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

Comparing libev/ev.pod (file contents):
Revision 1.138 by root, Mon Mar 31 01:14:12 2008 UTC vs.
Revision 1.158 by root, Wed May 21 12:51:38 2008 UTC

…		…
64		64
65	=head1 DESCRIPTION	65	=head1 DESCRIPTION
66		66
67	The newest version of this document is also available as an html-formatted	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	68	web page you might find easier to navigate when reading it for the first
69	time: L<http://cvs.schmorp.de/libev/ev.html>.	69	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
70		70
71	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
72	file descriptor being readable or a timeout occurring), and it will manage	72	file descriptor being readable or a timeout occurring), and it will manage
73	these event sources and provide your program with events.	73	these event sources and provide your program with events.
74		74
…		…
196	See the description of C<ev_embed> watchers for more info.	196	See the description of C<ev_embed> watchers for more info.
197		197
198	=item ev_set_allocator (void (cb)(void *ptr, long size))	198	=item ev_set_allocator (void (cb)(void *ptr, long size))
199		199
200	Sets the allocation function to use (the prototype is similar - the	200	Sets the allocation function to use (the prototype is similar - the
201	semantics is identical - to the realloc C function). It is used to	201	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
202	allocate and free memory (no surprises here). If it returns zero when	202	used to allocate and free memory (no surprises here). If it returns zero
203	memory needs to be allocated, the library might abort or take some	203	when memory needs to be allocated (C<size != 0>), the library might abort
204	potentially destructive action. The default is your system realloc	204	or take some potentially destructive action.
205	function.	205
		206	Since some systems (at least OpenBSD and Darwin) fail to implement
		207	correct C<realloc> semantics, libev will use a wrapper around the system
		208	C<realloc> and C<free> functions by default.
206		209
207	You could override this function in high-availability programs to, say,	210	You could override this function in high-availability programs to, say,
208	free some memory if it cannot allocate memory, to use a special allocator,	211	free some memory if it cannot allocate memory, to use a special allocator,
209	or even to sleep a while and retry until some memory is available.	212	or even to sleep a while and retry until some memory is available.
210		213
211	Example: Replace the libev allocator with one that waits a bit and then	214	Example: Replace the libev allocator with one that waits a bit and then
212	retries).	215	retries (example requires a standards-compliant C<realloc>).
213		216
214	static void *	217	static void *
215	persistent_realloc (void *ptr, size_t size)	218	persistent_realloc (void *ptr, size_t size)
216	{	219	{
217	for (;;)	220	for (;;)
…		…
256		259
257	An event loop is described by a C<struct ev_loop *>. The library knows two	260	An event loop is described by a C<struct ev_loop *>. The library knows two
258	types of such loops, the I<default> loop, which supports signals and child	261	types of such loops, the I<default> loop, which supports signals and child
259	events, and dynamically created loops which do not.	262	events, and dynamically created loops which do not.
260		263
261	If you use threads, a common model is to run the default event loop
262	in your main thread (or in a separate thread) and for each thread you
263	create, you also create another event loop. Libev itself does no locking
264	whatsoever, so if you mix calls to the same event loop in different
265	threads, make sure you lock (this is usually a bad idea, though, even if
266	done correctly, because it's hideous and inefficient).
267
268	=over 4	264	=over 4
269		265
270	=item struct ev_loop *ev_default_loop (unsigned int flags)	266	=item struct ev_loop *ev_default_loop (unsigned int flags)
271		267
272	This will initialise the default event loop if it hasn't been initialised	268	This will initialise the default event loop if it hasn't been initialised
…		…
274	false. If it already was initialised it simply returns it (and ignores the	270	false. If it already was initialised it simply returns it (and ignores the
275	flags. If that is troubling you, check C<ev_backend ()> afterwards).	271	flags. If that is troubling you, check C<ev_backend ()> afterwards).
276		272
277	If you don't know what event loop to use, use the one returned from this	273	If you don't know what event loop to use, use the one returned from this
278	function.	274	function.
		275
		276	Note that this function is I<not> thread-safe, so if you want to use it
		277	from multiple threads, you have to lock (note also that this is unlikely,
		278	as loops cannot bes hared easily between threads anyway).
279		279
280	The default loop is the only loop that can handle C<ev_signal> and	280	The default loop is the only loop that can handle C<ev_signal> and
281	C<ev_child> watchers, and to do this, it always registers a handler	281	C<ev_child> watchers, and to do this, it always registers a handler
282	for C<SIGCHLD>. If this is a problem for your app you can either	282	for C<SIGCHLD>. If this is a problem for your app you can either
283	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you	283	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
…		…
336	To get good performance out of this backend you need a high amount of	336	To get good performance out of this backend you need a high amount of
337	parallelity (most of the file descriptors should be busy). If you are	337	parallelity (most of the file descriptors should be busy). If you are
338	writing a server, you should C<accept ()> in a loop to accept as many	338	writing a server, you should C<accept ()> in a loop to accept as many
339	connections as possible during one iteration. You might also want to have	339	connections as possible during one iteration. You might also want to have
340	a look at C<ev_set_io_collect_interval ()> to increase the amount of	340	a look at C<ev_set_io_collect_interval ()> to increase the amount of
341	readyness notifications you get per iteration.	341	readiness notifications you get per iteration.
342		342
343	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)	343	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
344		344
345	And this is your standard poll(2) backend. It's more complicated	345	And this is your standard poll(2) backend. It's more complicated
346	than select, but handles sparse fds better and has no artificial	346	than select, but handles sparse fds better and has no artificial
…		…
354	For few fds, this backend is a bit little slower than poll and select,	354	For few fds, this backend is a bit little slower than poll and select,
355	but it scales phenomenally better. While poll and select usually scale	355	but it scales phenomenally better. While poll and select usually scale
356	like O(total_fds) where n is the total number of fds (or the highest fd),	356	like O(total_fds) where n is the total number of fds (or the highest fd),
357	epoll scales either O(1) or O(active_fds). The epoll design has a number	357	epoll scales either O(1) or O(active_fds). The epoll design has a number
358	of shortcomings, such as silently dropping events in some hard-to-detect	358	of shortcomings, such as silently dropping events in some hard-to-detect
359	cases and rewiring a syscall per fd change, no fork support and bad	359	cases and requiring a syscall per fd change, no fork support and bad
360	support for dup.	360	support for dup.
361		361
362	While stopping, setting and starting an I/O watcher in the same iteration	362	While stopping, setting and starting an I/O watcher in the same iteration
363	will result in some caching, there is still a syscall per such incident	363	will result in some caching, there is still a syscall per such incident
364	(because the fd could point to a different file description now), so its	364	(because the fd could point to a different file description now), so its
…		…
425	While this backend scales well, it requires one system call per active	425	While this backend scales well, it requires one system call per active
426	file descriptor per loop iteration. For small and medium numbers of file	426	file descriptor per loop iteration. For small and medium numbers of file
427	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend	427	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
428	might perform better.	428	might perform better.
429		429
430	On the positive side, ignoring the spurious readyness notifications, this	430	On the positive side, ignoring the spurious readiness notifications, this
431	backend actually performed to specification in all tests and is fully	431	backend actually performed to specification in all tests and is fully
432	embeddable, which is a rare feat among the OS-specific backends.	432	embeddable, which is a rare feat among the OS-specific backends.
433		433
434	=item C<EVBACKEND_ALL>	434	=item C<EVBACKEND_ALL>
435		435
…		…
465		465
466	Similar to C<ev_default_loop>, but always creates a new event loop that is	466	Similar to C<ev_default_loop>, but always creates a new event loop that is
467	always distinct from the default loop. Unlike the default loop, it cannot	467	always distinct from the default loop. Unlike the default loop, it cannot
468	handle signal and child watchers, and attempts to do so will be greeted by	468	handle signal and child watchers, and attempts to do so will be greeted by
469	undefined behaviour (or a failed assertion if assertions are enabled).	469	undefined behaviour (or a failed assertion if assertions are enabled).
		470
		471	Note that this function I<is> thread-safe, and the recommended way to use
		472	libev with threads is indeed to create one loop per thread, and using the
		473	default loop in the "main" or "initial" thread.
470		474
471	Example: Try to create a event loop that uses epoll and nothing else.	475	Example: Try to create a event loop that uses epoll and nothing else.
472		476
473	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL \| EVFLAG_NOENV);	477	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL \| EVFLAG_NOENV);
474	if (!epoller)	478	if (!epoller)
…		…
1028	If you must do this, then force the use of a known-to-be-good backend	1032	If you must do this, then force the use of a known-to-be-good backend
1029	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and	1033	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
1030	C<EVBACKEND_POLL>).	1034	C<EVBACKEND_POLL>).
1031		1035
1032	Another thing you have to watch out for is that it is quite easy to	1036	Another thing you have to watch out for is that it is quite easy to
1033	receive "spurious" readyness notifications, that is your callback might	1037	receive "spurious" readiness notifications, that is your callback might
1034	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1038	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1035	because there is no data. Not only are some backends known to create a	1039	because there is no data. Not only are some backends known to create a
1036	lot of those (for example solaris ports), it is very easy to get into	1040	lot of those (for example solaris ports), it is very easy to get into
1037	this situation even with a relatively standard program structure. Thus	1041	this situation even with a relatively standard program structure. Thus
1038	it is best to always use non-blocking I/O: An extra C<read>(2) returning	1042	it is best to always use non-blocking I/O: An extra C<read>(2) returning
…		…
1147		1151
1148	Timer watchers are simple relative timers that generate an event after a	1152	Timer watchers are simple relative timers that generate an event after a
1149	given time, and optionally repeating in regular intervals after that.	1153	given time, and optionally repeating in regular intervals after that.
1150		1154
1151	The timers are based on real time, that is, if you register an event that	1155	The timers are based on real time, that is, if you register an event that
1152	times out after an hour and you reset your system clock to last years	1156	times out after an hour and you reset your system clock to january last
1153	time, it will still time out after (roughly) and hour. "Roughly" because	1157	year, it will still time out after (roughly) and hour. "Roughly" because
1154	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1158	detecting time jumps is hard, and some inaccuracies are unavoidable (the
1155	monotonic clock option helps a lot here).	1159	monotonic clock option helps a lot here).
1156		1160
1157	The relative timeouts are calculated relative to the C<ev_now ()>	1161	The relative timeouts are calculated relative to the C<ev_now ()>
1158	time. This is usually the right thing as this timestamp refers to the time	1162	time. This is usually the right thing as this timestamp refers to the time
…		…
1160	you suspect event processing to be delayed and you I<need> to base the timeout	1164	you suspect event processing to be delayed and you I<need> to base the timeout
1161	on the current time, use something like this to adjust for this:	1165	on the current time, use something like this to adjust for this:
1162		1166
1163	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1167	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1164		1168
1165	The callback is guarenteed to be invoked only when its timeout has passed,	1169	The callback is guarenteed to be invoked only after its timeout has passed,
1166	but if multiple timers become ready during the same loop iteration then	1170	but if multiple timers become ready during the same loop iteration then
1167	order of execution is undefined.	1171	order of execution is undefined.
1168		1172
1169	=head3 Watcher-Specific Functions and Data Members	1173	=head3 Watcher-Specific Functions and Data Members
1170		1174
…		…
1172		1176
1173	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1177	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1174		1178
1175	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1179	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1176		1180
1177	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1181	Configure the timer to trigger after C<after> seconds. If C<repeat>
1178	C<0.>, then it will automatically be stopped. If it is positive, then the	1182	is C<0.>, then it will automatically be stopped once the timeout is
1179	timer will automatically be configured to trigger again C<repeat> seconds	1183	reached. If it is positive, then the timer will automatically be
1180	later, again, and again, until stopped manually.	1184	configured to trigger again C<repeat> seconds later, again, and again,
		1185	until stopped manually.
1181		1186
1182	The timer itself will do a best-effort at avoiding drift, that is, if you	1187	The timer itself will do a best-effort at avoiding drift, that is, if
1183	configure a timer to trigger every 10 seconds, then it will trigger at	1188	you configure a timer to trigger every 10 seconds, then it will normally
1184	exactly 10 second intervals. If, however, your program cannot keep up with	1189	trigger at exactly 10 second intervals. If, however, your program cannot
1185	the timer (because it takes longer than those 10 seconds to do stuff) the	1190	keep up with the timer (because it takes longer than those 10 seconds to
1186	timer will not fire more than once per event loop iteration.	1191	do stuff) the timer will not fire more than once per event loop iteration.
1187		1192
1188	=item ev_timer_again (loop, ev_timer *)	1193	=item ev_timer_again (loop, ev_timer *)
1189		1194
1190	This will act as if the timer timed out and restart it again if it is	1195	This will act as if the timer timed out and restart it again if it is
1191	repeating. The exact semantics are:	1196	repeating. The exact semantics are:
…		…
1268	Periodic watchers are also timers of a kind, but they are very versatile	1273	Periodic watchers are also timers of a kind, but they are very versatile
1269	(and unfortunately a bit complex).	1274	(and unfortunately a bit complex).
1270		1275
1271	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1276	Unlike C<ev_timer>'s, they are not based on real time (or relative time)
1272	but on wallclock time (absolute time). You can tell a periodic watcher	1277	but on wallclock time (absolute time). You can tell a periodic watcher
1273	to trigger "at" some specific point in time. For example, if you tell a	1278	to trigger after some specific point in time. For example, if you tell a
1274	periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()	1279	periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()
1275	+ 10.>) and then reset your system clock to the last year, then it will	1280	+ 10.>, that is, an absolute time not a delay) and then reset your system
		1281	clock to january of the previous year, then it will take more than year
1276	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	1282	to trigger the event (unlike an C<ev_timer>, which would still trigger
1277	roughly 10 seconds later).	1283	roughly 10 seconds later as it uses a relative timeout).
1278		1284
1279	They can also be used to implement vastly more complex timers, such as	1285	C<ev_periodic>s can also be used to implement vastly more complex timers,
1280	triggering an event on each midnight, local time or other, complicated,	1286	such as triggering an event on each "midnight, local time", or other
1281	rules.	1287	complicated, rules.
1282		1288
1283	As with timers, the callback is guarenteed to be invoked only when the	1289	As with timers, the callback is guarenteed to be invoked only when the
1284	time (C<at>) has been passed, but if multiple periodic timers become ready	1290	time (C<at>) has passed, but if multiple periodic timers become ready
1285	during the same loop iteration then order of execution is undefined.	1291	during the same loop iteration then order of execution is undefined.
1286		1292
1287	=head3 Watcher-Specific Functions and Data Members	1293	=head3 Watcher-Specific Functions and Data Members
1288		1294
1289	=over 4	1295	=over 4
…		…
1297		1303
1298	=over 4	1304	=over 4
1299		1305
1300	=item * absolute timer (at = time, interval = reschedule_cb = 0)	1306	=item * absolute timer (at = time, interval = reschedule_cb = 0)
1301		1307
1302	In this configuration the watcher triggers an event at the wallclock time	1308	In this configuration the watcher triggers an event after the wallclock
1303	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1309	time C<at> has passed and doesn't repeat. It will not adjust when a time
1304	that is, if it is to be run at January 1st 2011 then it will run when the	1310	jump occurs, that is, if it is to be run at January 1st 2011 then it will
1305	system time reaches or surpasses this time.	1311	run when the system time reaches or surpasses this time.
1306		1312
1307	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)	1313	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1308		1314
1309	In this mode the watcher will always be scheduled to time out at the next	1315	In this mode the watcher will always be scheduled to time out at the next
1310	C<at + N * interval> time (for some integer N, which can also be negative)	1316	C<at + N * interval> time (for some integer N, which can also be negative)
1311	and then repeat, regardless of any time jumps.	1317	and then repeat, regardless of any time jumps.
1312		1318
1313	This can be used to create timers that do not drift with respect to system	1319	This can be used to create timers that do not drift with respect to system
1314	time:	1320	time, for example, here is a C<ev_periodic> that triggers each hour, on
		1321	the hour:
1315		1322
1316	ev_periodic_set (&periodic, 0., 3600., 0);	1323	ev_periodic_set (&periodic, 0., 3600., 0);
1317		1324
1318	This doesn't mean there will always be 3600 seconds in between triggers,	1325	This doesn't mean there will always be 3600 seconds in between triggers,
1319	but only that the the callback will be called when the system time shows a	1326	but only that the the callback will be called when the system time shows a
…		…
1324	C<ev_periodic> will try to run the callback in this mode at the next possible	1331	C<ev_periodic> will try to run the callback in this mode at the next possible
1325	time where C<time = at (mod interval)>, regardless of any time jumps.	1332	time where C<time = at (mod interval)>, regardless of any time jumps.
1326		1333
1327	For numerical stability it is preferable that the C<at> value is near	1334	For numerical stability it is preferable that the C<at> value is near
1328	C<ev_now ()> (the current time), but there is no range requirement for	1335	C<ev_now ()> (the current time), but there is no range requirement for
1329	this value.	1336	this value, and in fact is often specified as zero.
		1337
		1338	Note also that there is an upper limit to how often a timer can fire (cpu
		1339	speed for example), so if C<interval> is very small then timing stability
		1340	will of course detoriate. Libev itself tries to be exact to be about one
		1341	millisecond (if the OS supports it and the machine is fast enough).
1330		1342
1331	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)	1343	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1332		1344
1333	In this mode the values for C<interval> and C<at> are both being	1345	In this mode the values for C<interval> and C<at> are both being
1334	ignored. Instead, each time the periodic watcher gets scheduled, the	1346	ignored. Instead, each time the periodic watcher gets scheduled, the
1335	reschedule callback will be called with the watcher as first, and the	1347	reschedule callback will be called with the watcher as first, and the
1336	current time as second argument.	1348	current time as second argument.
1337		1349
1338	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	1350	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
1339	ever, or make any event loop modifications>. If you need to stop it,	1351	ever, or make ANY event loop modifications whatsoever>.
1340	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
1341	starting an C<ev_prepare> watcher, which is legal).
1342		1352
		1353	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		1354	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		1355	only event loop modification you are allowed to do).
		1356
1343	Its prototype is C<ev_tstamp (reschedule_cb)(struct ev_periodic w,	1357	The callback prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic
1344	ev_tstamp now)>, e.g.:	1358	*w, ev_tstamp now)>, e.g.:
1345		1359
1346	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	1360	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
1347	{	1361	{
1348	return now + 60.;	1362	return now + 60.;
1349	}	1363	}
…		…
1351	It must return the next time to trigger, based on the passed time value	1365	It must return the next time to trigger, based on the passed time value
1352	(that is, the lowest time value larger than to the second argument). It	1366	(that is, the lowest time value larger than to the second argument). It
1353	will usually be called just before the callback will be triggered, but	1367	will usually be called just before the callback will be triggered, but
1354	might be called at other times, too.	1368	might be called at other times, too.
1355		1369
1356	NOTE: I<< This callback must always return a time that is later than the	1370	NOTE: I<< This callback must always return a time that is higher than or
1357	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	1371	equal to the passed C<now> value >>.
1358		1372
1359	This can be used to create very complex timers, such as a timer that	1373	This can be used to create very complex timers, such as a timer that
1360	triggers on each midnight, local time. To do this, you would calculate the	1374	triggers on "next midnight, local time". To do this, you would calculate the
1361	next midnight after C<now> and return the timestamp value for this. How	1375	next midnight after C<now> and return the timestamp value for this. How
1362	you do this is, again, up to you (but it is not trivial, which is the main	1376	you do this is, again, up to you (but it is not trivial, which is the main
1363	reason I omitted it as an example).	1377	reason I omitted it as an example).
1364		1378
1365	=back	1379	=back
…		…
1369	Simply stops and restarts the periodic watcher again. This is only useful	1383	Simply stops and restarts the periodic watcher again. This is only useful
1370	when you changed some parameters or the reschedule callback would return	1384	when you changed some parameters or the reschedule callback would return
1371	a different time than the last time it was called (e.g. in a crond like	1385	a different time than the last time it was called (e.g. in a crond like
1372	program when the crontabs have changed).	1386	program when the crontabs have changed).
1373		1387
		1388	=item ev_tstamp ev_periodic_at (ev_periodic *)
		1389
		1390	When active, returns the absolute time that the watcher is supposed to
		1391	trigger next.
		1392
1374	=item ev_tstamp offset [read-write]	1393	=item ev_tstamp offset [read-write]
1375		1394
1376	When repeating, this contains the offset value, otherwise this is the	1395	When repeating, this contains the offset value, otherwise this is the
1377	absolute point in time (the C<at> value passed to C<ev_periodic_set>).	1396	absolute point in time (the C<at> value passed to C<ev_periodic_set>).
1378		1397
…		…
1388	=item ev_tstamp (reschedule_cb)(struct ev_periodic w, ev_tstamp now) [read-write]	1407	=item ev_tstamp (reschedule_cb)(struct ev_periodic w, ev_tstamp now) [read-write]
1389		1408
1390	The current reschedule callback, or C<0>, if this functionality is	1409	The current reschedule callback, or C<0>, if this functionality is
1391	switched off. Can be changed any time, but changes only take effect when	1410	switched off. Can be changed any time, but changes only take effect when
1392	the periodic timer fires or C<ev_periodic_again> is being called.	1411	the periodic timer fires or C<ev_periodic_again> is being called.
1393
1394	=item ev_tstamp at [read-only]
1395
1396	When active, contains the absolute time that the watcher is supposed to
1397	trigger next.
1398		1412
1399	=back	1413	=back
1400		1414
1401	=head3 Examples	1415	=head3 Examples
1402		1416
…		…
1606	as even with OS-supported change notifications, this can be	1620	as even with OS-supported change notifications, this can be
1607	resource-intensive.	1621	resource-intensive.
1608		1622
1609	At the time of this writing, only the Linux inotify interface is	1623	At the time of this writing, only the Linux inotify interface is
1610	implemented (implementing kqueue support is left as an exercise for the	1624	implemented (implementing kqueue support is left as an exercise for the
		1625	reader, note, however, that the author sees no way of implementing ev_stat
1611	reader). Inotify will be used to give hints only and should not change the	1626	semantics with kqueue). Inotify will be used to give hints only and should
1612	semantics of C<ev_stat> watchers, which means that libev sometimes needs	1627	not change the semantics of C<ev_stat> watchers, which means that libev
1613	to fall back to regular polling again even with inotify, but changes are	1628	sometimes needs to fall back to regular polling again even with inotify,
1614	usually detected immediately, and if the file exists there will be no	1629	but changes are usually detected immediately, and if the file exists there
1615	polling.	1630	will be no polling.
1616		1631
1617	=head3 ABI Issues (Largefile Support)	1632	=head3 ABI Issues (Largefile Support)
1618		1633
1619	Libev by default (unless the user overrides this) uses the default	1634	Libev by default (unless the user overrides this) uses the default
1620	compilation environment, which means that on systems with optionally	1635	compilation environment, which means that on systems with optionally
…		…
1630	When C<inotify (7)> support has been compiled into libev (generally only	1645	When C<inotify (7)> support has been compiled into libev (generally only
1631	available on Linux) and present at runtime, it will be used to speed up	1646	available on Linux) and present at runtime, it will be used to speed up
1632	change detection where possible. The inotify descriptor will be created lazily	1647	change detection where possible. The inotify descriptor will be created lazily
1633	when the first C<ev_stat> watcher is being started.	1648	when the first C<ev_stat> watcher is being started.
1634		1649
1635	Inotify presense does not change the semantics of C<ev_stat> watchers	1650	Inotify presence does not change the semantics of C<ev_stat> watchers
1636	except that changes might be detected earlier, and in some cases, to avoid	1651	except that changes might be detected earlier, and in some cases, to avoid
1637	making regular C<stat> calls. Even in the presense of inotify support	1652	making regular C<stat> calls. Even in the presence of inotify support
1638	there are many cases where libev has to resort to regular C<stat> polling.	1653	there are many cases where libev has to resort to regular C<stat> polling.
1639		1654
1640	(There is no support for kqueue, as apparently it cannot be used to	1655	(There is no support for kqueue, as apparently it cannot be used to
1641	implement this functionality, due to the requirement of having a file	1656	implement this functionality, due to the requirement of having a file
1642	descriptor open on the object at all times).	1657	descriptor open on the object at all times).
…		…
1645		1660
1646	The C<stat ()> syscall only supports full-second resolution portably, and	1661	The C<stat ()> syscall only supports full-second resolution portably, and
1647	even on systems where the resolution is higher, many filesystems still	1662	even on systems where the resolution is higher, many filesystems still
1648	only support whole seconds.	1663	only support whole seconds.
1649		1664
1650	That means that, if the time is the only thing that changes, you might	1665	That means that, if the time is the only thing that changes, you can
1651	miss updates: on the first update, C<ev_stat> detects a change and calls	1666	easily miss updates: on the first update, C<ev_stat> detects a change and
1652	your callback, which does something. When there is another update within	1667	calls your callback, which does something. When there is another update
1653	the same second, C<ev_stat> will be unable to detect it.	1668	within the same second, C<ev_stat> will be unable to detect it as the stat
		1669	data does not change.
1654		1670
1655	The solution to this is to delay acting on a change for a second (or till	1671	The solution to this is to delay acting on a change for slightly more
1656	the next second boundary), using a roughly one-second delay C<ev_timer>	1672	than a second (or till slightly after the next full second boundary), using
1657	(C<ev_timer_set (w, 0., 1.01); ev_timer_again (loop, w)>). The C<.01>	1673	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
1658	is added to work around small timing inconsistencies of some operating	1674	ev_timer_again (loop, w)>).
1659	systems.	1675
		1676	The C<.02> offset is added to work around small timing inconsistencies
		1677	of some operating systems (where the second counter of the current time
		1678	might be be delayed. One such system is the Linux kernel, where a call to
		1679	C<gettimeofday> might return a timestamp with a full second later than
		1680	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		1681	update file times then there will be a small window where the kernel uses
		1682	the previous second to update file times but libev might already execute
		1683	the timer callback).
1660		1684
1661	=head3 Watcher-Specific Functions and Data Members	1685	=head3 Watcher-Specific Functions and Data Members
1662		1686
1663	=over 4	1687	=over 4
1664		1688
…		…
1670	C<path>. The C<interval> is a hint on how quickly a change is expected to	1694	C<path>. The C<interval> is a hint on how quickly a change is expected to
1671	be detected and should normally be specified as C<0> to let libev choose	1695	be detected and should normally be specified as C<0> to let libev choose
1672	a suitable value. The memory pointed to by C<path> must point to the same	1696	a suitable value. The memory pointed to by C<path> must point to the same
1673	path for as long as the watcher is active.	1697	path for as long as the watcher is active.
1674		1698
1675	The callback will be receive C<EV_STAT> when a change was detected,	1699	The callback will receive C<EV_STAT> when a change was detected, relative
1676	relative to the attributes at the time the watcher was started (or the	1700	to the attributes at the time the watcher was started (or the last change
1677	last change was detected).	1701	was detected).
1678		1702
1679	=item ev_stat_stat (loop, ev_stat *)	1703	=item ev_stat_stat (loop, ev_stat *)
1680		1704
1681	Updates the stat buffer immediately with new values. If you change the	1705	Updates the stat buffer immediately with new values. If you change the
1682	watched path in your callback, you could call this fucntion to avoid	1706	watched path in your callback, you could call this function to avoid
1683	detecting this change (while introducing a race condition). Can also be	1707	detecting this change (while introducing a race condition if you are not
1684	useful simply to find out the new values.	1708	the only one changing the path). Can also be useful simply to find out the
		1709	new values.
1685		1710
1686	=item ev_statdata attr [read-only]	1711	=item ev_statdata attr [read-only]
1687		1712
1688	The most-recently detected attributes of the file. Although the type is of	1713	The most-recently detected attributes of the file. Although the type is
1689	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types	1714	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
1690	suitable for your system. If the C<st_nlink> member is C<0>, then there	1715	suitable for your system, but you can only rely on the POSIX-standardised
		1716	members to be present. If the C<st_nlink> member is C<0>, then there was
1691	was some error while C<stat>ing the file.	1717	some error while C<stat>ing the file.
1692		1718
1693	=item ev_statdata prev [read-only]	1719	=item ev_statdata prev [read-only]
1694		1720
1695	The previous attributes of the file. The callback gets invoked whenever	1721	The previous attributes of the file. The callback gets invoked whenever
1696	C<prev> != C<attr>.	1722	C<prev> != C<attr>, or, more precisely, one or more of these members
		1723	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		1724	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
1697		1725
1698	=item ev_tstamp interval [read-only]	1726	=item ev_tstamp interval [read-only]
1699		1727
1700	The specified interval.	1728	The specified interval.
1701		1729
…		…
1755	}	1783	}
1756		1784
1757	...	1785	...
1758	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);	1786	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
1759	ev_stat_start (loop, &passwd);	1787	ev_stat_start (loop, &passwd);
1760	ev_timer_init (&timer, timer_cb, 0., 1.01);	1788	ev_timer_init (&timer, timer_cb, 0., 1.02);
1761		1789
1762		1790
1763	=head2 C<ev_idle> - when you've got nothing better to do...	1791	=head2 C<ev_idle> - when you've got nothing better to do...
1764		1792
1765	Idle watchers trigger events when no other events of the same or higher	1793	Idle watchers trigger events when no other events of the same or higher
…		…
1853		1881
1854	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)	1882	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1855	priority, to ensure that they are being run before any other watchers	1883	priority, to ensure that they are being run before any other watchers
1856	after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,	1884	after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,
1857	too) should not activate ("feed") events into libev. While libev fully	1885	too) should not activate ("feed") events into libev. While libev fully
1858	supports this, they will be called before other C<ev_check> watchers	1886	supports this, they might get executed before other C<ev_check> watchers
1859	did their job. As C<ev_check> watchers are often used to embed other	1887	did their job. As C<ev_check> watchers are often used to embed other
1860	(non-libev) event loops those other event loops might be in an unusable	1888	(non-libev) event loops those other event loops might be in an unusable
1861	state until their C<ev_check> watcher ran (always remind yourself to	1889	state until their C<ev_check> watcher ran (always remind yourself to
1862	coexist peacefully with others).	1890	coexist peacefully with others).
1863		1891
…		…
1878	=head3 Examples	1906	=head3 Examples
1879		1907
1880	There are a number of principal ways to embed other event loops or modules	1908	There are a number of principal ways to embed other event loops or modules
1881	into libev. Here are some ideas on how to include libadns into libev	1909	into libev. Here are some ideas on how to include libadns into libev
1882	(there is a Perl module named C<EV::ADNS> that does this, which you could	1910	(there is a Perl module named C<EV::ADNS> that does this, which you could
1883	use for an actually working example. Another Perl module named C<EV::Glib>	1911	use as a working example. Another Perl module named C<EV::Glib> embeds a
1884	embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV	1912	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
1885	into the Glib event loop).	1913	Glib event loop).
1886		1914
1887	Method 1: Add IO watchers and a timeout watcher in a prepare handler,	1915	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1888	and in a check watcher, destroy them and call into libadns. What follows	1916	and in a check watcher, destroy them and call into libadns. What follows
1889	is pseudo-code only of course. This requires you to either use a low	1917	is pseudo-code only of course. This requires you to either use a low
1890	priority for the check watcher or use C<ev_clear_pending> explicitly, as	1918	priority for the check watcher or use C<ev_clear_pending> explicitly, as
…		…
2280		2308
2281	This call incurs the overhead of a syscall only once per loop iteration,	2309	This call incurs the overhead of a syscall only once per loop iteration,
2282	so while the overhead might be noticable, it doesn't apply to repeated	2310	so while the overhead might be noticable, it doesn't apply to repeated
2283	calls to C<ev_async_send>.	2311	calls to C<ev_async_send>.
2284		2312
		2313	=item bool = ev_async_pending (ev_async *)
		2314
		2315	Returns a non-zero value when C<ev_async_send> has been called on the
		2316	watcher but the event has not yet been processed (or even noted) by the
		2317	event loop.
		2318
		2319	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		2320	the loop iterates next and checks for the watcher to have become active,
		2321	it will reset the flag again. C<ev_async_pending> can be used to very
		2322	quickly check wether invoking the loop might be a good idea.
		2323
		2324	Not that this does I<not> check wether the watcher itself is pending, only
		2325	wether it has been requested to make this watcher pending.
		2326
2285	=back	2327	=back
2286		2328
2287		2329
2288	=head1 OTHER FUNCTIONS	2330	=head1 OTHER FUNCTIONS
2289		2331
…		…
2360		2402
2361	=item * Priorities are not currently supported. Initialising priorities	2403	=item * Priorities are not currently supported. Initialising priorities
2362	will fail and all watchers will have the same priority, even though there	2404	will fail and all watchers will have the same priority, even though there
2363	is an ev_pri field.	2405	is an ev_pri field.
2364		2406
		2407	=item * In libevent, the last base created gets the signals, in libev, the
		2408	first base created (== the default loop) gets the signals.
		2409
2365	=item * Other members are not supported.	2410	=item * Other members are not supported.
2366		2411
2367	=item * The libev emulation is I<not> ABI compatible to libevent, you need	2412	=item * The libev emulation is I<not> ABI compatible to libevent, you need
2368	to use the libev header file and library.	2413	to use the libev header file and library.
2369		2414
…		…
2611	=item C<EV_DEFAULT>, C<EV_DEFAULT_>	2656	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
2612		2657
2613	Similar to the other two macros, this gives you the value of the default	2658	Similar to the other two macros, this gives you the value of the default
2614	loop, if multiple loops are supported ("ev loop default").	2659	loop, if multiple loops are supported ("ev loop default").
2615		2660
		2661	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		2662
		2663	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		2664	default loop has been initialised (C<UC> == unchecked). Their behaviour
		2665	is undefined when the default loop has not been initialised by a previous
		2666	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		2667
		2668	It is often prudent to use C<EV_DEFAULT> when initialising the first
		2669	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		2670
2616	=back	2671	=back
2617		2672
2618	Example: Declare and initialise a check watcher, utilising the above	2673	Example: Declare and initialise a check watcher, utilising the above
2619	macros so it will work regardless of whether multiple loops are supported	2674	macros so it will work regardless of whether multiple loops are supported
2620	or not.	2675	or not.
…		…
2715		2770
2716	libev.m4	2771	libev.m4
2717		2772
2718	=head2 PREPROCESSOR SYMBOLS/MACROS	2773	=head2 PREPROCESSOR SYMBOLS/MACROS
2719		2774
2720	Libev can be configured via a variety of preprocessor symbols you have to define	2775	Libev can be configured via a variety of preprocessor symbols you have to
2721	before including any of its files. The default is not to build for multiplicity	2776	define before including any of its files. The default in the absense of
2722	and only include the select backend.	2777	autoconf is noted for every option.
2723		2778
2724	=over 4	2779	=over 4
2725		2780
2726	=item EV_STANDALONE	2781	=item EV_STANDALONE
2727		2782
…		…
2753	=item EV_USE_NANOSLEEP	2808	=item EV_USE_NANOSLEEP
2754		2809
2755	If defined to be C<1>, libev will assume that C<nanosleep ()> is available	2810	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
2756	and will use it for delays. Otherwise it will use C<select ()>.	2811	and will use it for delays. Otherwise it will use C<select ()>.
2757		2812
		2813	=item EV_USE_EVENTFD
		2814
		2815	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		2816	available and will probe for kernel support at runtime. This will improve
		2817	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		2818	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		2819	2.7 or newer, otherwise disabled.
		2820
2758	=item EV_USE_SELECT	2821	=item EV_USE_SELECT
2759		2822
2760	If undefined or defined to be C<1>, libev will compile in support for the	2823	If undefined or defined to be C<1>, libev will compile in support for the
2761	C<select>(2) backend. No attempt at autodetection will be done: if no	2824	C<select>(2) backend. No attempt at autodetection will be done: if no
2762	other method takes over, select will be it. Otherwise the select backend	2825	other method takes over, select will be it. Otherwise the select backend
…		…
2798		2861
2799	=item EV_USE_EPOLL	2862	=item EV_USE_EPOLL
2800		2863
2801	If defined to be C<1>, libev will compile in support for the Linux	2864	If defined to be C<1>, libev will compile in support for the Linux
2802	C<epoll>(7) backend. Its availability will be detected at runtime,	2865	C<epoll>(7) backend. Its availability will be detected at runtime,
2803	otherwise another method will be used as fallback. This is the	2866	otherwise another method will be used as fallback. This is the preferred
2804	preferred backend for GNU/Linux systems.	2867	backend for GNU/Linux systems. If undefined, it will be enabled if the
		2868	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
2805		2869
2806	=item EV_USE_KQUEUE	2870	=item EV_USE_KQUEUE
2807		2871
2808	If defined to be C<1>, libev will compile in support for the BSD style	2872	If defined to be C<1>, libev will compile in support for the BSD style
2809	C<kqueue>(2) backend. Its actual availability will be detected at runtime,	2873	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
…		…
2828		2892
2829	=item EV_USE_INOTIFY	2893	=item EV_USE_INOTIFY
2830		2894
2831	If defined to be C<1>, libev will compile in support for the Linux inotify	2895	If defined to be C<1>, libev will compile in support for the Linux inotify
2832	interface to speed up C<ev_stat> watchers. Its actual availability will	2896	interface to speed up C<ev_stat> watchers. Its actual availability will
2833	be detected at runtime.	2897	be detected at runtime. If undefined, it will be enabled if the headers
		2898	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
2834		2899
2835	=item EV_ATOMIC_T	2900	=item EV_ATOMIC_T
2836		2901
2837	Libev requires an integer type (suitable for storing C<0> or C<1>) whose	2902	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
2838	access is atomic with respect to other threads or signal contexts. No such	2903	access is atomic with respect to other threads or signal contexts. No such
…		…
2925	defined to be C<0>, then they are not.	2990	defined to be C<0>, then they are not.
2926		2991
2927	=item EV_MINIMAL	2992	=item EV_MINIMAL
2928		2993
2929	If you need to shave off some kilobytes of code at the expense of some	2994	If you need to shave off some kilobytes of code at the expense of some
2930	speed, define this symbol to C<1>. Currently only used for gcc to override	2995	speed, define this symbol to C<1>. Currently this is used to override some
2931	some inlining decisions, saves roughly 30% codesize of amd64.	2996	inlining decisions, saves roughly 30% codesize of amd64. It also selects a
		2997	much smaller 2-heap for timer management over the default 4-heap.
2932		2998
2933	=item EV_PID_HASHSIZE	2999	=item EV_PID_HASHSIZE
2934		3000
2935	C<ev_child> watchers use a small hash table to distribute workload by	3001	C<ev_child> watchers use a small hash table to distribute workload by
2936	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more	3002	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
…		…
2942	C<ev_stat> watchers use a small hash table to distribute workload by	3008	C<ev_stat> watchers use a small hash table to distribute workload by
2943	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),	3009	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
2944	usually more than enough. If you need to manage thousands of C<ev_stat>	3010	usually more than enough. If you need to manage thousands of C<ev_stat>
2945	watchers you might want to increase this value (I<must> be a power of	3011	watchers you might want to increase this value (I<must> be a power of
2946	two).	3012	two).
		3013
		3014	=item EV_USE_4HEAP
		3015
		3016	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3017	timer and periodics heap, libev uses a 4-heap when this symbol is defined
		3018	to C<1>. The 4-heap uses more complicated (longer) code but has
		3019	noticably faster performance with many (thousands) of watchers.
		3020
		3021	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3022	(disabled).
		3023
		3024	=item EV_HEAP_CACHE_AT
		3025
		3026	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3027	timer and periodics heap, libev can cache the timestamp (I<at>) within
		3028	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3029	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3030	but avoids random read accesses on heap changes. This improves performance
		3031	noticably with with many (hundreds) of watchers.
		3032
		3033	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3034	(disabled).
2947		3035
2948	=item EV_COMMON	3036	=item EV_COMMON
2949		3037
2950	By default, all watchers have a C<void *data> member. By redefining	3038	By default, all watchers have a C<void *data> member. By redefining
2951	this macro to a something else you can include more and other types of	3039	this macro to a something else you can include more and other types of
…		…
3025		3113
3026	#include "ev_cpp.h"	3114	#include "ev_cpp.h"
3027	#include "ev.c"	3115	#include "ev.c"
3028		3116
3029		3117
		3118	=head1 THREADS AND COROUTINES
		3119
		3120	=head2 THREADS
		3121
		3122	Libev itself is completely threadsafe, but it uses no locking. This
		3123	means that you can use as many loops as you want in parallel, as long as
		3124	only one thread ever calls into one libev function with the same loop
		3125	parameter.
		3126
		3127	Or put differently: calls with different loop parameters can be done in
		3128	parallel from multiple threads, calls with the same loop parameter must be
		3129	done serially (but can be done from different threads, as long as only one
		3130	thread ever is inside a call at any point in time, e.g. by using a mutex
		3131	per loop).
		3132
		3133	If you want to know which design is best for your problem, then I cannot
		3134	help you but by giving some generic advice:
		3135
		3136	=over 4
		3137
		3138	=item * most applications have a main thread: use the default libev loop
		3139	in that thread, or create a seperate thread running only the default loop.
		3140
		3141	This helps integrating other libraries or software modules that use libev
		3142	themselves and don't care/know about threading.
		3143
		3144	=item * one loop per thread is usually a good model.
		3145
		3146	Doing this is almost never wrong, sometimes a better-performance model
		3147	exists, but it is always a good start.
		3148
		3149	=item * other models exist, such as the leader/follower pattern, where one
		3150	loop is handed through multiple threads in a kind of round-robbin fashion.
		3151
		3152	Chosing a model is hard - look around, learn, know that usually you cna do
		3153	better than you currently do :-)
		3154
		3155	=item * often you need to talk to some other thread which blocks in the
		3156	event loop - C<ev_async> watchers can be used to wake them up from other
		3157	threads safely (or from signal contexts...).
		3158
		3159	=back
		3160
		3161	=head2 COROUTINES
		3162
		3163	Libev is much more accomodating to coroutines ("cooperative threads"):
		3164	libev fully supports nesting calls to it's functions from different
		3165	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		3166	different coroutines and switch freely between both coroutines running the
		3167	loop, as long as you don't confuse yourself). The only exception is that
		3168	you must not do this from C<ev_periodic> reschedule callbacks.
		3169
		3170	Care has been invested into making sure that libev does not keep local
		3171	state inside C<ev_loop>, and other calls do not usually allow coroutine
		3172	switches.
		3173
		3174
3030	=head1 COMPLEXITIES	3175	=head1 COMPLEXITIES
3031		3176
3032	In this section the complexities of (many of) the algorithms used inside	3177	In this section the complexities of (many of) the algorithms used inside
3033	libev will be explained. For complexity discussions about backends see the	3178	libev will be explained. For complexity discussions about backends see the
3034	documentation for C<ev_default_init>.	3179	documentation for C<ev_default_init>.
…		…
3064	correct watcher to remove. The lists are usually short (you don't usually	3209	correct watcher to remove. The lists are usually short (you don't usually
3065	have many watchers waiting for the same fd or signal).	3210	have many watchers waiting for the same fd or signal).
3066		3211
3067	=item Finding the next timer in each loop iteration: O(1)	3212	=item Finding the next timer in each loop iteration: O(1)
3068		3213
3069	By virtue of using a binary heap, the next timer is always found at the	3214	By virtue of using a binary or 4-heap, the next timer is always found at a
3070	beginning of the storage array.	3215	fixed position in the storage array.
3071		3216
3072	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)	3217	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
3073		3218
3074	A change means an I/O watcher gets started or stopped, which requires	3219	A change means an I/O watcher gets started or stopped, which requires
3075	libev to recalculate its status (and possibly tell the kernel, depending	3220	libev to recalculate its status (and possibly tell the kernel, depending
…		…
3104	model. Libev still offers limited functionality on this platform in	3249	model. Libev still offers limited functionality on this platform in
3105	the form of the C<EVBACKEND_SELECT> backend, and only supports socket	3250	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
3106	descriptors. This only applies when using Win32 natively, not when using	3251	descriptors. This only applies when using Win32 natively, not when using
3107	e.g. cygwin.	3252	e.g. cygwin.
3108		3253
		3254	Lifting these limitations would basically require the full
		3255	re-implementation of the I/O system. If you are into these kinds of
		3256	things, then note that glib does exactly that for you in a very portable
		3257	way (note also that glib is the slowest event library known to man).
		3258
3109	There is no supported compilation method available on windows except	3259	There is no supported compilation method available on windows except
3110	embedding it into other applications.	3260	embedding it into other applications.
3111		3261
3112	Due to the many, low, and arbitrary limits on the win32 platform and the	3262	Due to the many, low, and arbitrary limits on the win32 platform and
3113	abysmal performance of winsockets, using a large number of sockets is not	3263	the abysmal performance of winsockets, using a large number of sockets
3114	recommended (and not reasonable). If your program needs to use more than	3264	is not recommended (and not reasonable). If your program needs to use
3115	a hundred or so sockets, then likely it needs to use a totally different	3265	more than a hundred or so sockets, then likely it needs to use a totally
3116	implementation for windows, as libev offers the POSIX model, which cannot	3266	different implementation for windows, as libev offers the POSIX readiness
3117	be implemented efficiently on windows (microsoft monopoly games).	3267	notification model, which cannot be implemented efficiently on windows
		3268	(microsoft monopoly games).
3118		3269
3119	=over 4	3270	=over 4
3120		3271
3121	=item The winsocket select function	3272	=item The winsocket select function
3122		3273
…		…
3136	Note that winsockets handling of fd sets is O(n), so you can easily get a	3287	Note that winsockets handling of fd sets is O(n), so you can easily get a
3137	complexity in the O(n²) range when using win32.	3288	complexity in the O(n²) range when using win32.
3138		3289
3139	=item Limited number of file descriptors	3290	=item Limited number of file descriptors
3140		3291
3141	Windows has numerous arbitrary (and low) limits on things. Early versions	3292	Windows has numerous arbitrary (and low) limits on things.
3142	of winsocket's select only supported waiting for a max. of C<64> handles	3293
		3294	Early versions of winsocket's select only supported waiting for a maximum
3143	(probably owning to the fact that all windows kernels can only wait for	3295	of C<64> handles (probably owning to the fact that all windows kernels
3144	C<64> things at the same time internally; microsoft recommends spawning a	3296	can only wait for C<64> things at the same time internally; microsoft
3145	chain of threads and wait for 63 handles and the previous thread in each).	3297	recommends spawning a chain of threads and wait for 63 handles and the
		3298	previous thread in each. Great).
3146		3299
3147	Newer versions support more handles, but you need to define C<FD_SETSIZE>	3300	Newer versions support more handles, but you need to define C<FD_SETSIZE>
3148	to some high number (e.g. C<2048>) before compiling the winsocket select	3301	to some high number (e.g. C<2048>) before compiling the winsocket select
3149	call (which might be in libev or elsewhere, for example, perl does its own	3302	call (which might be in libev or elsewhere, for example, perl does its own
3150	select emulation on windows).	3303	select emulation on windows).
…		…
3162	calling select (O(n²)) will likely make this unworkable.	3315	calling select (O(n²)) will likely make this unworkable.
3163		3316
3164	=back	3317	=back
3165		3318
3166		3319
		3320	=head1 PORTABILITY REQUIREMENTS
		3321
		3322	In addition to a working ISO-C implementation, libev relies on a few
		3323	additional extensions:
		3324
		3325	=over 4
		3326
		3327	=item C<sig_atomic_t volatile> must be thread-atomic as well
		3328
		3329	The type C<sig_atomic_t volatile> (or whatever is defined as
		3330	C<EV_ATOMIC_T>) must be atomic w.r.t. accesses from different
		3331	threads. This is not part of the specification for C<sig_atomic_t>, but is
		3332	believed to be sufficiently portable.
		3333
		3334	=item C<sigprocmask> must work in a threaded environment
		3335
		3336	Libev uses C<sigprocmask> to temporarily block signals. This is not
		3337	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		3338	pthread implementations will either allow C<sigprocmask> in the "main
		3339	thread" or will block signals process-wide, both behaviours would
		3340	be compatible with libev. Interaction between C<sigprocmask> and
		3341	C<pthread_sigmask> could complicate things, however.
		3342
		3343	The most portable way to handle signals is to block signals in all threads
		3344	except the initial one, and run the default loop in the initial thread as
		3345	well.
		3346
		3347	=item C<long> must be large enough for common memory allocation sizes
		3348
		3349	To improve portability and simplify using libev, libev uses C<long>
		3350	internally instead of C<size_t> when allocating its data structures. On
		3351	non-POSIX systems (Microsoft...) this might be unexpectedly low, but
		3352	is still at least 31 bits everywhere, which is enough for hundreds of
		3353	millions of watchers.
		3354
		3355	=item C<double> must hold a time value in seconds with enough accuracy
		3356
		3357	The type C<double> is used to represent timestamps. It is required to
		3358	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		3359	enough for at least into the year 4000. This requirement is fulfilled by
		3360	implementations implementing IEEE 754 (basically all existing ones).
		3361
		3362	=back
		3363
		3364	If you know of other additional requirements drop me a note.
		3365
		3366
		3367	=head1 VALGRIND
		3368
		3369	Valgrind has a special section here because it is a popular tool that is
		3370	highly useful, but valgrind reports are very hard to interpret.
		3371
		3372	If you think you found a bug (memory leak, uninitialised data access etc.)
		3373	in libev, then check twice: If valgrind reports something like:
		3374
		3375	==2274== definitely lost: 0 bytes in 0 blocks.
		3376	==2274== possibly lost: 0 bytes in 0 blocks.
		3377	==2274== still reachable: 256 bytes in 1 blocks.
		3378
		3379	then there is no memory leak. Similarly, under some circumstances,
		3380	valgrind might report kernel bugs as if it were a bug in libev, or it
		3381	might be confused (it is a very good tool, but only a tool).
		3382
		3383	If you are unsure about something, feel free to contact the mailing list
		3384	with the full valgrind report and an explanation on why you think this is
		3385	a bug in libev. However, don't be annoyed when you get a brisk "this is
		3386	no bug" answer and take the chance of learning how to interpret valgrind
		3387	properly.
		3388
		3389	If you need, for some reason, empty reports from valgrind for your project
		3390	I suggest using suppression lists.
		3391
		3392
3167	=head1 AUTHOR	3393	=head1 AUTHOR
3168		3394
3169	Marc Lehmann <libev@schmorp.de>.	3395	Marc Lehmann <libev@schmorp.de>.
3170		3396

Diff Legend

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

Comparing libev/ev.pod (file contents): Revision 1.138 by root, Mon Mar 31 01:14:12 2008 UTC vs. Revision 1.158 by root, Wed May 21 12:51:38 2008 UTC

Diff Legend

Comparing libev/ev.pod (file contents):
Revision 1.138 by root, Mon Mar 31 01:14:12 2008 UTC vs.
Revision 1.158 by root, Wed May 21 12:51:38 2008 UTC