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

Comparing libev/ev.pod (file contents):
Revision 1.184 by root, Tue Sep 23 09:11:14 2008 UTC vs.
Revision 1.197 by root, Tue Oct 21 20:52:30 2008 UTC

…		…
214	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for	214	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
215	recommended ones.	215	recommended ones.
216		216
217	See the description of C<ev_embed> watchers for more info.	217	See the description of C<ev_embed> watchers for more info.
218		218
219	=item ev_set_allocator (void (cb)(void *ptr, long size))	219	=item ev_set_allocator (void (cb)(void *ptr, long size)) [NOT REENTRANT]
220		220
221	Sets the allocation function to use (the prototype is similar - the	221	Sets the allocation function to use (the prototype is similar - the
222	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is	222	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
223	used to allocate and free memory (no surprises here). If it returns zero	223	used to allocate and free memory (no surprises here). If it returns zero
224	when memory needs to be allocated (C<size != 0>), the library might abort	224	when memory needs to be allocated (C<size != 0>), the library might abort
…		…
250	}	250	}
251		251
252	...	252	...
253	ev_set_allocator (persistent_realloc);	253	ev_set_allocator (persistent_realloc);
254		254
255	=item ev_set_syserr_cb (void (cb)(const char msg));	255	=item ev_set_syserr_cb (void (cb)(const char msg)); [NOT REENTRANT]
256		256
257	Set the callback function to call on a retryable system call error (such	257	Set the callback function to call on a retryable system call error (such
258	as failed select, poll, epoll_wait). The message is a printable string	258	as failed select, poll, epoll_wait). The message is a printable string
259	indicating the system call or subsystem causing the problem. If this	259	indicating the system call or subsystem causing the problem. If this
260	callback is set, then libev will expect it to remedy the situation, no	260	callback is set, then libev will expect it to remedy the situation, no
…		…
685	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	685	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
686	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	686	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
687		687
688	This "unloop state" will be cleared when entering C<ev_loop> again.	688	This "unloop state" will be cleared when entering C<ev_loop> again.
689		689
		690	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
		691
690	=item ev_ref (loop)	692	=item ev_ref (loop)
691		693
692	=item ev_unref (loop)	694	=item ev_unref (loop)
693		695
694	Ref/unref can be used to add or remove a reference count on the event	696	Ref/unref can be used to add or remove a reference count on the event
…		…
892	=item C<EV_ERROR>	894	=item C<EV_ERROR>
893		895
894	An unspecified error has occurred, the watcher has been stopped. This might	896	An unspecified error has occurred, the watcher has been stopped. This might
895	happen because the watcher could not be properly started because libev	897	happen because the watcher could not be properly started because libev
896	ran out of memory, a file descriptor was found to be closed or any other	898	ran out of memory, a file descriptor was found to be closed or any other
		899	problem. Libev considers these application bugs.
		900
897	problem. You best act on it by reporting the problem and somehow coping	901	You best act on it by reporting the problem and somehow coping with the
898	with the watcher being stopped.	902	watcher being stopped. Note that well-written programs should not receive
		903	an error ever, so when your watcher receives it, this usually indicates a
		904	bug in your program.
899		905
900	Libev will usually signal a few "dummy" events together with an error, for	906	Libev will usually signal a few "dummy" events together with an error, for
901	example it might indicate that a fd is readable or writable, and if your	907	example it might indicate that a fd is readable or writable, and if your
902	callbacks is well-written it can just attempt the operation and cope with	908	callbacks is well-written it can just attempt the operation and cope with
903	the error from read() or write(). This will not work in multi-threaded	909	the error from read() or write(). This will not work in multi-threaded
…		…
967		973
968	ev_io_start (EV_DEFAULT_UC, &w);	974	ev_io_start (EV_DEFAULT_UC, &w);
969		975
970	=item C<ev_TYPE_stop> (loop , ev_TYPE watcher)	976	=item C<ev_TYPE_stop> (loop , ev_TYPE watcher)
971		977
972	Stops the given watcher again (if active) and clears the pending	978	Stops the given watcher if active, and clears the pending status (whether
		979	the watcher was active or not).
		980
973	status. It is possible that stopped watchers are pending (for example,	981	It is possible that stopped watchers are pending - for example,
974	non-repeating timers are being stopped when they become pending), but	982	non-repeating timers are being stopped when they become pending - but
975	C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If	983	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
976	you want to free or reuse the memory used by the watcher it is therefore a	984	pending. If you want to free or reuse the memory used by the watcher it is
977	good idea to always call its C<ev_TYPE_stop> function.	985	therefore a good idea to always call its C<ev_TYPE_stop> function.
978		986
979	=item bool ev_is_active (ev_TYPE *watcher)	987	=item bool ev_is_active (ev_TYPE *watcher)
980		988
981	Returns a true value iff the watcher is active (i.e. it has been started	989	Returns a true value iff the watcher is active (i.e. it has been started
982	and not yet been stopped). As long as a watcher is active you must not modify	990	and not yet been stopped). As long as a watcher is active you must not modify
…		…
1625		1633
1626	=back	1634	=back
1627		1635
1628	=head3 Examples	1636	=head3 Examples
1629		1637
1630	Example: Try to exit cleanly on SIGINT and SIGTERM.	1638	Example: Try to exit cleanly on SIGINT.
1631		1639
1632	static void	1640	static void
1633	sigint_cb (struct ev_loop loop, struct ev_signal w, int revents)	1641	sigint_cb (struct ev_loop loop, struct ev_signal w, int revents)
1634	{	1642	{
1635	ev_unloop (loop, EVUNLOOP_ALL);	1643	ev_unloop (loop, EVUNLOOP_ALL);
1636	}	1644	}
1637		1645
1638	struct ev_signal signal_watcher;	1646	struct ev_signal signal_watcher;
1639	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);	1647	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1640	ev_signal_start (loop, &sigint_cb);	1648	ev_signal_start (loop, &signal_watcher);
1641		1649
1642		1650
1643	=head2 C<ev_child> - watch out for process status changes	1651	=head2 C<ev_child> - watch out for process status changes
1644		1652
1645	Child watchers trigger when your process receives a SIGCHLD in response to	1653	Child watchers trigger when your process receives a SIGCHLD in response to
…		…
1792	to exchange stat structures with application programs compiled using the	1800	to exchange stat structures with application programs compiled using the
1793	default compilation environment.	1801	default compilation environment.
1794		1802
1795	=head3 Inotify and Kqueue	1803	=head3 Inotify and Kqueue
1796		1804
1797	When C<inotify (7)> support has been compiled into libev (generally only	1805	When C<inotify (7)> support has been compiled into libev (generally
		1806	only available with Linux 2.6.25 or above due to bugs in earlier
1798	available with Linux) and present at runtime, it will be used to speed up	1807	implementations) and present at runtime, it will be used to speed up
1799	change detection where possible. The inotify descriptor will be created lazily	1808	change detection where possible. The inotify descriptor will be created
1800	when the first C<ev_stat> watcher is being started.	1809	lazily when the first C<ev_stat> watcher is being started.
1801		1810
1802	Inotify presence does not change the semantics of C<ev_stat> watchers	1811	Inotify presence does not change the semantics of C<ev_stat> watchers
1803	except that changes might be detected earlier, and in some cases, to avoid	1812	except that changes might be detected earlier, and in some cases, to avoid
1804	making regular C<stat> calls. Even in the presence of inotify support	1813	making regular C<stat> calls. Even in the presence of inotify support
1805	there are many cases where libev has to resort to regular C<stat> polling,	1814	there are many cases where libev has to resort to regular C<stat> polling,
…		…
2242	So when you want to use this feature you will always have to be prepared	2251	So when you want to use this feature you will always have to be prepared
2243	that you cannot get an embeddable loop. The recommended way to get around	2252	that you cannot get an embeddable loop. The recommended way to get around
2244	this is to have a separate variables for your embeddable loop, try to	2253	this is to have a separate variables for your embeddable loop, try to
2245	create it, and if that fails, use the normal loop for everything.	2254	create it, and if that fails, use the normal loop for everything.
2246		2255
		2256	=head3 C<ev_embed> and fork
		2257
		2258	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2259	automatically be applied to the embedded loop as well, so no special
		2260	fork handling is required in that case. When the watcher is not running,
		2261	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2262	as applicable.
		2263
2247	=head3 Watcher-Specific Functions and Data Members	2264	=head3 Watcher-Specific Functions and Data Members
2248		2265
2249	=over 4	2266	=over 4
2250		2267
2251	=item ev_embed_init (ev_embed , callback, struct ev_loop embedded_loop)	2268	=item ev_embed_init (ev_embed , callback, struct ev_loop embedded_loop)
…		…
2376	=over 4	2393	=over 4
2377		2394
2378	=item queueing from a signal handler context	2395	=item queueing from a signal handler context
2379		2396
2380	To implement race-free queueing, you simply add to the queue in the signal	2397	To implement race-free queueing, you simply add to the queue in the signal
2381	handler but you block the signal handler in the watcher callback. Here is an example that does that for	2398	handler but you block the signal handler in the watcher callback. Here is
2382	some fictitious SIGUSR1 handler:	2399	an example that does that for some fictitious SIGUSR1 handler:
2383		2400
2384	static ev_async mysig;	2401	static ev_async mysig;
2385		2402
2386	static void	2403	static void
2387	sigusr1_handler (void)	2404	sigusr1_handler (void)
…		…
2494	=over 4	2511	=over 4
2495		2512
2496	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	2513	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
2497		2514
2498	This function combines a simple timer and an I/O watcher, calls your	2515	This function combines a simple timer and an I/O watcher, calls your
2499	callback on whichever event happens first and automatically stop both	2516	callback on whichever event happens first and automatically stops both
2500	watchers. This is useful if you want to wait for a single event on an fd	2517	watchers. This is useful if you want to wait for a single event on an fd
2501	or timeout without having to allocate/configure/start/stop/free one or	2518	or timeout without having to allocate/configure/start/stop/free one or
2502	more watchers yourself.	2519	more watchers yourself.
2503		2520
2504	If C<fd> is less than 0, then no I/O watcher will be started and events	2521	If C<fd> is less than 0, then no I/O watcher will be started and the
2505	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	2522	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
2506	C<events> set will be created and started.	2523	the given C<fd> and C<events> set will be created and started.
2507		2524
2508	If C<timeout> is less than 0, then no timeout watcher will be	2525	If C<timeout> is less than 0, then no timeout watcher will be
2509	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	2526	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
2510	repeat = 0) will be started. While C<0> is a valid timeout, it is of	2527	repeat = 0) will be started. C<0> is a valid timeout.
2511	dubious value.
2512		2528
2513	The callback has the type C<void (cb)(int revents, void arg)> and gets	2529	The callback has the type C<void (cb)(int revents, void arg)> and gets
2514	passed an C<revents> set like normal event callbacks (a combination of	2530	passed an C<revents> set like normal event callbacks (a combination of
2515	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	2531	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
2516	value passed to C<ev_once>:	2532	value passed to C<ev_once>. Note that it is possible to receive I<both>
		2533	a timeout and an io event at the same time - you probably should give io
		2534	events precedence.
		2535
		2536	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
2517		2537
2518	static void stdin_ready (int revents, void *arg)	2538	static void stdin_ready (int revents, void *arg)
2519	{	2539	{
		2540	if (revents & EV_READ)
		2541	/* stdin might have data for us, joy! */;
2520	if (revents & EV_TIMEOUT)	2542	else if (revents & EV_TIMEOUT)
2521	/* doh, nothing entered */;	2543	/* doh, nothing entered */;
2522	else if (revents & EV_READ)
2523	/* stdin might have data for us, joy! */;
2524	}	2544	}
2525		2545
2526	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	2546	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
2527		2547
2528	=item ev_feed_event (ev_loop , watcher , int revents)	2548	=item ev_feed_event (ev_loop , watcher , int revents)
…		…
3241	definition and a statement, respectively. See the F<ev.h> header file for	3261	definition and a statement, respectively. See the F<ev.h> header file for
3242	their default definitions. One possible use for overriding these is to	3262	their default definitions. One possible use for overriding these is to
3243	avoid the C<struct ev_loop *> as first argument in all cases, or to use	3263	avoid the C<struct ev_loop *> as first argument in all cases, or to use
3244	method calls instead of plain function calls in C++.	3264	method calls instead of plain function calls in C++.
3245		3265
		3266	=back
		3267
3246	=head2 EXPORTED API SYMBOLS	3268	=head2 EXPORTED API SYMBOLS
3247		3269
3248	If you need to re-export the API (e.g. via a DLL) and you need a list of	3270	If you need to re-export the API (e.g. via a DLL) and you need a list of
3249	exported symbols, you can use the provided F<Symbol.*> files which list	3271	exported symbols, you can use the provided F<Symbol.*> files which list
3250	all public symbols, one per line:	3272	all public symbols, one per line:
…		…
3296	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	3318	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
3297		3319
3298	#include "ev_cpp.h"	3320	#include "ev_cpp.h"
3299	#include "ev.c"	3321	#include "ev.c"
3300		3322
		3323	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
3301		3324
3302	=head1 THREADS AND COROUTINES	3325	=head2 THREADS AND COROUTINES
3303		3326
3304	=head2 THREADS	3327	=head3 THREADS
3305		3328
3306	Libev itself is thread-safe (unless the opposite is specifically	3329	All libev functions are reentrant and thread-safe unless explicitly
3307	documented for a function), but it uses no locking itself. This means that	3330	documented otherwise, but libev implements no locking itself. This means
3308	you can use as many loops as you want in parallel, as long as only one	3331	that you can use as many loops as you want in parallel, as long as there
3309	thread ever calls into one libev function with the same loop parameter:	3332	are no concurrent calls into any libev function with the same loop
		3333	parameter (C<ev_default_*> calls have an implicit default loop parameter,
3310	libev guarantees that different event loops share no data structures that	3334	of course): libev guarantees that different event loops share no data
3311	need locking.	3335	structures that need any locking.
3312		3336
3313	Or to put it differently: calls with different loop parameters can be done	3337	Or to put it differently: calls with different loop parameters can be done
3314	concurrently from multiple threads, calls with the same loop parameter	3338	concurrently from multiple threads, calls with the same loop parameter
3315	must be done serially (but can be done from different threads, as long as	3339	must be done serially (but can be done from different threads, as long as
3316	only one thread ever is inside a call at any point in time, e.g. by using	3340	only one thread ever is inside a call at any point in time, e.g. by using
3317	a mutex per loop).	3341	a mutex per loop).
3318		3342
3319	Specifically to support threads (and signal handlers), libev implements	3343	Specifically to support threads (and signal handlers), libev implements
3320	so-called C<ev_async> watchers, which allow some limited form of	3344	so-called C<ev_async> watchers, which allow some limited form of
3321	concurrency on the same event loop.	3345	concurrency on the same event loop, namely waking it up "from the
		3346	outside".
3322		3347
3323	If you want to know which design (one loop, locking, or multiple loops	3348	If you want to know which design (one loop, locking, or multiple loops
3324	without or something else still) is best for your problem, then I cannot	3349	without or something else still) is best for your problem, then I cannot
3325	help you. I can give some generic advice however:	3350	help you, but here is some generic advice:
3326		3351
3327	=over 4	3352	=over 4
3328		3353
3329	=item * most applications have a main thread: use the default libev loop	3354	=item * most applications have a main thread: use the default libev loop
3330	in that thread, or create a separate thread running only the default loop.	3355	in that thread, or create a separate thread running only the default loop.
…		…
3354	default loop and triggering an C<ev_async> watcher from the default loop	3379	default loop and triggering an C<ev_async> watcher from the default loop
3355	watcher callback into the event loop interested in the signal.	3380	watcher callback into the event loop interested in the signal.
3356		3381
3357	=back	3382	=back
3358		3383
3359	=head2 COROUTINES	3384	=head3 COROUTINES
3360		3385
3361	Libev is much more accommodating to coroutines ("cooperative threads"):	3386	Libev is very accommodating to coroutines ("cooperative threads"):
3362	libev fully supports nesting calls to it's functions from different	3387	libev fully supports nesting calls to its functions from different
3363	coroutines (e.g. you can call C<ev_loop> on the same loop from two	3388	coroutines (e.g. you can call C<ev_loop> on the same loop from two
3364	different coroutines and switch freely between both coroutines running the	3389	different coroutines, and switch freely between both coroutines running the
3365	loop, as long as you don't confuse yourself). The only exception is that	3390	loop, as long as you don't confuse yourself). The only exception is that
3366	you must not do this from C<ev_periodic> reschedule callbacks.	3391	you must not do this from C<ev_periodic> reschedule callbacks.
3367		3392
3368	Care has been taken to ensure that libev does not keep local state inside	3393	Care has been taken to ensure that libev does not keep local state inside
3369	C<ev_loop>, and other calls do not usually allow coroutine switches.	3394	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		3395	they do not clal any callbacks.
3370		3396
		3397	=head2 COMPILER WARNINGS
3371		3398
3372	=head1 COMPLEXITIES	3399	Depending on your compiler and compiler settings, you might get no or a
		3400	lot of warnings when compiling libev code. Some people are apparently
		3401	scared by this.
3373		3402
3374	In this section the complexities of (many of) the algorithms used inside	3403	However, these are unavoidable for many reasons. For one, each compiler
3375	libev will be explained. For complexity discussions about backends see the	3404	has different warnings, and each user has different tastes regarding
3376	documentation for C<ev_default_init>.	3405	warning options. "Warn-free" code therefore cannot be a goal except when
		3406	targeting a specific compiler and compiler-version.
3377		3407
3378	All of the following are about amortised time: If an array needs to be	3408	Another reason is that some compiler warnings require elaborate
3379	extended, libev needs to realloc and move the whole array, but this	3409	workarounds, or other changes to the code that make it less clear and less
3380	happens asymptotically never with higher number of elements, so O(1) might	3410	maintainable.
3381	mean it might do a lengthy realloc operation in rare cases, but on average
3382	it is much faster and asymptotically approaches constant time.
3383		3411
3384	=over 4	3412	And of course, some compiler warnings are just plain stupid, or simply
		3413	wrong (because they don't actually warn about the condition their message
		3414	seems to warn about). For example, certain older gcc versions had some
		3415	warnings that resulted an extreme number of false positives. These have
		3416	been fixed, but some people still insist on making code warn-free with
		3417	such buggy versions.
3385		3418
3386	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)	3419	While libev is written to generate as few warnings as possible,
		3420	"warn-free" code is not a goal, and it is recommended not to build libev
		3421	with any compiler warnings enabled unless you are prepared to cope with
		3422	them (e.g. by ignoring them). Remember that warnings are just that:
		3423	warnings, not errors, or proof of bugs.
3387		3424
3388	This means that, when you have a watcher that triggers in one hour and
3389	there are 100 watchers that would trigger before that then inserting will
3390	have to skip roughly seven (C<ld 100>) of these watchers.
3391		3425
3392	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)	3426	=head2 VALGRIND
3393		3427
3394	That means that changing a timer costs less than removing/adding them	3428	Valgrind has a special section here because it is a popular tool that is
3395	as only the relative motion in the event queue has to be paid for.	3429	highly useful. Unfortunately, valgrind reports are very hard to interpret.
3396		3430
3397	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)	3431	If you think you found a bug (memory leak, uninitialised data access etc.)
		3432	in libev, then check twice: If valgrind reports something like:
3398		3433
3399	These just add the watcher into an array or at the head of a list.	3434	==2274== definitely lost: 0 bytes in 0 blocks.
		3435	==2274== possibly lost: 0 bytes in 0 blocks.
		3436	==2274== still reachable: 256 bytes in 1 blocks.
3400		3437
3401	=item Stopping check/prepare/idle/fork/async watchers: O(1)	3438	Then there is no memory leak, just as memory accounted to global variables
		3439	is not a memleak - the memory is still being refernced, and didn't leak.
3402		3440
3403	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))	3441	Similarly, under some circumstances, valgrind might report kernel bugs
		3442	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		3443	although an acceptable workaround has been found here), or it might be
		3444	confused.
3404		3445
3405	These watchers are stored in lists then need to be walked to find the	3446	Keep in mind that valgrind is a very good tool, but only a tool. Don't
3406	correct watcher to remove. The lists are usually short (you don't usually	3447	make it into some kind of religion.
3407	have many watchers waiting for the same fd or signal).
3408		3448
3409	=item Finding the next timer in each loop iteration: O(1)	3449	If you are unsure about something, feel free to contact the mailing list
		3450	with the full valgrind report and an explanation on why you think this
		3451	is a bug in libev (best check the archives, too :). However, don't be
		3452	annoyed when you get a brisk "this is no bug" answer and take the chance
		3453	of learning how to interpret valgrind properly.
3410		3454
3411	By virtue of using a binary or 4-heap, the next timer is always found at a	3455	If you need, for some reason, empty reports from valgrind for your project
3412	fixed position in the storage array.	3456	I suggest using suppression lists.
3413		3457
3414	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
3415		3458
3416	A change means an I/O watcher gets started or stopped, which requires	3459	=head1 PORTABILITY NOTES
3417	libev to recalculate its status (and possibly tell the kernel, depending
3418	on backend and whether C<ev_io_set> was used).
3419		3460
3420	=item Activating one watcher (putting it into the pending state): O(1)
3421
3422	=item Priority handling: O(number_of_priorities)
3423
3424	Priorities are implemented by allocating some space for each
3425	priority. When doing priority-based operations, libev usually has to
3426	linearly search all the priorities, but starting/stopping and activating
3427	watchers becomes O(1) with respect to priority handling.
3428
3429	=item Sending an ev_async: O(1)
3430
3431	=item Processing ev_async_send: O(number_of_async_watchers)
3432
3433	=item Processing signals: O(max_signal_number)
3434
3435	Sending involves a system call I<iff> there were no other C<ev_async_send>
3436	calls in the current loop iteration. Checking for async and signal events
3437	involves iterating over all running async watchers or all signal numbers.
3438
3439	=back
3440
3441
3442	=head1 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS	3461	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
3443		3462
3444	Win32 doesn't support any of the standards (e.g. POSIX) that libev	3463	Win32 doesn't support any of the standards (e.g. POSIX) that libev
3445	requires, and its I/O model is fundamentally incompatible with the POSIX	3464	requires, and its I/O model is fundamentally incompatible with the POSIX
3446	model. Libev still offers limited functionality on this platform in	3465	model. Libev still offers limited functionality on this platform in
3447	the form of the C<EVBACKEND_SELECT> backend, and only supports socket	3466	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
…		…
3534	wrap all I/O functions and provide your own fd management, but the cost of	3553	wrap all I/O functions and provide your own fd management, but the cost of
3535	calling select (O(n²)) will likely make this unworkable.	3554	calling select (O(n²)) will likely make this unworkable.
3536		3555
3537	=back	3556	=back
3538		3557
3539
3540	=head1 PORTABILITY REQUIREMENTS	3558	=head2 PORTABILITY REQUIREMENTS
3541		3559
3542	In addition to a working ISO-C implementation, libev relies on a few	3560	In addition to a working ISO-C implementation and of course the
3543	additional extensions:	3561	backend-specific APIs, libev relies on a few additional extensions:
3544		3562
3545	=over 4	3563	=over 4
3546		3564
3547	=item C<void ()(ev_watcher_type , int revents)> must have compatible	3565	=item C<void ()(ev_watcher_type , int revents)> must have compatible
3548	calling conventions regardless of C<ev_watcher_type *>.	3566	calling conventions regardless of C<ev_watcher_type *>.
…		…
3573	except the initial one, and run the default loop in the initial thread as	3591	except the initial one, and run the default loop in the initial thread as
3574	well.	3592	well.
3575		3593
3576	=item C<long> must be large enough for common memory allocation sizes	3594	=item C<long> must be large enough for common memory allocation sizes
3577		3595
3578	To improve portability and simplify using libev, libev uses C<long>	3596	To improve portability and simplify its API, libev uses C<long> internally
3579	internally instead of C<size_t> when allocating its data structures. On	3597	instead of C<size_t> when allocating its data structures. On non-POSIX
3580	non-POSIX systems (Microsoft...) this might be unexpectedly low, but	3598	systems (Microsoft...) this might be unexpectedly low, but is still at
3581	is still at least 31 bits everywhere, which is enough for hundreds of	3599	least 31 bits everywhere, which is enough for hundreds of millions of
3582	millions of watchers.	3600	watchers.
3583		3601
3584	=item C<double> must hold a time value in seconds with enough accuracy	3602	=item C<double> must hold a time value in seconds with enough accuracy
3585		3603
3586	The type C<double> is used to represent timestamps. It is required to	3604	The type C<double> is used to represent timestamps. It is required to
3587	have at least 51 bits of mantissa (and 9 bits of exponent), which is good	3605	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
…		…
3591	=back	3609	=back
3592		3610
3593	If you know of other additional requirements drop me a note.	3611	If you know of other additional requirements drop me a note.
3594		3612
3595		3613
3596	=head1 COMPILER WARNINGS	3614	=head1 ALGORITHMIC COMPLEXITIES
3597		3615
3598	Depending on your compiler and compiler settings, you might get no or a	3616	In this section the complexities of (many of) the algorithms used inside
3599	lot of warnings when compiling libev code. Some people are apparently	3617	libev will be documented. For complexity discussions about backends see
3600	scared by this.	3618	the documentation for C<ev_default_init>.
3601		3619
3602	However, these are unavoidable for many reasons. For one, each compiler	3620	All of the following are about amortised time: If an array needs to be
3603	has different warnings, and each user has different tastes regarding	3621	extended, libev needs to realloc and move the whole array, but this
3604	warning options. "Warn-free" code therefore cannot be a goal except when	3622	happens asymptotically rarer with higher number of elements, so O(1) might
3605	targeting a specific compiler and compiler-version.	3623	mean that libev does a lengthy realloc operation in rare cases, but on
		3624	average it is much faster and asymptotically approaches constant time.
3606		3625
3607	Another reason is that some compiler warnings require elaborate	3626	=over 4
3608	workarounds, or other changes to the code that make it less clear and less
3609	maintainable.
3610		3627
3611	And of course, some compiler warnings are just plain stupid, or simply	3628	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
3612	wrong (because they don't actually warn about the condition their message
3613	seems to warn about).
3614		3629
3615	While libev is written to generate as few warnings as possible,	3630	This means that, when you have a watcher that triggers in one hour and
3616	"warn-free" code is not a goal, and it is recommended not to build libev	3631	there are 100 watchers that would trigger before that, then inserting will
3617	with any compiler warnings enabled unless you are prepared to cope with	3632	have to skip roughly seven (C<ld 100>) of these watchers.
3618	them (e.g. by ignoring them). Remember that warnings are just that:
3619	warnings, not errors, or proof of bugs.
3620		3633
		3634	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
3621		3635
3622	=head1 VALGRIND	3636	That means that changing a timer costs less than removing/adding them,
		3637	as only the relative motion in the event queue has to be paid for.
3623		3638
3624	Valgrind has a special section here because it is a popular tool that is	3639	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
3625	highly useful, but valgrind reports are very hard to interpret.
3626		3640
3627	If you think you found a bug (memory leak, uninitialised data access etc.)	3641	These just add the watcher into an array or at the head of a list.
3628	in libev, then check twice: If valgrind reports something like:
3629		3642
3630	==2274== definitely lost: 0 bytes in 0 blocks.	3643	=item Stopping check/prepare/idle/fork/async watchers: O(1)
3631	==2274== possibly lost: 0 bytes in 0 blocks.
3632	==2274== still reachable: 256 bytes in 1 blocks.
3633		3644
3634	Then there is no memory leak. Similarly, under some circumstances,	3645	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
3635	valgrind might report kernel bugs as if it were a bug in libev, or it
3636	might be confused (it is a very good tool, but only a tool).
3637		3646
3638	If you are unsure about something, feel free to contact the mailing list	3647	These watchers are stored in lists, so they need to be walked to find the
3639	with the full valgrind report and an explanation on why you think this is	3648	correct watcher to remove. The lists are usually short (you don't usually
3640	a bug in libev. However, don't be annoyed when you get a brisk "this is	3649	have many watchers waiting for the same fd or signal: one is typical, two
3641	no bug" answer and take the chance of learning how to interpret valgrind	3650	is rare).
3642	properly.
3643		3651
3644	If you need, for some reason, empty reports from valgrind for your project	3652	=item Finding the next timer in each loop iteration: O(1)
3645	I suggest using suppression lists.	3653
		3654	By virtue of using a binary or 4-heap, the next timer is always found at a
		3655	fixed position in the storage array.
		3656
		3657	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		3658
		3659	A change means an I/O watcher gets started or stopped, which requires
		3660	libev to recalculate its status (and possibly tell the kernel, depending
		3661	on backend and whether C<ev_io_set> was used).
		3662
		3663	=item Activating one watcher (putting it into the pending state): O(1)
		3664
		3665	=item Priority handling: O(number_of_priorities)
		3666
		3667	Priorities are implemented by allocating some space for each
		3668	priority. When doing priority-based operations, libev usually has to
		3669	linearly search all the priorities, but starting/stopping and activating
		3670	watchers becomes O(1) with respect to priority handling.
		3671
		3672	=item Sending an ev_async: O(1)
		3673
		3674	=item Processing ev_async_send: O(number_of_async_watchers)
		3675
		3676	=item Processing signals: O(max_signal_number)
		3677
		3678	Sending involves a system call I<iff> there were no other C<ev_async_send>
		3679	calls in the current loop iteration. Checking for async and signal events
		3680	involves iterating over all running async watchers or all signal numbers.
		3681
		3682	=back
3646		3683
3647		3684
3648	=head1 AUTHOR	3685	=head1 AUTHOR
3649		3686
3650	Marc Lehmann <libev@schmorp.de>.	3687	Marc Lehmann <libev@schmorp.de>.

Diff Legend

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

Comparing libev/ev.pod (file contents): Revision 1.184 by root, Tue Sep 23 09:11:14 2008 UTC vs. Revision 1.197 by root, Tue Oct 21 20:52:30 2008 UTC

Diff Legend

Comparing libev/ev.pod (file contents):
Revision 1.184 by root, Tue Sep 23 09:11:14 2008 UTC vs.
Revision 1.197 by root, Tue Oct 21 20:52:30 2008 UTC