds down timeouts, but gives you
		518	no way to know when and by how much, so sometimes you have to busy-wait
		519	because epoll returns immediately despite a nonzero timeout. And last
508	not least, it also refuses to work with some file descriptors which work	520	not least, it also refuses to work with some file descriptors which work
509	perfectly fine with C<select> (files, many character devices...).	521	perfectly fine with C<select> (files, many character devices...).
510		522
511	Epoll is truly the train wreck analog among event poll mechanisms.	523	Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
		524	cobbled together in a hurry, no thought to design or interaction with
		525	others. Oh, the pain, will it ever stop...
512		526
513	While stopping, setting and starting an I/O watcher in the same iteration	527	While stopping, setting and starting an I/O watcher in the same iteration
514	will result in some caching, there is still a system call per such	528	will result in some caching, there is still a system call per such
515	incident (because the same I<file descriptor> could point to a different	529	incident (because the same I<file descriptor> could point to a different
516	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed	530	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
…		…
582	=item C<EVBACKEND_PORT> (value 32, Solaris 10)	596	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
583		597
584	This uses the Solaris 10 event port mechanism. As with everything on Solaris,	598	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
585	it's really slow, but it still scales very well (O(active_fds)).	599	it's really slow, but it still scales very well (O(active_fds)).
586		600
587	Please note that Solaris event ports can deliver a lot of spurious
588	notifications, so you need to use non-blocking I/O or other means to avoid
589	blocking when no data (or space) is available.
590
591	While this backend scales well, it requires one system call per active	601	While this backend scales well, it requires one system call per active
592	file descriptor per loop iteration. For small and medium numbers of file	602	file descriptor per loop iteration. For small and medium numbers of file
593	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend	603	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
594	might perform better.	604	might perform better.
595		605
596	On the positive side, with the exception of the spurious readiness	606	On the positive side, this backend actually performed fully to
597	notifications, this backend actually performed fully to specification
598	in all tests and is fully embeddable, which is a rare feat among the	607	specification in all tests and is fully embeddable, which is a rare feat
599	OS-specific backends (I vastly prefer correctness over speed hacks).	608	among the OS-specific backends (I vastly prefer correctness over speed
		609	hacks).
		610
		611	On the negative side, the interface is I<bizarre> - so bizarre that
		612	even sun itself gets it wrong in their code examples: The event polling
		613	function sometimes returning events to the caller even though an error
		614	occurred, but with no indication whether it has done so or not (yes, it's
		615	even documented that way) - deadly for edge-triggered interfaces where
		616	you absolutely have to know whether an event occurred or not because you
		617	have to re-arm the watcher.
		618
		619	Fortunately libev seems to be able to work around these idiocies.
600		620
601	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as	621	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
602	C<EVBACKEND_POLL>.	622	C<EVBACKEND_POLL>.
603		623
604	=item C<EVBACKEND_ALL>	624	=item C<EVBACKEND_ALL>
…		…
814	This is useful if you are waiting for some external event in conjunction	834	This is useful if you are waiting for some external event in conjunction
815	with something not expressible using other libev watchers (i.e. "roll your	835	with something not expressible using other libev watchers (i.e. "roll your
816	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is	836	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
817	usually a better approach for this kind of thing.	837	usually a better approach for this kind of thing.
818		838
819	Here are the gory details of what C<ev_run> does:	839	Here are the gory details of what C<ev_run> does (this is for your
		840	understanding, not a guarantee that things will work exactly like this in
		841	future versions):
820		842
821	- Increment loop depth.	843	- Increment loop depth.
822	- Reset the ev_break status.	844	- Reset the ev_break status.
823	- Before the first iteration, call any pending watchers.	845	- Before the first iteration, call any pending watchers.
824	LOOP:	846	LOOP:
…		…
857	anymore.	879	anymore.
858		880
859	... queue jobs here, make sure they register event watchers as long	881	... queue jobs here, make sure they register event watchers as long
860	... as they still have work to do (even an idle watcher will do..)	882	... as they still have work to do (even an idle watcher will do..)
861	ev_run (my_loop, 0);	883	ev_run (my_loop, 0);
862	... jobs done or somebody called unloop. yeah!	884	... jobs done or somebody called break. yeah!
863		885
864	=item ev_break (loop, how)	886	=item ev_break (loop, how)
865		887
866	Can be used to make a call to C<ev_run> return early (but only after it	888	Can be used to make a call to C<ev_run> return early (but only after it
867	has processed all outstanding events). The C<how> argument must be either	889	has processed all outstanding events). The C<how> argument must be either
…		…
1349	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related	1371	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
1350	functions that do not need a watcher.	1372	functions that do not need a watcher.
1351		1373
1352	=back	1374	=back
1353		1375
1354	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1376	See also the L<ASSOCIATING CUSTOM DATA WITH A WATCHER> and L<BUILDING YOUR
1355		1377	OWN COMPOSITE WATCHERS> idioms.
1356	Each watcher has, by default, a member C<void *data> that you can change
1357	and read at any time: libev will completely ignore it. This can be used
1358	to associate arbitrary data with your watcher. If you need more data and
1359	don't want to allocate memory and store a pointer to it in that data
1360	member, you can also "subclass" the watcher type and provide your own
1361	data:
1362
1363	struct my_io
1364	{
1365	ev_io io;
1366	int otherfd;
1367	void *somedata;
1368	struct whatever *mostinteresting;
1369	};
1370
1371	...
1372	struct my_io w;
1373	ev_io_init (&w.io, my_cb, fd, EV_READ);
1374
1375	And since your callback will be called with a pointer to the watcher, you
1376	can cast it back to your own type:
1377
1378	static void my_cb (struct ev_loop loop, ev_io w_, int revents)
1379	{
1380	struct my_io w = (struct my_io )w_;
1381	...
1382	}
1383
1384	More interesting and less C-conformant ways of casting your callback type
1385	instead have been omitted.
1386
1387	Another common scenario is to use some data structure with multiple
1388	embedded watchers:
1389
1390	struct my_biggy
1391	{
1392	int some_data;
1393	ev_timer t1;
1394	ev_timer t2;
1395	}
1396
1397	In this case getting the pointer to C<my_biggy> is a bit more
1398	complicated: Either you store the address of your C<my_biggy> struct
1399	in the C<data> member of the watcher (for woozies), or you need to use
1400	some pointer arithmetic using C<offsetof> inside your watchers (for real
1401	programmers):
1402
1403	#include <stddef.h>
1404
1405	static void
1406	t1_cb (EV_P_ ev_timer *w, int revents)
1407	{
1408	struct my_biggy big = (struct my_biggy *)
1409	(((char *)w) - offsetof (struct my_biggy, t1));
1410	}
1411
1412	static void
1413	t2_cb (EV_P_ ev_timer *w, int revents)
1414	{
1415	struct my_biggy big = (struct my_biggy *)
1416	(((char *)w) - offsetof (struct my_biggy, t2));
1417	}
1418		1378
1419	=head2 WATCHER STATES	1379	=head2 WATCHER STATES
1420		1380
1421	There are various watcher states mentioned throughout this manual -	1381	There are various watcher states mentioned throughout this manual -
1422	active, pending and so on. In this section these states and the rules to	1382	active, pending and so on. In this section these states and the rules to
…		…
1429		1389
1430	Before a watcher can be registered with the event looop it has to be	1390	Before a watcher can be registered with the event looop it has to be
1431	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to	1391	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
1432	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.	1392	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
1433		1393
1434	In this state it is simply some block of memory that is suitable for use	1394	In this state it is simply some block of memory that is suitable for
1435	in an event loop. It can be moved around, freed, reused etc. at will.	1395	use in an event loop. It can be moved around, freed, reused etc. at
		1396	will - as long as you either keep the memory contents intact, or call
		1397	C<ev_TYPE_init> again.
1436		1398
1437	=item started/running/active	1399	=item started/running/active
1438		1400
1439	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes	1401	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
1440	property of the event loop, and is actively waiting for events. While in	1402	property of the event loop, and is actively waiting for events. While in
…		…
1468	latter will clear any pending state the watcher might be in, regardless	1430	latter will clear any pending state the watcher might be in, regardless
1469	of whether it was active or not, so stopping a watcher explicitly before	1431	of whether it was active or not, so stopping a watcher explicitly before
1470	freeing it is often a good idea.	1432	freeing it is often a good idea.
1471		1433
1472	While stopped (and not pending) the watcher is essentially in the	1434	While stopped (and not pending) the watcher is essentially in the
1473	initialised state, that is it can be reused, moved, modified in any way	1435	initialised state, that is, it can be reused, moved, modified in any way
1474	you wish.	1436	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1437	it again).
1475		1438
1476	=back	1439	=back
1477		1440
1478	=head2 WATCHER PRIORITY MODELS	1441	=head2 WATCHER PRIORITY MODELS
1479		1442
…		…
1608	In general you can register as many read and/or write event watchers per	1571	In general you can register as many read and/or write event watchers per
1609	fd as you want (as long as you don't confuse yourself). Setting all file	1572	fd as you want (as long as you don't confuse yourself). Setting all file
1610	descriptors to non-blocking mode is also usually a good idea (but not	1573	descriptors to non-blocking mode is also usually a good idea (but not
1611	required if you know what you are doing).	1574	required if you know what you are doing).
1612		1575
1613	If you cannot use non-blocking mode, then force the use of a
1614	known-to-be-good backend (at the time of this writing, this includes only
1615	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
1616	descriptors for which non-blocking operation makes no sense (such as
1617	files) - libev doesn't guarantee any specific behaviour in that case.
1618
1619	Another thing you have to watch out for is that it is quite easy to	1576	Another thing you have to watch out for is that it is quite easy to
1620	receive "spurious" readiness notifications, that is your callback might	1577	receive "spurious" readiness notifications, that is, your callback might
1621	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1578	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1622	because there is no data. Not only are some backends known to create a	1579	because there is no data. It is very easy to get into this situation even
1623	lot of those (for example Solaris ports), it is very easy to get into	1580	with a relatively standard program structure. Thus it is best to always
1624	this situation even with a relatively standard program structure. Thus	1581	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
1625	it is best to always use non-blocking I/O: An extra C<read>(2) returning
1626	C<EAGAIN> is far preferable to a program hanging until some data arrives.	1582	preferable to a program hanging until some data arrives.
1627		1583
1628	If you cannot run the fd in non-blocking mode (for example you should	1584	If you cannot run the fd in non-blocking mode (for example you should
1629	not play around with an Xlib connection), then you have to separately	1585	not play around with an Xlib connection), then you have to separately
1630	re-test whether a file descriptor is really ready with a known-to-be good	1586	re-test whether a file descriptor is really ready with a known-to-be good
1631	interface such as poll (fortunately in our Xlib example, Xlib already	1587	interface such as poll (fortunately in the case of Xlib, it already does
1632	does this on its own, so its quite safe to use). Some people additionally	1588	this on its own, so its quite safe to use). Some people additionally
1633	use C<SIGALRM> and an interval timer, just to be sure you won't block	1589	use C<SIGALRM> and an interval timer, just to be sure you won't block
1634	indefinitely.	1590	indefinitely.
1635		1591
1636	But really, best use non-blocking mode.	1592	But really, best use non-blocking mode.
1637		1593
…		…
1665		1621
1666	There is no workaround possible except not registering events	1622	There is no workaround possible except not registering events
1667	for potentially C<dup ()>'ed file descriptors, or to resort to	1623	for potentially C<dup ()>'ed file descriptors, or to resort to
1668	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.	1624	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1669		1625
		1626	=head3 The special problem of files
		1627
		1628	Many people try to use C<select> (or libev) on file descriptors
		1629	representing files, and expect it to become ready when their program
		1630	doesn't block on disk accesses (which can take a long time on their own).
		1631
		1632	However, this cannot ever work in the "expected" way - you get a readiness
		1633	notification as soon as the kernel knows whether and how much data is
		1634	there, and in the case of open files, that's always the case, so you
		1635	always get a readiness notification instantly, and your read (or possibly
		1636	write) will still block on the disk I/O.
		1637
		1638	Another way to view it is that in the case of sockets, pipes, character
		1639	devices and so on, there is another party (the sender) that delivers data
		1640	on its own, but in the case of files, there is no such thing: the disk
		1641	will not send data on its own, simply because it doesn't know what you
		1642	wish to read - you would first have to request some data.
		1643
		1644	Since files are typically not-so-well supported by advanced notification
		1645	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1646	to files, even though you should not use it. The reason for this is
		1647	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1648	usually a tty, often a pipe, but also sometimes files or special devices
		1649	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1650	F</dev/urandom>), and even though the file might better be served with
		1651	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1652	it "just works" instead of freezing.
		1653
		1654	So avoid file descriptors pointing to files when you know it (e.g. use
		1655	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1656	when you rarely read from a file instead of from a socket, and want to
		1657	reuse the same code path.
		1658
1670	=head3 The special problem of fork	1659	=head3 The special problem of fork
1671		1660
1672	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit	1661	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1673	useless behaviour. Libev fully supports fork, but needs to be told about	1662	useless behaviour. Libev fully supports fork, but needs to be told about
1674	it in the child.	1663	it in the child if you want to continue to use it in the child.
1675		1664
1676	To support fork in your programs, you either have to call	1665	To support fork in your child processes, you have to call C<ev_loop_fork
1677	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,	1666	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
1678	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or	1667	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1679	C<EVBACKEND_POLL>.
1680		1668
1681	=head3 The special problem of SIGPIPE	1669	=head3 The special problem of SIGPIPE
1682		1670
1683	While not really specific to libev, it is easy to forget about C<SIGPIPE>:	1671	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1684	when writing to a pipe whose other end has been closed, your program gets	1672	when writing to a pipe whose other end has been closed, your program gets
…		…
2174		2162
2175	Another way to think about it (for the mathematically inclined) is that	2163	Another way to think about it (for the mathematically inclined) is that
2176	C<ev_periodic> will try to run the callback in this mode at the next possible	2164	C<ev_periodic> will try to run the callback in this mode at the next possible
2177	time where C<time = offset (mod interval)>, regardless of any time jumps.	2165	time where C<time = offset (mod interval)>, regardless of any time jumps.
2178		2166
2179	For numerical stability it is preferable that the C<offset> value is near	2167	The C<interval> I<MUST> be positive, and for numerical stability, the
2180	C<ev_now ()> (the current time), but there is no range requirement for	2168	interval value should be higher than C<1/8192> (which is around 100
2181	this value, and in fact is often specified as zero.	2169	microseconds) and C<offset> should be higher than C<0> and should have
		2170	at most a similar magnitude as the current time (say, within a factor of
		2171	ten). Typical values for offset are, in fact, C<0> or something between
		2172	C<0> and C<interval>, which is also the recommended range.
2182		2173
2183	Note also that there is an upper limit to how often a timer can fire (CPU	2174	Note also that there is an upper limit to how often a timer can fire (CPU
2184	speed for example), so if C<interval> is very small then timing stability	2175	speed for example), so if C<interval> is very small then timing stability
2185	will of course deteriorate. Libev itself tries to be exact to be about one	2176	will of course deteriorate. Libev itself tries to be exact to be about one
2186	millisecond (if the OS supports it and the machine is fast enough).	2177	millisecond (if the OS supports it and the machine is fast enough).
…		…
2329	=head3 The special problem of inheritance over fork/execve/pthread_create	2320	=head3 The special problem of inheritance over fork/execve/pthread_create
2330		2321
2331	Both the signal mask (C<sigprocmask>) and the signal disposition	2322	Both the signal mask (C<sigprocmask>) and the signal disposition
2332	(C<sigaction>) are unspecified after starting a signal watcher (and after	2323	(C<sigaction>) are unspecified after starting a signal watcher (and after
2333	stopping it again), that is, libev might or might not block the signal,	2324	stopping it again), that is, libev might or might not block the signal,
2334	and might or might not set or restore the installed signal handler.	2325	and might or might not set or restore the installed signal handler (but
		2326	see C<EVFLAG_NOSIGMASK>).
2335		2327
2336	While this does not matter for the signal disposition (libev never	2328	While this does not matter for the signal disposition (libev never
2337	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on	2329	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
2338	C<execve>), this matters for the signal mask: many programs do not expect	2330	C<execve>), this matters for the signal mask: many programs do not expect
2339	certain signals to be blocked.	2331	certain signals to be blocked.
…		…
3210	atexit (program_exits);	3202	atexit (program_exits);
3211		3203
3212		3204
3213	=head2 C<ev_async> - how to wake up an event loop	3205	=head2 C<ev_async> - how to wake up an event loop
3214		3206
3215	In general, you cannot use an C<ev_run> from multiple threads or other	3207	In general, you cannot use an C<ev_loop> from multiple threads or other
3216	asynchronous sources such as signal handlers (as opposed to multiple event	3208	asynchronous sources such as signal handlers (as opposed to multiple event
3217	loops - those are of course safe to use in different threads).	3209	loops - those are of course safe to use in different threads).
3218		3210
3219	Sometimes, however, you need to wake up an event loop you do not control,	3211	Sometimes, however, you need to wake up an event loop you do not control,
3220	for example because it belongs to another thread. This is what C<ev_async>	3212	for example because it belongs to another thread. This is what C<ev_async>
…		…
3330	trust me.	3322	trust me.
3331		3323
3332	=item ev_async_send (loop, ev_async *)	3324	=item ev_async_send (loop, ev_async *)
3333		3325
3334	Sends/signals/activates the given C<ev_async> watcher, that is, feeds	3326	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
3335	an C<EV_ASYNC> event on the watcher into the event loop. Unlike	3327	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3328	returns.
		3329
3336	C<ev_feed_event>, this call is safe to do from other threads, signal or	3330	Unlike C<ev_feed_event>, this call is safe to do from other threads,
3337	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding	3331	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
3338	section below on what exactly this means).	3332	embedding section below on what exactly this means).
3339		3333
3340	Note that, as with other watchers in libev, multiple events might get	3334	Note that, as with other watchers in libev, multiple events might get
3341	compressed into a single callback invocation (another way to look at this	3335	compressed into a single callback invocation (another way to look at this
3342	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,	3336	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
3343	reset when the event loop detects that).	3337	reset when the event loop detects that).
…		…
3423		3417
3424	This section explains some common idioms that are not immediately	3418	This section explains some common idioms that are not immediately
3425	obvious. Note that examples are sprinkled over the whole manual, and this	3419	obvious. Note that examples are sprinkled over the whole manual, and this
3426	section only contains stuff that wouldn't fit anywhere else.	3420	section only contains stuff that wouldn't fit anywhere else.
3427		3421
3428	=over 4	3422	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
3429		3423
3430	=item Model/nested event loop invocations and exit conditions.	3424	Each watcher has, by default, a C<void *data> member that you can read
		3425	or modify at any time: libev will completely ignore it. This can be used
		3426	to associate arbitrary data with your watcher. If you need more data and
		3427	don't want to allocate memory separately and store a pointer to it in that
		3428	data member, you can also "subclass" the watcher type and provide your own
		3429	data:
		3430
		3431	struct my_io
		3432	{
		3433	ev_io io;
		3434	int otherfd;
		3435	void *somedata;
		3436	struct whatever *mostinteresting;
		3437	};
		3438
		3439	...
		3440	struct my_io w;
		3441	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3442
		3443	And since your callback will be called with a pointer to the watcher, you
		3444	can cast it back to your own type:
		3445
		3446	static void my_cb (struct ev_loop loop, ev_io w_, int revents)
		3447	{
		3448	struct my_io w = (struct my_io )w_;
		3449	...
		3450	}
		3451
		3452	More interesting and less C-conformant ways of casting your callback
		3453	function type instead have been omitted.
		3454
		3455	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3456
		3457	Another common scenario is to use some data structure with multiple
		3458	embedded watchers, in effect creating your own watcher that combines
		3459	multiple libev event sources into one "super-watcher":
		3460
		3461	struct my_biggy
		3462	{
		3463	int some_data;
		3464	ev_timer t1;
		3465	ev_timer t2;
		3466	}
		3467
		3468	In this case getting the pointer to C<my_biggy> is a bit more
		3469	complicated: Either you store the address of your C<my_biggy> struct in
		3470	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3471	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3472	real programmers):
		3473
		3474	#include <stddef.h>
		3475
		3476	static void
		3477	t1_cb (EV_P_ ev_timer *w, int revents)
		3478	{
		3479	struct my_biggy big = (struct my_biggy *)
		3480	(((char *)w) - offsetof (struct my_biggy, t1));
		3481	}
		3482
		3483	static void
		3484	t2_cb (EV_P_ ev_timer *w, int revents)
		3485	{
		3486	struct my_biggy big = (struct my_biggy *)
		3487	(((char *)w) - offsetof (struct my_biggy, t2));
		3488	}
		3489
		3490	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
3431		3491
3432	Often (especially in GUI toolkits) there are places where you have	3492	Often (especially in GUI toolkits) there are places where you have
3433	I<modal> interaction, which is most easily implemented by recursively	3493	I<modal> interaction, which is most easily implemented by recursively
3434	invoking C<ev_run>.	3494	invoking C<ev_run>.
3435		3495
…		…
3464	exit_main_loop = 1;	3524	exit_main_loop = 1;
3465		3525
3466	// exit both	3526	// exit both
3467	exit_main_loop = exit_nested_loop = 1;	3527	exit_main_loop = exit_nested_loop = 1;
3468		3528
3469	=back	3529	=head2 THREAD LOCKING EXAMPLE
		3530
		3531	Here is a fictitious example of how to run an event loop in a different
		3532	thread from where callbacks are being invoked and watchers are
		3533	created/added/removed.
		3534
		3535	For a real-world example, see the C<EV::Loop::Async> perl module,
		3536	which uses exactly this technique (which is suited for many high-level
		3537	languages).
		3538
		3539	The example uses a pthread mutex to protect the loop data, a condition
		3540	variable to wait for callback invocations, an async watcher to notify the
		3541	event loop thread and an unspecified mechanism to wake up the main thread.
		3542
		3543	First, you need to associate some data with the event loop:
		3544
		3545	typedef struct {
		3546	mutex_t lock; /* global loop lock */
		3547	ev_async async_w;
		3548	thread_t tid;
		3549	cond_t invoke_cv;
		3550	} userdata;
		3551
		3552	void prepare_loop (EV_P)
		3553	{
		3554	// for simplicity, we use a static userdata struct.
		3555	static userdata u;
		3556
		3557	ev_async_init (&u->async_w, async_cb);
		3558	ev_async_start (EV_A_ &u->async_w);
		3559
		3560	pthread_mutex_init (&u->lock, 0);
		3561	pthread_cond_init (&u->invoke_cv, 0);
		3562
		3563	// now associate this with the loop
		3564	ev_set_userdata (EV_A_ u);
		3565	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3566	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3567
		3568	// then create the thread running ev_run
		3569	pthread_create (&u->tid, 0, l_run, EV_A);
		3570	}
		3571
		3572	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3573	solely to wake up the event loop so it takes notice of any new watchers
		3574	that might have been added:
		3575
		3576	static void
		3577	async_cb (EV_P_ ev_async *w, int revents)
		3578	{
		3579	// just used for the side effects
		3580	}
		3581
		3582	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3583	protecting the loop data, respectively.
		3584
		3585	static void
		3586	l_release (EV_P)
		3587	{
		3588	userdata *u = ev_userdata (EV_A);
		3589	pthread_mutex_unlock (&u->lock);
		3590	}
		3591
		3592	static void
		3593	l_acquire (EV_P)
		3594	{
		3595	userdata *u = ev_userdata (EV_A);
		3596	pthread_mutex_lock (&u->lock);
		3597	}
		3598
		3599	The event loop thread first acquires the mutex, and then jumps straight
		3600	into C<ev_run>:
		3601
		3602	void *
		3603	l_run (void *thr_arg)
		3604	{
		3605	struct ev_loop loop = (struct ev_loop )thr_arg;
		3606
		3607	l_acquire (EV_A);
		3608	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3609	ev_run (EV_A_ 0);
		3610	l_release (EV_A);
		3611
		3612	return 0;
		3613	}
		3614
		3615	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3616	signal the main thread via some unspecified mechanism (signals? pipe
		3617	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3618	have been called (in a while loop because a) spurious wakeups are possible
		3619	and b) skipping inter-thread-communication when there are no pending
		3620	watchers is very beneficial):
		3621
		3622	static void
		3623	l_invoke (EV_P)
		3624	{
		3625	userdata *u = ev_userdata (EV_A);
		3626
		3627	while (ev_pending_count (EV_A))
		3628	{
		3629	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3630	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3631	}
		3632	}
		3633
		3634	Now, whenever the main thread gets told to invoke pending watchers, it
		3635	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3636	thread to continue:
		3637
		3638	static void
		3639	real_invoke_pending (EV_P)
		3640	{
		3641	userdata *u = ev_userdata (EV_A);
		3642
		3643	pthread_mutex_lock (&u->lock);
		3644	ev_invoke_pending (EV_A);
		3645	pthread_cond_signal (&u->invoke_cv);
		3646	pthread_mutex_unlock (&u->lock);
		3647	}
		3648
		3649	Whenever you want to start/stop a watcher or do other modifications to an
		3650	event loop, you will now have to lock:
		3651
		3652	ev_timer timeout_watcher;
		3653	userdata *u = ev_userdata (EV_A);
		3654
		3655	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3656
		3657	pthread_mutex_lock (&u->lock);
		3658	ev_timer_start (EV_A_ &timeout_watcher);
		3659	ev_async_send (EV_A_ &u->async_w);
		3660	pthread_mutex_unlock (&u->lock);
		3661
		3662	Note that sending the C<ev_async> watcher is required because otherwise
		3663	an event loop currently blocking in the kernel will have no knowledge
		3664	about the newly added timer. By waking up the loop it will pick up any new
		3665	watchers in the next event loop iteration.
		3666
		3667	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3668
		3669	While the overhead of a callback that e.g. schedules a thread is small, it
		3670	is still an overhead. If you embed libev, and your main usage is with some
		3671	kind of threads or coroutines, you might want to customise libev so that
		3672	doesn't need callbacks anymore.
		3673
		3674	Imagine you have coroutines that you can switch to using a function
		3675	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3676	and that due to some magic, the currently active coroutine is stored in a
		3677	global called C<current_coro>. Then you can build your own "wait for libev
		3678	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3679	the differing C<;> conventions):
		3680
		3681	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3682	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3683
		3684	That means instead of having a C callback function, you store the
		3685	coroutine to switch to in each watcher, and instead of having libev call
		3686	your callback, you instead have it switch to that coroutine.
		3687
		3688	A coroutine might now wait for an event with a function called
		3689	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3690	matter when, or whether the watcher is active or not when this function is
		3691	called):
		3692
		3693	void
		3694	wait_for_event (ev_watcher *w)
		3695	{
		3696	ev_cb_set (w) = current_coro;
		3697	switch_to (libev_coro);
		3698	}
		3699
		3700	That basically suspends the coroutine inside C<wait_for_event> and
		3701	continues the libev coroutine, which, when appropriate, switches back to
		3702	this or any other coroutine. I am sure if you sue this your own :)
		3703
		3704	You can do similar tricks if you have, say, threads with an event queue -
		3705	instead of storing a coroutine, you store the queue object and instead of
		3706	switching to a coroutine, you push the watcher onto the queue and notify
		3707	any waiters.
		3708
		3709	To embed libev, see L<EMBEDDING>, but in short, it's easiest to create two
		3710	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3711
		3712	// my_ev.h
		3713	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3714	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb);
		3715	#include "../libev/ev.h"
		3716
		3717	// my_ev.c
		3718	#define EV_H "my_ev.h"
		3719	#include "../libev/ev.c"
		3720
		3721	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		3722	F<my_ev.c> into your project. When properly specifying include paths, you
		3723	can even use F<ev.h> as header file name directly.
3470		3724
3471		3725
3472	=head1 LIBEVENT EMULATION	3726	=head1 LIBEVENT EMULATION
3473		3727
3474	Libev offers a compatibility emulation layer for libevent. It cannot	3728	Libev offers a compatibility emulation layer for libevent. It cannot
…		…
3964	F<event.h> that are not directly supported by the libev core alone.	4218	F<event.h> that are not directly supported by the libev core alone.
3965		4219
3966	In standalone mode, libev will still try to automatically deduce the	4220	In standalone mode, libev will still try to automatically deduce the
3967	configuration, but has to be more conservative.	4221	configuration, but has to be more conservative.
3968		4222
		4223	=item EV_USE_FLOOR
		4224
		4225	If defined to be C<1>, libev will use the C<floor ()> function for its
		4226	periodic reschedule calculations, otherwise libev will fall back on a
		4227	portable (slower) implementation. If you enable this, you usually have to
		4228	link against libm or something equivalent. Enabling this when the C<floor>
		4229	function is not available will fail, so the safe default is to not enable
		4230	this.
		4231
3969	=item EV_USE_MONOTONIC	4232	=item EV_USE_MONOTONIC
3970		4233
3971	If defined to be C<1>, libev will try to detect the availability of the	4234	If defined to be C<1>, libev will try to detect the availability of the
3972	monotonic clock option at both compile time and runtime. Otherwise no	4235	monotonic clock option at both compile time and runtime. Otherwise no
3973	use of the monotonic clock option will be attempted. If you enable this,	4236	use of the monotonic clock option will be attempted. If you enable this,
…		…
4404	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	4667	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
4405		4668
4406	#include "ev_cpp.h"	4669	#include "ev_cpp.h"
4407	#include "ev.c"	4670	#include "ev.c"
4408		4671
4409	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES	4672	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
4410		4673
4411	=head2 THREADS AND COROUTINES	4674	=head2 THREADS AND COROUTINES
4412		4675
4413	=head3 THREADS	4676	=head3 THREADS
4414		4677
…		…
4465	default loop and triggering an C<ev_async> watcher from the default loop	4728	default loop and triggering an C<ev_async> watcher from the default loop
4466	watcher callback into the event loop interested in the signal.	4729	watcher callback into the event loop interested in the signal.
4467		4730
4468	=back	4731	=back
4469		4732
4470	=head4 THREAD LOCKING EXAMPLE	4733	See also L<THREAD LOCKING EXAMPLE>.
4471
4472	Here is a fictitious example of how to run an event loop in a different
4473	thread than where callbacks are being invoked and watchers are
4474	created/added/removed.
4475
4476	For a real-world example, see the C<EV::Loop::Async> perl module,
4477	which uses exactly this technique (which is suited for many high-level
4478	languages).
4479
4480	The example uses a pthread mutex to protect the loop data, a condition
4481	variable to wait for callback invocations, an async watcher to notify the
4482	event loop thread and an unspecified mechanism to wake up the main thread.
4483
4484	First, you need to associate some data with the event loop:
4485
4486	typedef struct {
4487	mutex_t lock; /* global loop lock */
4488	ev_async async_w;
4489	thread_t tid;
4490	cond_t invoke_cv;
4491	} userdata;
4492
4493	void prepare_loop (EV_P)
4494	{
4495	// for simplicity, we use a static userdata struct.
4496	static userdata u;
4497
4498	ev_async_init (&u->async_w, async_cb);
4499	ev_async_start (EV_A_ &u->async_w);
4500
4501	pthread_mutex_init (&u->lock, 0);
4502	pthread_cond_init (&u->invoke_cv, 0);
4503
4504	// now associate this with the loop
4505	ev_set_userdata (EV_A_ u);
4506	ev_set_invoke_pending_cb (EV_A_ l_invoke);
4507	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
4508
4509	// then create the thread running ev_loop
4510	pthread_create (&u->tid, 0, l_run, EV_A);
4511	}
4512
4513	The callback for the C<ev_async> watcher does nothing: the watcher is used
4514	solely to wake up the event loop so it takes notice of any new watchers
4515	that might have been added:
4516
4517	static void
4518	async_cb (EV_P_ ev_async *w, int revents)
4519	{
4520	// just used for the side effects
4521	}
4522
4523	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
4524	protecting the loop data, respectively.
4525
4526	static void
4527	l_release (EV_P)
4528	{
4529	userdata *u = ev_userdata (EV_A);
4530	pthread_mutex_unlock (&u->lock);
4531	}
4532
4533	static void
4534	l_acquire (EV_P)
4535	{
4536	userdata *u = ev_userdata (EV_A);
4537	pthread_mutex_lock (&u->lock);
4538	}
4539
4540	The event loop thread first acquires the mutex, and then jumps straight
4541	into C<ev_run>:
4542
4543	void *
4544	l_run (void *thr_arg)
4545	{
4546	struct ev_loop loop = (struct ev_loop )thr_arg;
4547
4548	l_acquire (EV_A);
4549	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
4550	ev_run (EV_A_ 0);
4551	l_release (EV_A);
4552
4553	return 0;
4554	}
4555
4556	Instead of invoking all pending watchers, the C<l_invoke> callback will
4557	signal the main thread via some unspecified mechanism (signals? pipe
4558	writes? C<Async::Interrupt>?) and then waits until all pending watchers
4559	have been called (in a while loop because a) spurious wakeups are possible
4560	and b) skipping inter-thread-communication when there are no pending
4561	watchers is very beneficial):
4562
4563	static void
4564	l_invoke (EV_P)
4565	{
4566	userdata *u = ev_userdata (EV_A);
4567
4568	while (ev_pending_count (EV_A))
4569	{
4570	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
4571	pthread_cond_wait (&u->invoke_cv, &u->lock);
4572	}
4573	}
4574
4575	Now, whenever the main thread gets told to invoke pending watchers, it
4576	will grab the lock, call C<ev_invoke_pending> and then signal the loop
4577	thread to continue:
4578
4579	static void
4580	real_invoke_pending (EV_P)
4581	{
4582	userdata *u = ev_userdata (EV_A);
4583
4584	pthread_mutex_lock (&u->lock);
4585	ev_invoke_pending (EV_A);
4586	pthread_cond_signal (&u->invoke_cv);
4587	pthread_mutex_unlock (&u->lock);
4588	}
4589
4590	Whenever you want to start/stop a watcher or do other modifications to an
4591	event loop, you will now have to lock:
4592
4593	ev_timer timeout_watcher;
4594	userdata *u = ev_userdata (EV_A);
4595
4596	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
4597
4598	pthread_mutex_lock (&u->lock);
4599	ev_timer_start (EV_A_ &timeout_watcher);
4600	ev_async_send (EV_A_ &u->async_w);
4601	pthread_mutex_unlock (&u->lock);
4602
4603	Note that sending the C<ev_async> watcher is required because otherwise
4604	an event loop currently blocking in the kernel will have no knowledge
4605	about the newly added timer. By waking up the loop it will pick up any new
4606	watchers in the next event loop iteration.
4607		4734
4608	=head3 COROUTINES	4735	=head3 COROUTINES
4609		4736
4610	Libev is very accommodating to coroutines ("cooperative threads"):	4737	Libev is very accommodating to coroutines ("cooperative threads"):
4611	libev fully supports nesting calls to its functions from different	4738	libev fully supports nesting calls to its functions from different
…		…
5120	The physical time that is observed. It is apparently strictly monotonic :)	5247	The physical time that is observed. It is apparently strictly monotonic :)
5121		5248
5122	=item wall-clock time	5249	=item wall-clock time
5123		5250
5124	The time and date as shown on clocks. Unlike real time, it can actually	5251	The time and date as shown on clocks. Unlike real time, it can actually
5125	be wrong and jump forwards and backwards, e.g. when the you adjust your	5252	be wrong and jump forwards and backwards, e.g. when you adjust your
5126	clock.	5253	clock.
5127		5254
5128	=item watcher	5255	=item watcher
5129		5256
5130	A data structure that describes interest in certain events. Watchers need	5257	A data structure that describes interest in certain events. Watchers need
…		…
5133	=back	5260	=back
5134		5261
5135	=head1 AUTHOR	5262	=head1 AUTHOR
5136		5263
5137	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael	5264	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
5138	Magnusson and Emanuele Giaquinta.	5265	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
5139		5266

Diff Legend

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

Comparing libev/ev.pod (file contents): Revision 1.350 by sf-exg, Mon Jan 10 08:36:41 2011 UTC vs. Revision 1.371 by root, Sat Jun 4 05:25:03 2011 UTC

Diff Legend

Comparing libev/ev.pod (file contents):
Revision 1.350 by sf-exg, Mon Jan 10 08:36:41 2011 UTC vs.
Revision 1.371 by root, Sat Jun 4 05:25:03 2011 UTC