ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev.pod
(Generate patch)

Comparing libev/ev.pod (file contents):
Revision 1.275 by root, Sat Dec 26 09:21:54 2009 UTC vs.
Revision 1.451 by root, Mon Jun 24 00:19:26 2019 UTC

1=encoding utf-8
2
1=head1 NAME 3=head1 NAME
2 4
3libev - a high performance full-featured event loop written in C 5libev - a high performance full-featured event loop written in C
4 6
5=head1 SYNOPSIS 7=head1 SYNOPSIS
26 puts ("stdin ready"); 28 puts ("stdin ready");
27 // for one-shot events, one must manually stop the watcher 29 // for one-shot events, one must manually stop the watcher
28 // with its corresponding stop function. 30 // with its corresponding stop function.
29 ev_io_stop (EV_A_ w); 31 ev_io_stop (EV_A_ w);
30 32
31 // this causes all nested ev_loop's to stop iterating 33 // this causes all nested ev_run's to stop iterating
32 ev_unloop (EV_A_ EVUNLOOP_ALL); 34 ev_break (EV_A_ EVBREAK_ALL);
33 } 35 }
34 36
35 // another callback, this time for a time-out 37 // another callback, this time for a time-out
36 static void 38 static void
37 timeout_cb (EV_P_ ev_timer *w, int revents) 39 timeout_cb (EV_P_ ev_timer *w, int revents)
38 { 40 {
39 puts ("timeout"); 41 puts ("timeout");
40 // this causes the innermost ev_loop to stop iterating 42 // this causes the innermost ev_run to stop iterating
41 ev_unloop (EV_A_ EVUNLOOP_ONE); 43 ev_break (EV_A_ EVBREAK_ONE);
42 } 44 }
43 45
44 int 46 int
45 main (void) 47 main (void)
46 { 48 {
47 // use the default event loop unless you have special needs 49 // use the default event loop unless you have special needs
48 struct ev_loop *loop = ev_default_loop (0); 50 struct ev_loop *loop = EV_DEFAULT;
49 51
50 // initialise an io watcher, then start it 52 // initialise an io watcher, then start it
51 // this one will watch for stdin to become readable 53 // this one will watch for stdin to become readable
52 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ); 54 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
53 ev_io_start (loop, &stdin_watcher); 55 ev_io_start (loop, &stdin_watcher);
56 // simple non-repeating 5.5 second timeout 58 // simple non-repeating 5.5 second timeout
57 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.); 59 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
58 ev_timer_start (loop, &timeout_watcher); 60 ev_timer_start (loop, &timeout_watcher);
59 61
60 // now wait for events to arrive 62 // now wait for events to arrive
61 ev_loop (loop, 0); 63 ev_run (loop, 0);
62 64
63 // unloop was called, so exit 65 // break was called, so exit
64 return 0; 66 return 0;
65 } 67 }
66 68
67=head1 ABOUT THIS DOCUMENT 69=head1 ABOUT THIS DOCUMENT
68 70
75While this document tries to be as complete as possible in documenting 77While this document tries to be as complete as possible in documenting
76libev, its usage and the rationale behind its design, it is not a tutorial 78libev, its usage and the rationale behind its design, it is not a tutorial
77on event-based programming, nor will it introduce event-based programming 79on event-based programming, nor will it introduce event-based programming
78with libev. 80with libev.
79 81
80Familarity with event based programming techniques in general is assumed 82Familiarity with event based programming techniques in general is assumed
81throughout this document. 83throughout this document.
84
85=head1 WHAT TO READ WHEN IN A HURRY
86
87This manual tries to be very detailed, but unfortunately, this also makes
88it very long. If you just want to know the basics of libev, I suggest
89reading L</ANATOMY OF A WATCHER>, then the L</EXAMPLE PROGRAM> above and
90look up the missing functions in L</GLOBAL FUNCTIONS> and the C<ev_io> and
91C<ev_timer> sections in L</WATCHER TYPES>.
82 92
83=head1 ABOUT LIBEV 93=head1 ABOUT LIBEV
84 94
85Libev is an event loop: you register interest in certain events (such as a 95Libev is an event loop: you register interest in certain events (such as a
86file descriptor being readable or a timeout occurring), and it will manage 96file descriptor being readable or a timeout occurring), and it will manage
95details of the event, and then hand it over to libev by I<starting> the 105details of the event, and then hand it over to libev by I<starting> the
96watcher. 106watcher.
97 107
98=head2 FEATURES 108=head2 FEATURES
99 109
100Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the 110Libev supports C<select>, C<poll>, the Linux-specific aio and C<epoll>
101BSD-specific C<kqueue> and the Solaris-specific event port mechanisms 111interfaces, the BSD-specific C<kqueue> and the Solaris-specific event port
102for file descriptor events (C<ev_io>), the Linux C<inotify> interface 112mechanisms for file descriptor events (C<ev_io>), the Linux C<inotify>
103(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner 113interface (for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
104inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative 114inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
105timers (C<ev_timer>), absolute timers with customised rescheduling 115timers (C<ev_timer>), absolute timers with customised rescheduling
106(C<ev_periodic>), synchronous signals (C<ev_signal>), process status 116(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
107change events (C<ev_child>), and event watchers dealing with the event 117change events (C<ev_child>), and event watchers dealing with the event
108loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and 118loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
124this argument. 134this argument.
125 135
126=head2 TIME REPRESENTATION 136=head2 TIME REPRESENTATION
127 137
128Libev represents time as a single floating point number, representing 138Libev represents time as a single floating point number, representing
129the (fractional) number of seconds since the (POSIX) epoch (somewhere 139the (fractional) number of seconds since the (POSIX) epoch (in practice
130near the beginning of 1970, details are complicated, don't ask). This 140somewhere near the beginning of 1970, details are complicated, don't
131type is called C<ev_tstamp>, which is what you should use too. It usually 141ask). This type is called C<ev_tstamp>, which is what you should use
132aliases to the C<double> type in C. When you need to do any calculations 142too. It usually aliases to the C<double> type in C. When you need to do
133on it, you should treat it as some floating point value. Unlike the name 143any calculations on it, you should treat it as some floating point value.
144
134component C<stamp> might indicate, it is also used for time differences 145Unlike the name component C<stamp> might indicate, it is also used for
135throughout libev. 146time differences (e.g. delays) throughout libev.
136 147
137=head1 ERROR HANDLING 148=head1 ERROR HANDLING
138 149
139Libev knows three classes of errors: operating system errors, usage errors 150Libev knows three classes of errors: operating system errors, usage errors
140and internal errors (bugs). 151and internal errors (bugs).
164 175
165=item ev_tstamp ev_time () 176=item ev_tstamp ev_time ()
166 177
167Returns the current time as libev would use it. Please note that the 178Returns the current time as libev would use it. Please note that the
168C<ev_now> function is usually faster and also often returns the timestamp 179C<ev_now> function is usually faster and also often returns the timestamp
169you actually want to know. 180you actually want to know. Also interesting is the combination of
181C<ev_now_update> and C<ev_now>.
170 182
171=item ev_sleep (ev_tstamp interval) 183=item ev_sleep (ev_tstamp interval)
172 184
173Sleep for the given interval: The current thread will be blocked until 185Sleep for the given interval: The current thread will be blocked
174either it is interrupted or the given time interval has passed. Basically 186until either it is interrupted or the given time interval has
187passed (approximately - it might return a bit earlier even if not
188interrupted). Returns immediately if C<< interval <= 0 >>.
189
175this is a sub-second-resolution C<sleep ()>. 190Basically this is a sub-second-resolution C<sleep ()>.
191
192The range of the C<interval> is limited - libev only guarantees to work
193with sleep times of up to one day (C<< interval <= 86400 >>).
176 194
177=item int ev_version_major () 195=item int ev_version_major ()
178 196
179=item int ev_version_minor () 197=item int ev_version_minor ()
180 198
191as this indicates an incompatible change. Minor versions are usually 209as this indicates an incompatible change. Minor versions are usually
192compatible to older versions, so a larger minor version alone is usually 210compatible to older versions, so a larger minor version alone is usually
193not a problem. 211not a problem.
194 212
195Example: Make sure we haven't accidentally been linked against the wrong 213Example: Make sure we haven't accidentally been linked against the wrong
196version. 214version (note, however, that this will not detect other ABI mismatches,
215such as LFS or reentrancy).
197 216
198 assert (("libev version mismatch", 217 assert (("libev version mismatch",
199 ev_version_major () == EV_VERSION_MAJOR 218 ev_version_major () == EV_VERSION_MAJOR
200 && ev_version_minor () >= EV_VERSION_MINOR)); 219 && ev_version_minor () >= EV_VERSION_MINOR));
201 220
212 assert (("sorry, no epoll, no sex", 231 assert (("sorry, no epoll, no sex",
213 ev_supported_backends () & EVBACKEND_EPOLL)); 232 ev_supported_backends () & EVBACKEND_EPOLL));
214 233
215=item unsigned int ev_recommended_backends () 234=item unsigned int ev_recommended_backends ()
216 235
217Return the set of all backends compiled into this binary of libev and also 236Return the set of all backends compiled into this binary of libev and
218recommended for this platform. This set is often smaller than the one 237also recommended for this platform, meaning it will work for most file
238descriptor types. This set is often smaller than the one returned by
219returned by C<ev_supported_backends>, as for example kqueue is broken on 239C<ev_supported_backends>, as for example kqueue is broken on most BSDs
220most BSDs and will not be auto-detected unless you explicitly request it 240and will not be auto-detected unless you explicitly request it (assuming
221(assuming you know what you are doing). This is the set of backends that 241you know what you are doing). This is the set of backends that libev will
222libev will probe for if you specify no backends explicitly. 242probe for if you specify no backends explicitly.
223 243
224=item unsigned int ev_embeddable_backends () 244=item unsigned int ev_embeddable_backends ()
225 245
226Returns the set of backends that are embeddable in other event loops. This 246Returns the set of backends that are embeddable in other event loops. This
227is the theoretical, all-platform, value. To find which backends 247value is platform-specific but can include backends not available on the
228might be supported on the current system, you would need to look at 248current system. To find which embeddable backends might be supported on
229C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for 249the current system, you would need to look at C<ev_embeddable_backends ()
230recommended ones. 250& ev_supported_backends ()>, likewise for recommended ones.
231 251
232See the description of C<ev_embed> watchers for more info. 252See the description of C<ev_embed> watchers for more info.
233 253
234=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT] 254=item ev_set_allocator (void *(*cb)(void *ptr, long size) throw ())
235 255
236Sets the allocation function to use (the prototype is similar - the 256Sets the allocation function to use (the prototype is similar - the
237semantics are identical to the C<realloc> C89/SuS/POSIX function). It is 257semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
238used to allocate and free memory (no surprises here). If it returns zero 258used to allocate and free memory (no surprises here). If it returns zero
239when memory needs to be allocated (C<size != 0>), the library might abort 259when memory needs to be allocated (C<size != 0>), the library might abort
245 265
246You could override this function in high-availability programs to, say, 266You could override this function in high-availability programs to, say,
247free some memory if it cannot allocate memory, to use a special allocator, 267free some memory if it cannot allocate memory, to use a special allocator,
248or even to sleep a while and retry until some memory is available. 268or even to sleep a while and retry until some memory is available.
249 269
270Example: The following is the C<realloc> function that libev itself uses
271which should work with C<realloc> and C<free> functions of all kinds and
272is probably a good basis for your own implementation.
273
274 static void *
275 ev_realloc_emul (void *ptr, long size) EV_NOEXCEPT
276 {
277 if (size)
278 return realloc (ptr, size);
279
280 free (ptr);
281 return 0;
282 }
283
250Example: Replace the libev allocator with one that waits a bit and then 284Example: Replace the libev allocator with one that waits a bit and then
251retries (example requires a standards-compliant C<realloc>). 285retries.
252 286
253 static void * 287 static void *
254 persistent_realloc (void *ptr, size_t size) 288 persistent_realloc (void *ptr, size_t size)
255 { 289 {
290 if (!size)
291 {
292 free (ptr);
293 return 0;
294 }
295
256 for (;;) 296 for (;;)
257 { 297 {
258 void *newptr = realloc (ptr, size); 298 void *newptr = realloc (ptr, size);
259 299
260 if (newptr) 300 if (newptr)
265 } 305 }
266 306
267 ... 307 ...
268 ev_set_allocator (persistent_realloc); 308 ev_set_allocator (persistent_realloc);
269 309
270=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT] 310=item ev_set_syserr_cb (void (*cb)(const char *msg) throw ())
271 311
272Set the callback function to call on a retryable system call error (such 312Set the callback function to call on a retryable system call error (such
273as failed select, poll, epoll_wait). The message is a printable string 313as failed select, poll, epoll_wait). The message is a printable string
274indicating the system call or subsystem causing the problem. If this 314indicating the system call or subsystem causing the problem. If this
275callback is set, then libev will expect it to remedy the situation, no 315callback is set, then libev will expect it to remedy the situation, no
287 } 327 }
288 328
289 ... 329 ...
290 ev_set_syserr_cb (fatal_error); 330 ev_set_syserr_cb (fatal_error);
291 331
332=item ev_feed_signal (int signum)
333
334This function can be used to "simulate" a signal receive. It is completely
335safe to call this function at any time, from any context, including signal
336handlers or random threads.
337
338Its main use is to customise signal handling in your process, especially
339in the presence of threads. For example, you could block signals
340by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
341creating any loops), and in one thread, use C<sigwait> or any other
342mechanism to wait for signals, then "deliver" them to libev by calling
343C<ev_feed_signal>.
344
292=back 345=back
293 346
294=head1 FUNCTIONS CONTROLLING THE EVENT LOOP 347=head1 FUNCTIONS CONTROLLING EVENT LOOPS
295 348
296An event loop is described by a C<struct ev_loop *> (the C<struct> 349An event loop is described by a C<struct ev_loop *> (the C<struct> is
297is I<not> optional in this case, as there is also an C<ev_loop> 350I<not> optional in this case unless libev 3 compatibility is disabled, as
298I<function>). 351libev 3 had an C<ev_loop> function colliding with the struct name).
299 352
300The library knows two types of such loops, the I<default> loop, which 353The library knows two types of such loops, the I<default> loop, which
301supports signals and child events, and dynamically created loops which do 354supports child process events, and dynamically created event loops which
302not. 355do not.
303 356
304=over 4 357=over 4
305 358
306=item struct ev_loop *ev_default_loop (unsigned int flags) 359=item struct ev_loop *ev_default_loop (unsigned int flags)
307 360
308This will initialise the default event loop if it hasn't been initialised 361This returns the "default" event loop object, which is what you should
309yet and return it. If the default loop could not be initialised, returns 362normally use when you just need "the event loop". Event loop objects and
310false. If it already was initialised it simply returns it (and ignores the 363the C<flags> parameter are described in more detail in the entry for
311flags. If that is troubling you, check C<ev_backend ()> afterwards). 364C<ev_loop_new>.
365
366If the default loop is already initialised then this function simply
367returns it (and ignores the flags. If that is troubling you, check
368C<ev_backend ()> afterwards). Otherwise it will create it with the given
369flags, which should almost always be C<0>, unless the caller is also the
370one calling C<ev_run> or otherwise qualifies as "the main program".
312 371
313If you don't know what event loop to use, use the one returned from this 372If you don't know what event loop to use, use the one returned from this
314function. 373function (or via the C<EV_DEFAULT> macro).
315 374
316Note that this function is I<not> thread-safe, so if you want to use it 375Note that this function is I<not> thread-safe, so if you want to use it
317from multiple threads, you have to lock (note also that this is unlikely, 376from multiple threads, you have to employ some kind of mutex (note also
318as loops cannot be shared easily between threads anyway). 377that this case is unlikely, as loops cannot be shared easily between
378threads anyway).
319 379
320The default loop is the only loop that can handle C<ev_signal> and 380The default loop is the only loop that can handle C<ev_child> watchers,
321C<ev_child> watchers, and to do this, it always registers a handler 381and to do this, it always registers a handler for C<SIGCHLD>. If this is
322for C<SIGCHLD>. If this is a problem for your application you can either 382a problem for your application you can either create a dynamic loop with
323create a dynamic loop with C<ev_loop_new> that doesn't do that, or you 383C<ev_loop_new> which doesn't do that, or you can simply overwrite the
324can simply overwrite the C<SIGCHLD> signal handler I<after> calling 384C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
325C<ev_default_init>. 385
386Example: This is the most typical usage.
387
388 if (!ev_default_loop (0))
389 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
390
391Example: Restrict libev to the select and poll backends, and do not allow
392environment settings to be taken into account:
393
394 ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
395
396=item struct ev_loop *ev_loop_new (unsigned int flags)
397
398This will create and initialise a new event loop object. If the loop
399could not be initialised, returns false.
400
401This function is thread-safe, and one common way to use libev with
402threads is indeed to create one loop per thread, and using the default
403loop in the "main" or "initial" thread.
326 404
327The flags argument can be used to specify special behaviour or specific 405The flags argument can be used to specify special behaviour or specific
328backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>). 406backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
329 407
330The following flags are supported: 408The following flags are supported:
340 418
341If this flag bit is or'ed into the flag value (or the program runs setuid 419If this flag bit is or'ed into the flag value (or the program runs setuid
342or setgid) then libev will I<not> look at the environment variable 420or setgid) then libev will I<not> look at the environment variable
343C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will 421C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
344override the flags completely if it is found in the environment. This is 422override the flags completely if it is found in the environment. This is
345useful to try out specific backends to test their performance, or to work 423useful to try out specific backends to test their performance, to work
346around bugs. 424around bugs, or to make libev threadsafe (accessing environment variables
425cannot be done in a threadsafe way, but usually it works if no other
426thread modifies them).
347 427
348=item C<EVFLAG_FORKCHECK> 428=item C<EVFLAG_FORKCHECK>
349 429
350Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after 430Instead of calling C<ev_loop_fork> manually after a fork, you can also
351a fork, you can also make libev check for a fork in each iteration by 431make libev check for a fork in each iteration by enabling this flag.
352enabling this flag.
353 432
354This works by calling C<getpid ()> on every iteration of the loop, 433This works by calling C<getpid ()> on every iteration of the loop,
355and thus this might slow down your event loop if you do a lot of loop 434and thus this might slow down your event loop if you do a lot of loop
356iterations and little real work, but is usually not noticeable (on my 435iterations and little real work, but is usually not noticeable (on my
357GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence 436GNU/Linux system for example, C<getpid> is actually a simple 5-insn
358without a system call and thus I<very> fast, but my GNU/Linux system also has 437sequence without a system call and thus I<very> fast, but my GNU/Linux
359C<pthread_atfork> which is even faster). 438system also has C<pthread_atfork> which is even faster). (Update: glibc
439versions 2.25 apparently removed the C<getpid> optimisation again).
360 440
361The big advantage of this flag is that you can forget about fork (and 441The big advantage of this flag is that you can forget about fork (and
362forget about forgetting to tell libev about forking) when you use this 442forget about forgetting to tell libev about forking, although you still
363flag. 443have to ignore C<SIGPIPE>) when you use this flag.
364 444
365This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS> 445This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
366environment variable. 446environment variable.
367 447
368=item C<EVFLAG_NOINOTIFY> 448=item C<EVFLAG_NOINOTIFY>
369 449
370When this flag is specified, then libev will not attempt to use the 450When this flag is specified, then libev will not attempt to use the
371I<inotify> API for it's C<ev_stat> watchers. Apart from debugging and 451I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
372testing, this flag can be useful to conserve inotify file descriptors, as 452testing, this flag can be useful to conserve inotify file descriptors, as
373otherwise each loop using C<ev_stat> watchers consumes one inotify handle. 453otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
374 454
375=item C<EVFLAG_NOSIGFD> 455=item C<EVFLAG_SIGNALFD>
376 456
377When this flag is specified, then libev will not attempt to use the 457When this flag is specified, then libev will attempt to use the
378I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This is 458I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
379probably only useful to work around any bugs in libev. Consequently, this 459delivers signals synchronously, which makes it both faster and might make
380flag might go away once the signalfd functionality is considered stable, 460it possible to get the queued signal data. It can also simplify signal
381so it's useful mostly in environment variables and not in program code. 461handling with threads, as long as you properly block signals in your
462threads that are not interested in handling them.
463
464Signalfd will not be used by default as this changes your signal mask, and
465there are a lot of shoddy libraries and programs (glib's threadpool for
466example) that can't properly initialise their signal masks.
467
468=item C<EVFLAG_NOSIGMASK>
469
470When this flag is specified, then libev will avoid to modify the signal
471mask. Specifically, this means you have to make sure signals are unblocked
472when you want to receive them.
473
474This behaviour is useful when you want to do your own signal handling, or
475want to handle signals only in specific threads and want to avoid libev
476unblocking the signals.
477
478It's also required by POSIX in a threaded program, as libev calls
479C<sigprocmask>, whose behaviour is officially unspecified.
480
481This flag's behaviour will become the default in future versions of libev.
382 482
383=item C<EVBACKEND_SELECT> (value 1, portable select backend) 483=item C<EVBACKEND_SELECT> (value 1, portable select backend)
384 484
385This is your standard select(2) backend. Not I<completely> standard, as 485This is your standard select(2) backend. Not I<completely> standard, as
386libev tries to roll its own fd_set with no limits on the number of fds, 486libev tries to roll its own fd_set with no limits on the number of fds,
414=item C<EVBACKEND_EPOLL> (value 4, Linux) 514=item C<EVBACKEND_EPOLL> (value 4, Linux)
415 515
416Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9 516Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
417kernels). 517kernels).
418 518
419For few fds, this backend is a bit little slower than poll and select, 519For few fds, this backend is a bit little slower than poll and select, but
420but it scales phenomenally better. While poll and select usually scale 520it scales phenomenally better. While poll and select usually scale like
421like O(total_fds) where n is the total number of fds (or the highest fd), 521O(total_fds) where total_fds is the total number of fds (or the highest
422epoll scales either O(1) or O(active_fds). 522fd), epoll scales either O(1) or O(active_fds).
423 523
424The epoll mechanism deserves honorable mention as the most misdesigned 524The epoll mechanism deserves honorable mention as the most misdesigned
425of the more advanced event mechanisms: mere annoyances include silently 525of the more advanced event mechanisms: mere annoyances include silently
426dropping file descriptors, requiring a system call per change per file 526dropping file descriptors, requiring a system call per change per file
427descriptor (and unnecessary guessing of parameters), problems with dup and 527descriptor (and unnecessary guessing of parameters), problems with dup,
528returning before the timeout value, resulting in additional iterations
529(and only giving 5ms accuracy while select on the same platform gives
428so on. The biggest issue is fork races, however - if a program forks then 5300.1ms) and so on. The biggest issue is fork races, however - if a program
429I<both> parent and child process have to recreate the epoll set, which can 531forks then I<both> parent and child process have to recreate the epoll
430take considerable time (one syscall per file descriptor) and is of course 532set, which can take considerable time (one syscall per file descriptor)
431hard to detect. 533and is of course hard to detect.
432 534
433Epoll is also notoriously buggy - embedding epoll fds I<should> work, but 535Epoll is also notoriously buggy - embedding epoll fds I<should> work,
434of course I<doesn't>, and epoll just loves to report events for totally 536but of course I<doesn't>, and epoll just loves to report events for
435I<different> file descriptors (even already closed ones, so one cannot 537totally I<different> file descriptors (even already closed ones, so
436even remove them from the set) than registered in the set (especially 538one cannot even remove them from the set) than registered in the set
437on SMP systems). Libev tries to counter these spurious notifications by 539(especially on SMP systems). Libev tries to counter these spurious
438employing an additional generation counter and comparing that against the 540notifications by employing an additional generation counter and comparing
439events to filter out spurious ones, recreating the set when required. 541that against the events to filter out spurious ones, recreating the set
542when required. Epoll also erroneously rounds down timeouts, but gives you
543no way to know when and by how much, so sometimes you have to busy-wait
544because epoll returns immediately despite a nonzero timeout. And last
545not least, it also refuses to work with some file descriptors which work
546perfectly fine with C<select> (files, many character devices...).
547
548Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
549cobbled together in a hurry, no thought to design or interaction with
550others. Oh, the pain, will it ever stop...
440 551
441While stopping, setting and starting an I/O watcher in the same iteration 552While stopping, setting and starting an I/O watcher in the same iteration
442will result in some caching, there is still a system call per such 553will result in some caching, there is still a system call per such
443incident (because the same I<file descriptor> could point to a different 554incident (because the same I<file descriptor> could point to a different
444I<file description> now), so its best to avoid that. Also, C<dup ()>'ed 555I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
456All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or 567All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
457faster than epoll for maybe up to a hundred file descriptors, depending on 568faster than epoll for maybe up to a hundred file descriptors, depending on
458the usage. So sad. 569the usage. So sad.
459 570
460While nominally embeddable in other event loops, this feature is broken in 571While nominally embeddable in other event loops, this feature is broken in
461all kernel versions tested so far. 572a lot of kernel revisions, but probably(!) works in current versions.
573
574This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
575C<EVBACKEND_POLL>.
576
577=item C<EVBACKEND_LINUXAIO> (value 64, Linux)
578
579Use the linux-specific linux aio (I<not> C<< aio(7) >> but C<<
580io_submit(2) >>) event interface available in post-4.18 kernels.
581
582If this backend works for you (as of this writing, it was very
583experimental), it is the best event interface available on linux and might
584be well worth enabling it - if it isn't available in your kernel this will
585be detected and this backend will be skipped.
586
587This backend can batch oneshot requests and supports a user-space ring
588buffer to receive events. It also doesn't suffer from most of the design
589problems of epoll (such as not being able to remove event sources from the
590epoll set), and generally sounds too good to be true. Because, this being
591the linux kernel, of course it suffers from a whole new set of limitations.
592
593For one, it is not easily embeddable (but probably could be done using
594an event fd at some extra overhead). It also is subject to a system wide
595limit that can be configured in F</proc/sys/fs/aio-max-nr> - each loop
596currently requires C<61> of this number. If no aio requests are left, this
597backend will be skipped during initialisation.
598
599Most problematic in practise, however, is that not all file descriptors
600work with it. For example, in linux 5.1, tcp sockets, pipes, event fds,
601files, F</dev/null> and a few others are supported, but ttys do not work
602properly (a known bug that the kernel developers don't care about, see
603L<https://lore.kernel.org/patchwork/patch/1047453/>), so this is not
604(yet?) a generic event polling interface.
605
606Overall, it seems the linux developers just don't want it to have a
607generic event handling mechanism other than C<select> or C<poll>.
608
609To work around the fd type problem, the current version of libev uses
610epoll as a fallback for file deescriptor types that do not work. Epoll
611is used in, kind of, slow mode that hopefully avoids most of its design
612problems and requires 1-3 extra syscalls per active fd every iteration.
462 613
463This backend maps C<EV_READ> and C<EV_WRITE> in the same way as 614This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
464C<EVBACKEND_POLL>. 615C<EVBACKEND_POLL>.
465 616
466=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones) 617=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
481 632
482It scales in the same way as the epoll backend, but the interface to the 633It scales in the same way as the epoll backend, but the interface to the
483kernel is more efficient (which says nothing about its actual speed, of 634kernel is more efficient (which says nothing about its actual speed, of
484course). While stopping, setting and starting an I/O watcher does never 635course). While stopping, setting and starting an I/O watcher does never
485cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to 636cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
486two event changes per incident. Support for C<fork ()> is very bad (but 637two event changes per incident. Support for C<fork ()> is very bad (you
487sane, unlike epoll) and it drops fds silently in similarly hard-to-detect 638might have to leak fd's on fork, but it's more sane than epoll) and it
488cases 639drops fds silently in similarly hard-to-detect cases.
489 640
490This backend usually performs well under most conditions. 641This backend usually performs well under most conditions.
491 642
492While nominally embeddable in other event loops, this doesn't work 643While nominally embeddable in other event loops, this doesn't work
493everywhere, so you might need to test for this. And since it is broken 644everywhere, so you might need to test for this. And since it is broken
510=item C<EVBACKEND_PORT> (value 32, Solaris 10) 661=item C<EVBACKEND_PORT> (value 32, Solaris 10)
511 662
512This uses the Solaris 10 event port mechanism. As with everything on Solaris, 663This uses the Solaris 10 event port mechanism. As with everything on Solaris,
513it's really slow, but it still scales very well (O(active_fds)). 664it's really slow, but it still scales very well (O(active_fds)).
514 665
515Please note that Solaris event ports can deliver a lot of spurious
516notifications, so you need to use non-blocking I/O or other means to avoid
517blocking when no data (or space) is available.
518
519While this backend scales well, it requires one system call per active 666While this backend scales well, it requires one system call per active
520file descriptor per loop iteration. For small and medium numbers of file 667file descriptor per loop iteration. For small and medium numbers of file
521descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend 668descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
522might perform better. 669might perform better.
523 670
524On the positive side, with the exception of the spurious readiness 671On the positive side, this backend actually performed fully to
525notifications, this backend actually performed fully to specification
526in all tests and is fully embeddable, which is a rare feat among the 672specification in all tests and is fully embeddable, which is a rare feat
527OS-specific backends (I vastly prefer correctness over speed hacks). 673among the OS-specific backends (I vastly prefer correctness over speed
674hacks).
675
676On the negative side, the interface is I<bizarre> - so bizarre that
677even sun itself gets it wrong in their code examples: The event polling
678function sometimes returns events to the caller even though an error
679occurred, but with no indication whether it has done so or not (yes, it's
680even documented that way) - deadly for edge-triggered interfaces where you
681absolutely have to know whether an event occurred or not because you have
682to re-arm the watcher.
683
684Fortunately libev seems to be able to work around these idiocies.
528 685
529This backend maps C<EV_READ> and C<EV_WRITE> in the same way as 686This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
530C<EVBACKEND_POLL>. 687C<EVBACKEND_POLL>.
531 688
532=item C<EVBACKEND_ALL> 689=item C<EVBACKEND_ALL>
533 690
534Try all backends (even potentially broken ones that wouldn't be tried 691Try all backends (even potentially broken ones that wouldn't be tried
535with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as 692with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
536C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>. 693C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
537 694
538It is definitely not recommended to use this flag. 695It is definitely not recommended to use this flag, use whatever
696C<ev_recommended_backends ()> returns, or simply do not specify a backend
697at all.
698
699=item C<EVBACKEND_MASK>
700
701Not a backend at all, but a mask to select all backend bits from a
702C<flags> value, in case you want to mask out any backends from a flags
703value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
539 704
540=back 705=back
541 706
542If one or more of the backend flags are or'ed into the flags value, 707If one or more of the backend flags are or'ed into the flags value,
543then only these backends will be tried (in the reverse order as listed 708then only these backends will be tried (in the reverse order as listed
544here). If none are specified, all backends in C<ev_recommended_backends 709here). If none are specified, all backends in C<ev_recommended_backends
545()> will be tried. 710()> will be tried.
546 711
547Example: This is the most typical usage.
548
549 if (!ev_default_loop (0))
550 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
551
552Example: Restrict libev to the select and poll backends, and do not allow
553environment settings to be taken into account:
554
555 ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
556
557Example: Use whatever libev has to offer, but make sure that kqueue is
558used if available (warning, breaks stuff, best use only with your own
559private event loop and only if you know the OS supports your types of
560fds):
561
562 ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
563
564=item struct ev_loop *ev_loop_new (unsigned int flags)
565
566Similar to C<ev_default_loop>, but always creates a new event loop that is
567always distinct from the default loop. Unlike the default loop, it cannot
568handle signal and child watchers, and attempts to do so will be greeted by
569undefined behaviour (or a failed assertion if assertions are enabled).
570
571Note that this function I<is> thread-safe, and the recommended way to use
572libev with threads is indeed to create one loop per thread, and using the
573default loop in the "main" or "initial" thread.
574
575Example: Try to create a event loop that uses epoll and nothing else. 712Example: Try to create a event loop that uses epoll and nothing else.
576 713
577 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV); 714 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
578 if (!epoller) 715 if (!epoller)
579 fatal ("no epoll found here, maybe it hides under your chair"); 716 fatal ("no epoll found here, maybe it hides under your chair");
580 717
718Example: Use whatever libev has to offer, but make sure that kqueue is
719used if available.
720
721 struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
722
723Example: Similarly, on linux, you mgiht want to take advantage of the
724linux aio backend if possible, but fall back to something else if that
725isn't available.
726
727 struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_LINUXAIO);
728
581=item ev_default_destroy () 729=item ev_loop_destroy (loop)
582 730
583Destroys the default loop again (frees all memory and kernel state 731Destroys an event loop object (frees all memory and kernel state
584etc.). None of the active event watchers will be stopped in the normal 732etc.). None of the active event watchers will be stopped in the normal
585sense, so e.g. C<ev_is_active> might still return true. It is your 733sense, so e.g. C<ev_is_active> might still return true. It is your
586responsibility to either stop all watchers cleanly yourself I<before> 734responsibility to either stop all watchers cleanly yourself I<before>
587calling this function, or cope with the fact afterwards (which is usually 735calling this function, or cope with the fact afterwards (which is usually
588the easiest thing, you can just ignore the watchers and/or C<free ()> them 736the easiest thing, you can just ignore the watchers and/or C<free ()> them
590 738
591Note that certain global state, such as signal state (and installed signal 739Note that certain global state, such as signal state (and installed signal
592handlers), will not be freed by this function, and related watchers (such 740handlers), will not be freed by this function, and related watchers (such
593as signal and child watchers) would need to be stopped manually. 741as signal and child watchers) would need to be stopped manually.
594 742
595In general it is not advisable to call this function except in the 743This function is normally used on loop objects allocated by
596rare occasion where you really need to free e.g. the signal handling 744C<ev_loop_new>, but it can also be used on the default loop returned by
745C<ev_default_loop>, in which case it is not thread-safe.
746
747Note that it is not advisable to call this function on the default loop
748except in the rare occasion where you really need to free its resources.
597pipe fds. If you need dynamically allocated loops it is better to use 749If you need dynamically allocated loops it is better to use C<ev_loop_new>
598C<ev_loop_new> and C<ev_loop_destroy>. 750and C<ev_loop_destroy>.
599 751
600=item ev_loop_destroy (loop) 752=item ev_loop_fork (loop)
601 753
602Like C<ev_default_destroy>, but destroys an event loop created by an
603earlier call to C<ev_loop_new>.
604
605=item ev_default_fork ()
606
607This function sets a flag that causes subsequent C<ev_loop> iterations 754This function sets a flag that causes subsequent C<ev_run> iterations
608to reinitialise the kernel state for backends that have one. Despite the 755to reinitialise the kernel state for backends that have one. Despite
609name, you can call it anytime, but it makes most sense after forking, in 756the name, you can call it anytime you are allowed to start or stop
610the child process (or both child and parent, but that again makes little 757watchers (except inside an C<ev_prepare> callback), but it makes most
611sense). You I<must> call it in the child before using any of the libev 758sense after forking, in the child process. You I<must> call it (or use
612functions, and it will only take effect at the next C<ev_loop> iteration. 759C<EVFLAG_FORKCHECK>) in the child before resuming or calling C<ev_run>.
760
761In addition, if you want to reuse a loop (via this function or
762C<EVFLAG_FORKCHECK>), you I<also> have to ignore C<SIGPIPE>.
763
764Again, you I<have> to call it on I<any> loop that you want to re-use after
765a fork, I<even if you do not plan to use the loop in the parent>. This is
766because some kernel interfaces *cough* I<kqueue> *cough* do funny things
767during fork.
613 768
614On the other hand, you only need to call this function in the child 769On the other hand, you only need to call this function in the child
615process if and only if you want to use the event library in the child. If 770process if and only if you want to use the event loop in the child. If
616you just fork+exec, you don't have to call it at all. 771you just fork+exec or create a new loop in the child, you don't have to
772call it at all (in fact, C<epoll> is so badly broken that it makes a
773difference, but libev will usually detect this case on its own and do a
774costly reset of the backend).
617 775
618The function itself is quite fast and it's usually not a problem to call 776The function itself is quite fast and it's usually not a problem to call
619it just in case after a fork. To make this easy, the function will fit in 777it just in case after a fork.
620quite nicely into a call to C<pthread_atfork>:
621 778
779Example: Automate calling C<ev_loop_fork> on the default loop when
780using pthreads.
781
782 static void
783 post_fork_child (void)
784 {
785 ev_loop_fork (EV_DEFAULT);
786 }
787
788 ...
622 pthread_atfork (0, 0, ev_default_fork); 789 pthread_atfork (0, 0, post_fork_child);
623
624=item ev_loop_fork (loop)
625
626Like C<ev_default_fork>, but acts on an event loop created by
627C<ev_loop_new>. Yes, you have to call this on every allocated event loop
628after fork that you want to re-use in the child, and how you do this is
629entirely your own problem.
630 790
631=item int ev_is_default_loop (loop) 791=item int ev_is_default_loop (loop)
632 792
633Returns true when the given loop is, in fact, the default loop, and false 793Returns true when the given loop is, in fact, the default loop, and false
634otherwise. 794otherwise.
635 795
636=item unsigned int ev_loop_count (loop) 796=item unsigned int ev_iteration (loop)
637 797
638Returns the count of loop iterations for the loop, which is identical to 798Returns the current iteration count for the event loop, which is identical
639the number of times libev did poll for new events. It starts at C<0> and 799to the number of times libev did poll for new events. It starts at C<0>
640happily wraps around with enough iterations. 800and happily wraps around with enough iterations.
641 801
642This value can sometimes be useful as a generation counter of sorts (it 802This value can sometimes be useful as a generation counter of sorts (it
643"ticks" the number of loop iterations), as it roughly corresponds with 803"ticks" the number of loop iterations), as it roughly corresponds with
644C<ev_prepare> and C<ev_check> calls. 804C<ev_prepare> and C<ev_check> calls - and is incremented between the
805prepare and check phases.
645 806
646=item unsigned int ev_loop_depth (loop) 807=item unsigned int ev_depth (loop)
647 808
648Returns the number of times C<ev_loop> was entered minus the number of 809Returns the number of times C<ev_run> was entered minus the number of
649times C<ev_loop> was exited, in other words, the recursion depth. 810times C<ev_run> was exited normally, in other words, the recursion depth.
650 811
651Outside C<ev_loop>, this number is zero. In a callback, this number is 812Outside C<ev_run>, this number is zero. In a callback, this number is
652C<1>, unless C<ev_loop> was invoked recursively (or from another thread), 813C<1>, unless C<ev_run> was invoked recursively (or from another thread),
653in which case it is higher. 814in which case it is higher.
654 815
655Leaving C<ev_loop> abnormally (setjmp/longjmp, cancelling the thread 816Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
656etc.), doesn't count as exit. 817throwing an exception etc.), doesn't count as "exit" - consider this
818as a hint to avoid such ungentleman-like behaviour unless it's really
819convenient, in which case it is fully supported.
657 820
658=item unsigned int ev_backend (loop) 821=item unsigned int ev_backend (loop)
659 822
660Returns one of the C<EVBACKEND_*> flags indicating the event backend in 823Returns one of the C<EVBACKEND_*> flags indicating the event backend in
661use. 824use.
670 833
671=item ev_now_update (loop) 834=item ev_now_update (loop)
672 835
673Establishes the current time by querying the kernel, updating the time 836Establishes the current time by querying the kernel, updating the time
674returned by C<ev_now ()> in the progress. This is a costly operation and 837returned by C<ev_now ()> in the progress. This is a costly operation and
675is usually done automatically within C<ev_loop ()>. 838is usually done automatically within C<ev_run ()>.
676 839
677This function is rarely useful, but when some event callback runs for a 840This function is rarely useful, but when some event callback runs for a
678very long time without entering the event loop, updating libev's idea of 841very long time without entering the event loop, updating libev's idea of
679the current time is a good idea. 842the current time is a good idea.
680 843
681See also L<The special problem of time updates> in the C<ev_timer> section. 844See also L</The special problem of time updates> in the C<ev_timer> section.
682 845
683=item ev_suspend (loop) 846=item ev_suspend (loop)
684 847
685=item ev_resume (loop) 848=item ev_resume (loop)
686 849
687These two functions suspend and resume a loop, for use when the loop is 850These two functions suspend and resume an event loop, for use when the
688not used for a while and timeouts should not be processed. 851loop is not used for a while and timeouts should not be processed.
689 852
690A typical use case would be an interactive program such as a game: When 853A typical use case would be an interactive program such as a game: When
691the user presses C<^Z> to suspend the game and resumes it an hour later it 854the user presses C<^Z> to suspend the game and resumes it an hour later it
692would be best to handle timeouts as if no time had actually passed while 855would be best to handle timeouts as if no time had actually passed while
693the program was suspended. This can be achieved by calling C<ev_suspend> 856the program was suspended. This can be achieved by calling C<ev_suspend>
695C<ev_resume> directly afterwards to resume timer processing. 858C<ev_resume> directly afterwards to resume timer processing.
696 859
697Effectively, all C<ev_timer> watchers will be delayed by the time spend 860Effectively, all C<ev_timer> watchers will be delayed by the time spend
698between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers 861between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
699will be rescheduled (that is, they will lose any events that would have 862will be rescheduled (that is, they will lose any events that would have
700occured while suspended). 863occurred while suspended).
701 864
702After calling C<ev_suspend> you B<must not> call I<any> function on the 865After calling C<ev_suspend> you B<must not> call I<any> function on the
703given loop other than C<ev_resume>, and you B<must not> call C<ev_resume> 866given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
704without a previous call to C<ev_suspend>. 867without a previous call to C<ev_suspend>.
705 868
706Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the 869Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
707event loop time (see C<ev_now_update>). 870event loop time (see C<ev_now_update>).
708 871
709=item ev_loop (loop, int flags) 872=item bool ev_run (loop, int flags)
710 873
711Finally, this is it, the event handler. This function usually is called 874Finally, this is it, the event handler. This function usually is called
712after you have initialised all your watchers and you want to start 875after you have initialised all your watchers and you want to start
713handling events. 876handling events. It will ask the operating system for any new events, call
877the watcher callbacks, and then repeat the whole process indefinitely: This
878is why event loops are called I<loops>.
714 879
715If the flags argument is specified as C<0>, it will not return until 880If the flags argument is specified as C<0>, it will keep handling events
716either no event watchers are active anymore or C<ev_unloop> was called. 881until either no event watchers are active anymore or C<ev_break> was
882called.
717 883
884The return value is false if there are no more active watchers (which
885usually means "all jobs done" or "deadlock"), and true in all other cases
886(which usually means " you should call C<ev_run> again").
887
718Please note that an explicit C<ev_unloop> is usually better than 888Please note that an explicit C<ev_break> is usually better than
719relying on all watchers to be stopped when deciding when a program has 889relying on all watchers to be stopped when deciding when a program has
720finished (especially in interactive programs), but having a program 890finished (especially in interactive programs), but having a program
721that automatically loops as long as it has to and no longer by virtue 891that automatically loops as long as it has to and no longer by virtue
722of relying on its watchers stopping correctly, that is truly a thing of 892of relying on its watchers stopping correctly, that is truly a thing of
723beauty. 893beauty.
724 894
895This function is I<mostly> exception-safe - you can break out of a
896C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
897exception and so on. This does not decrement the C<ev_depth> value, nor
898will it clear any outstanding C<EVBREAK_ONE> breaks.
899
725A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle 900A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
726those events and any already outstanding ones, but will not block your 901those events and any already outstanding ones, but will not wait and
727process in case there are no events and will return after one iteration of 902block your process in case there are no events and will return after one
728the loop. 903iteration of the loop. This is sometimes useful to poll and handle new
904events while doing lengthy calculations, to keep the program responsive.
729 905
730A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if 906A flags value of C<EVRUN_ONCE> will look for new events (waiting if
731necessary) and will handle those and any already outstanding ones. It 907necessary) and will handle those and any already outstanding ones. It
732will block your process until at least one new event arrives (which could 908will block your process until at least one new event arrives (which could
733be an event internal to libev itself, so there is no guarantee that a 909be an event internal to libev itself, so there is no guarantee that a
734user-registered callback will be called), and will return after one 910user-registered callback will be called), and will return after one
735iteration of the loop. 911iteration of the loop.
736 912
737This is useful if you are waiting for some external event in conjunction 913This is useful if you are waiting for some external event in conjunction
738with something not expressible using other libev watchers (i.e. "roll your 914with something not expressible using other libev watchers (i.e. "roll your
739own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is 915own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
740usually a better approach for this kind of thing. 916usually a better approach for this kind of thing.
741 917
742Here are the gory details of what C<ev_loop> does: 918Here are the gory details of what C<ev_run> does (this is for your
919understanding, not a guarantee that things will work exactly like this in
920future versions):
743 921
922 - Increment loop depth.
923 - Reset the ev_break status.
744 - Before the first iteration, call any pending watchers. 924 - Before the first iteration, call any pending watchers.
925 LOOP:
745 * If EVFLAG_FORKCHECK was used, check for a fork. 926 - If EVFLAG_FORKCHECK was used, check for a fork.
746 - If a fork was detected (by any means), queue and call all fork watchers. 927 - If a fork was detected (by any means), queue and call all fork watchers.
747 - Queue and call all prepare watchers. 928 - Queue and call all prepare watchers.
929 - If ev_break was called, goto FINISH.
748 - If we have been forked, detach and recreate the kernel state 930 - If we have been forked, detach and recreate the kernel state
749 as to not disturb the other process. 931 as to not disturb the other process.
750 - Update the kernel state with all outstanding changes. 932 - Update the kernel state with all outstanding changes.
751 - Update the "event loop time" (ev_now ()). 933 - Update the "event loop time" (ev_now ()).
752 - Calculate for how long to sleep or block, if at all 934 - Calculate for how long to sleep or block, if at all
753 (active idle watchers, EVLOOP_NONBLOCK or not having 935 (active idle watchers, EVRUN_NOWAIT or not having
754 any active watchers at all will result in not sleeping). 936 any active watchers at all will result in not sleeping).
755 - Sleep if the I/O and timer collect interval say so. 937 - Sleep if the I/O and timer collect interval say so.
938 - Increment loop iteration counter.
756 - Block the process, waiting for any events. 939 - Block the process, waiting for any events.
757 - Queue all outstanding I/O (fd) events. 940 - Queue all outstanding I/O (fd) events.
758 - Update the "event loop time" (ev_now ()), and do time jump adjustments. 941 - Update the "event loop time" (ev_now ()), and do time jump adjustments.
759 - Queue all expired timers. 942 - Queue all expired timers.
760 - Queue all expired periodics. 943 - Queue all expired periodics.
761 - Unless any events are pending now, queue all idle watchers. 944 - Queue all idle watchers with priority higher than that of pending events.
762 - Queue all check watchers. 945 - Queue all check watchers.
763 - Call all queued watchers in reverse order (i.e. check watchers first). 946 - Call all queued watchers in reverse order (i.e. check watchers first).
764 Signals and child watchers are implemented as I/O watchers, and will 947 Signals and child watchers are implemented as I/O watchers, and will
765 be handled here by queueing them when their watcher gets executed. 948 be handled here by queueing them when their watcher gets executed.
766 - If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK 949 - If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
767 were used, or there are no active watchers, return, otherwise 950 were used, or there are no active watchers, goto FINISH, otherwise
768 continue with step *. 951 continue with step LOOP.
952 FINISH:
953 - Reset the ev_break status iff it was EVBREAK_ONE.
954 - Decrement the loop depth.
955 - Return.
769 956
770Example: Queue some jobs and then loop until no events are outstanding 957Example: Queue some jobs and then loop until no events are outstanding
771anymore. 958anymore.
772 959
773 ... queue jobs here, make sure they register event watchers as long 960 ... queue jobs here, make sure they register event watchers as long
774 ... as they still have work to do (even an idle watcher will do..) 961 ... as they still have work to do (even an idle watcher will do..)
775 ev_loop (my_loop, 0); 962 ev_run (my_loop, 0);
776 ... jobs done or somebody called unloop. yeah! 963 ... jobs done or somebody called break. yeah!
777 964
778=item ev_unloop (loop, how) 965=item ev_break (loop, how)
779 966
780Can be used to make a call to C<ev_loop> return early (but only after it 967Can be used to make a call to C<ev_run> return early (but only after it
781has processed all outstanding events). The C<how> argument must be either 968has processed all outstanding events). The C<how> argument must be either
782C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or 969C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
783C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return. 970C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
784 971
785This "unloop state" will be cleared when entering C<ev_loop> again. 972This "break state" will be cleared on the next call to C<ev_run>.
786 973
787It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls. 974It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
975which case it will have no effect.
788 976
789=item ev_ref (loop) 977=item ev_ref (loop)
790 978
791=item ev_unref (loop) 979=item ev_unref (loop)
792 980
793Ref/unref can be used to add or remove a reference count on the event 981Ref/unref can be used to add or remove a reference count on the event
794loop: Every watcher keeps one reference, and as long as the reference 982loop: Every watcher keeps one reference, and as long as the reference
795count is nonzero, C<ev_loop> will not return on its own. 983count is nonzero, C<ev_run> will not return on its own.
796 984
797If you have a watcher you never unregister that should not keep C<ev_loop> 985This is useful when you have a watcher that you never intend to
798from returning, call ev_unref() after starting, and ev_ref() before 986unregister, but that nevertheless should not keep C<ev_run> from
987returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
799stopping it. 988before stopping it.
800 989
801As an example, libev itself uses this for its internal signal pipe: It 990As an example, libev itself uses this for its internal signal pipe: It
802is not visible to the libev user and should not keep C<ev_loop> from 991is not visible to the libev user and should not keep C<ev_run> from
803exiting if no event watchers registered by it are active. It is also an 992exiting if no event watchers registered by it are active. It is also an
804excellent way to do this for generic recurring timers or from within 993excellent way to do this for generic recurring timers or from within
805third-party libraries. Just remember to I<unref after start> and I<ref 994third-party libraries. Just remember to I<unref after start> and I<ref
806before stop> (but only if the watcher wasn't active before, or was active 995before stop> (but only if the watcher wasn't active before, or was active
807before, respectively. Note also that libev might stop watchers itself 996before, respectively. Note also that libev might stop watchers itself
808(e.g. non-repeating timers) in which case you have to C<ev_ref> 997(e.g. non-repeating timers) in which case you have to C<ev_ref>
809in the callback). 998in the callback).
810 999
811Example: Create a signal watcher, but keep it from keeping C<ev_loop> 1000Example: Create a signal watcher, but keep it from keeping C<ev_run>
812running when nothing else is active. 1001running when nothing else is active.
813 1002
814 ev_signal exitsig; 1003 ev_signal exitsig;
815 ev_signal_init (&exitsig, sig_cb, SIGINT); 1004 ev_signal_init (&exitsig, sig_cb, SIGINT);
816 ev_signal_start (loop, &exitsig); 1005 ev_signal_start (loop, &exitsig);
817 evf_unref (loop); 1006 ev_unref (loop);
818 1007
819Example: For some weird reason, unregister the above signal handler again. 1008Example: For some weird reason, unregister the above signal handler again.
820 1009
821 ev_ref (loop); 1010 ev_ref (loop);
822 ev_signal_stop (loop, &exitsig); 1011 ev_signal_stop (loop, &exitsig);
842overhead for the actual polling but can deliver many events at once. 1031overhead for the actual polling but can deliver many events at once.
843 1032
844By setting a higher I<io collect interval> you allow libev to spend more 1033By setting a higher I<io collect interval> you allow libev to spend more
845time collecting I/O events, so you can handle more events per iteration, 1034time collecting I/O events, so you can handle more events per iteration,
846at the cost of increasing latency. Timeouts (both C<ev_periodic> and 1035at the cost of increasing latency. Timeouts (both C<ev_periodic> and
847C<ev_timer>) will be not affected. Setting this to a non-null value will 1036C<ev_timer>) will not be affected. Setting this to a non-null value will
848introduce an additional C<ev_sleep ()> call into most loop iterations. The 1037introduce an additional C<ev_sleep ()> call into most loop iterations. The
849sleep time ensures that libev will not poll for I/O events more often then 1038sleep time ensures that libev will not poll for I/O events more often then
850once per this interval, on average. 1039once per this interval, on average (as long as the host time resolution is
1040good enough).
851 1041
852Likewise, by setting a higher I<timeout collect interval> you allow libev 1042Likewise, by setting a higher I<timeout collect interval> you allow libev
853to spend more time collecting timeouts, at the expense of increased 1043to spend more time collecting timeouts, at the expense of increased
854latency/jitter/inexactness (the watcher callback will be called 1044latency/jitter/inexactness (the watcher callback will be called
855later). C<ev_io> watchers will not be affected. Setting this to a non-null 1045later). C<ev_io> watchers will not be affected. Setting this to a non-null
861usually doesn't make much sense to set it to a lower value than C<0.01>, 1051usually doesn't make much sense to set it to a lower value than C<0.01>,
862as this approaches the timing granularity of most systems. Note that if 1052as this approaches the timing granularity of most systems. Note that if
863you do transactions with the outside world and you can't increase the 1053you do transactions with the outside world and you can't increase the
864parallelity, then this setting will limit your transaction rate (if you 1054parallelity, then this setting will limit your transaction rate (if you
865need to poll once per transaction and the I/O collect interval is 0.01, 1055need to poll once per transaction and the I/O collect interval is 0.01,
866then you can't do more than 100 transations per second). 1056then you can't do more than 100 transactions per second).
867 1057
868Setting the I<timeout collect interval> can improve the opportunity for 1058Setting the I<timeout collect interval> can improve the opportunity for
869saving power, as the program will "bundle" timer callback invocations that 1059saving power, as the program will "bundle" timer callback invocations that
870are "near" in time together, by delaying some, thus reducing the number of 1060are "near" in time together, by delaying some, thus reducing the number of
871times the process sleeps and wakes up again. Another useful technique to 1061times the process sleeps and wakes up again. Another useful technique to
879 ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01); 1069 ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
880 1070
881=item ev_invoke_pending (loop) 1071=item ev_invoke_pending (loop)
882 1072
883This call will simply invoke all pending watchers while resetting their 1073This call will simply invoke all pending watchers while resetting their
884pending state. Normally, C<ev_loop> does this automatically when required, 1074pending state. Normally, C<ev_run> does this automatically when required,
885but when overriding the invoke callback this call comes handy. 1075but when overriding the invoke callback this call comes handy. This
1076function can be invoked from a watcher - this can be useful for example
1077when you want to do some lengthy calculation and want to pass further
1078event handling to another thread (you still have to make sure only one
1079thread executes within C<ev_invoke_pending> or C<ev_run> of course).
886 1080
887=item int ev_pending_count (loop) 1081=item int ev_pending_count (loop)
888 1082
889Returns the number of pending watchers - zero indicates that no watchers 1083Returns the number of pending watchers - zero indicates that no watchers
890are pending. 1084are pending.
891 1085
892=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P)) 1086=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
893 1087
894This overrides the invoke pending functionality of the loop: Instead of 1088This overrides the invoke pending functionality of the loop: Instead of
895invoking all pending watchers when there are any, C<ev_loop> will call 1089invoking all pending watchers when there are any, C<ev_run> will call
896this callback instead. This is useful, for example, when you want to 1090this callback instead. This is useful, for example, when you want to
897invoke the actual watchers inside another context (another thread etc.). 1091invoke the actual watchers inside another context (another thread etc.).
898 1092
899If you want to reset the callback, use C<ev_invoke_pending> as new 1093If you want to reset the callback, use C<ev_invoke_pending> as new
900callback. 1094callback.
901 1095
902=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P)) 1096=item ev_set_loop_release_cb (loop, void (*release)(EV_P) throw (), void (*acquire)(EV_P) throw ())
903 1097
904Sometimes you want to share the same loop between multiple threads. This 1098Sometimes you want to share the same loop between multiple threads. This
905can be done relatively simply by putting mutex_lock/unlock calls around 1099can be done relatively simply by putting mutex_lock/unlock calls around
906each call to a libev function. 1100each call to a libev function.
907 1101
908However, C<ev_loop> can run an indefinite time, so it is not feasible to 1102However, C<ev_run> can run an indefinite time, so it is not feasible
909wait for it to return. One way around this is to wake up the loop via 1103to wait for it to return. One way around this is to wake up the event
910C<ev_unloop> and C<av_async_send>, another way is to set these I<release> 1104loop via C<ev_break> and C<ev_async_send>, another way is to set these
911and I<acquire> callbacks on the loop. 1105I<release> and I<acquire> callbacks on the loop.
912 1106
913When set, then C<release> will be called just before the thread is 1107When set, then C<release> will be called just before the thread is
914suspended waiting for new events, and C<acquire> is called just 1108suspended waiting for new events, and C<acquire> is called just
915afterwards. 1109afterwards.
916 1110
919 1113
920While event loop modifications are allowed between invocations of 1114While event loop modifications are allowed between invocations of
921C<release> and C<acquire> (that's their only purpose after all), no 1115C<release> and C<acquire> (that's their only purpose after all), no
922modifications done will affect the event loop, i.e. adding watchers will 1116modifications done will affect the event loop, i.e. adding watchers will
923have no effect on the set of file descriptors being watched, or the time 1117have no effect on the set of file descriptors being watched, or the time
924waited. Use an C<ev_async> watcher to wake up C<ev_loop> when you want it 1118waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
925to take note of any changes you made. 1119to take note of any changes you made.
926 1120
927In theory, threads executing C<ev_loop> will be async-cancel safe between 1121In theory, threads executing C<ev_run> will be async-cancel safe between
928invocations of C<release> and C<acquire>. 1122invocations of C<release> and C<acquire>.
929 1123
930See also the locking example in the C<THREADS> section later in this 1124See also the locking example in the C<THREADS> section later in this
931document. 1125document.
932 1126
933=item ev_set_userdata (loop, void *data) 1127=item ev_set_userdata (loop, void *data)
934 1128
935=item ev_userdata (loop) 1129=item void *ev_userdata (loop)
936 1130
937Set and retrieve a single C<void *> associated with a loop. When 1131Set and retrieve a single C<void *> associated with a loop. When
938C<ev_set_userdata> has never been called, then C<ev_userdata> returns 1132C<ev_set_userdata> has never been called, then C<ev_userdata> returns
939C<0.> 1133C<0>.
940 1134
941These two functions can be used to associate arbitrary data with a loop, 1135These two functions can be used to associate arbitrary data with a loop,
942and are intended solely for the C<invoke_pending_cb>, C<release> and 1136and are intended solely for the C<invoke_pending_cb>, C<release> and
943C<acquire> callbacks described above, but of course can be (ab-)used for 1137C<acquire> callbacks described above, but of course can be (ab-)used for
944any other purpose as well. 1138any other purpose as well.
945 1139
946=item ev_loop_verify (loop) 1140=item ev_verify (loop)
947 1141
948This function only does something when C<EV_VERIFY> support has been 1142This function only does something when C<EV_VERIFY> support has been
949compiled in, which is the default for non-minimal builds. It tries to go 1143compiled in, which is the default for non-minimal builds. It tries to go
950through all internal structures and checks them for validity. If anything 1144through all internal structures and checks them for validity. If anything
951is found to be inconsistent, it will print an error message to standard 1145is found to be inconsistent, it will print an error message to standard
962 1156
963In the following description, uppercase C<TYPE> in names stands for the 1157In the following description, uppercase C<TYPE> in names stands for the
964watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer 1158watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
965watchers and C<ev_io_start> for I/O watchers. 1159watchers and C<ev_io_start> for I/O watchers.
966 1160
967A watcher is a structure that you create and register to record your 1161A watcher is an opaque structure that you allocate and register to record
968interest in some event. For instance, if you want to wait for STDIN to 1162your interest in some event. To make a concrete example, imagine you want
969become readable, you would create an C<ev_io> watcher for that: 1163to wait for STDIN to become readable, you would create an C<ev_io> watcher
1164for that:
970 1165
971 static void my_cb (struct ev_loop *loop, ev_io *w, int revents) 1166 static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
972 { 1167 {
973 ev_io_stop (w); 1168 ev_io_stop (w);
974 ev_unloop (loop, EVUNLOOP_ALL); 1169 ev_break (loop, EVBREAK_ALL);
975 } 1170 }
976 1171
977 struct ev_loop *loop = ev_default_loop (0); 1172 struct ev_loop *loop = ev_default_loop (0);
978 1173
979 ev_io stdin_watcher; 1174 ev_io stdin_watcher;
980 1175
981 ev_init (&stdin_watcher, my_cb); 1176 ev_init (&stdin_watcher, my_cb);
982 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ); 1177 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
983 ev_io_start (loop, &stdin_watcher); 1178 ev_io_start (loop, &stdin_watcher);
984 1179
985 ev_loop (loop, 0); 1180 ev_run (loop, 0);
986 1181
987As you can see, you are responsible for allocating the memory for your 1182As you can see, you are responsible for allocating the memory for your
988watcher structures (and it is I<usually> a bad idea to do this on the 1183watcher structures (and it is I<usually> a bad idea to do this on the
989stack). 1184stack).
990 1185
991Each watcher has an associated watcher structure (called C<struct ev_TYPE> 1186Each watcher has an associated watcher structure (called C<struct ev_TYPE>
992or simply C<ev_TYPE>, as typedefs are provided for all watcher structs). 1187or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
993 1188
994Each watcher structure must be initialised by a call to C<ev_init 1189Each watcher structure must be initialised by a call to C<ev_init (watcher
995(watcher *, callback)>, which expects a callback to be provided. This 1190*, callback)>, which expects a callback to be provided. This callback is
996callback gets invoked each time the event occurs (or, in the case of I/O 1191invoked each time the event occurs (or, in the case of I/O watchers, each
997watchers, each time the event loop detects that the file descriptor given 1192time the event loop detects that the file descriptor given is readable
998is readable and/or writable). 1193and/or writable).
999 1194
1000Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >> 1195Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
1001macro to configure it, with arguments specific to the watcher type. There 1196macro to configure it, with arguments specific to the watcher type. There
1002is also a macro to combine initialisation and setting in one call: C<< 1197is also a macro to combine initialisation and setting in one call: C<<
1003ev_TYPE_init (watcher *, callback, ...) >>. 1198ev_TYPE_init (watcher *, callback, ...) >>.
1026=item C<EV_WRITE> 1221=item C<EV_WRITE>
1027 1222
1028The file descriptor in the C<ev_io> watcher has become readable and/or 1223The file descriptor in the C<ev_io> watcher has become readable and/or
1029writable. 1224writable.
1030 1225
1031=item C<EV_TIMEOUT> 1226=item C<EV_TIMER>
1032 1227
1033The C<ev_timer> watcher has timed out. 1228The C<ev_timer> watcher has timed out.
1034 1229
1035=item C<EV_PERIODIC> 1230=item C<EV_PERIODIC>
1036 1231
1054 1249
1055=item C<EV_PREPARE> 1250=item C<EV_PREPARE>
1056 1251
1057=item C<EV_CHECK> 1252=item C<EV_CHECK>
1058 1253
1059All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts 1254All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts to
1060to gather new events, and all C<ev_check> watchers are invoked just after 1255gather new events, and all C<ev_check> watchers are queued (not invoked)
1061C<ev_loop> has gathered them, but before it invokes any callbacks for any 1256just after C<ev_run> has gathered them, but before it queues any callbacks
1257for any received events. That means C<ev_prepare> watchers are the last
1258watchers invoked before the event loop sleeps or polls for new events, and
1259C<ev_check> watchers will be invoked before any other watchers of the same
1260or lower priority within an event loop iteration.
1261
1062received events. Callbacks of both watcher types can start and stop as 1262Callbacks of both watcher types can start and stop as many watchers as
1063many watchers as they want, and all of them will be taken into account 1263they want, and all of them will be taken into account (for example, a
1064(for example, a C<ev_prepare> watcher might start an idle watcher to keep 1264C<ev_prepare> watcher might start an idle watcher to keep C<ev_run> from
1065C<ev_loop> from blocking). 1265blocking).
1066 1266
1067=item C<EV_EMBED> 1267=item C<EV_EMBED>
1068 1268
1069The embedded event loop specified in the C<ev_embed> watcher needs attention. 1269The embedded event loop specified in the C<ev_embed> watcher needs attention.
1070 1270
1071=item C<EV_FORK> 1271=item C<EV_FORK>
1072 1272
1073The event loop has been resumed in the child process after fork (see 1273The event loop has been resumed in the child process after fork (see
1074C<ev_fork>). 1274C<ev_fork>).
1275
1276=item C<EV_CLEANUP>
1277
1278The event loop is about to be destroyed (see C<ev_cleanup>).
1075 1279
1076=item C<EV_ASYNC> 1280=item C<EV_ASYNC>
1077 1281
1078The given async watcher has been asynchronously notified (see C<ev_async>). 1282The given async watcher has been asynchronously notified (see C<ev_async>).
1079 1283
1189 1393
1190=item callback ev_cb (ev_TYPE *watcher) 1394=item callback ev_cb (ev_TYPE *watcher)
1191 1395
1192Returns the callback currently set on the watcher. 1396Returns the callback currently set on the watcher.
1193 1397
1194=item ev_cb_set (ev_TYPE *watcher, callback) 1398=item ev_set_cb (ev_TYPE *watcher, callback)
1195 1399
1196Change the callback. You can change the callback at virtually any time 1400Change the callback. You can change the callback at virtually any time
1197(modulo threads). 1401(modulo threads).
1198 1402
1199=item ev_set_priority (ev_TYPE *watcher, int priority) 1403=item ev_set_priority (ev_TYPE *watcher, int priority)
1217or might not have been clamped to the valid range. 1421or might not have been clamped to the valid range.
1218 1422
1219The default priority used by watchers when no priority has been set is 1423The default priority used by watchers when no priority has been set is
1220always C<0>, which is supposed to not be too high and not be too low :). 1424always C<0>, which is supposed to not be too high and not be too low :).
1221 1425
1222See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of 1426See L</WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
1223priorities. 1427priorities.
1224 1428
1225=item ev_invoke (loop, ev_TYPE *watcher, int revents) 1429=item ev_invoke (loop, ev_TYPE *watcher, int revents)
1226 1430
1227Invoke the C<watcher> with the given C<loop> and C<revents>. Neither 1431Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
1252See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related 1456See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
1253functions that do not need a watcher. 1457functions that do not need a watcher.
1254 1458
1255=back 1459=back
1256 1460
1461See also the L</ASSOCIATING CUSTOM DATA WITH A WATCHER> and L</BUILDING YOUR
1462OWN COMPOSITE WATCHERS> idioms.
1257 1463
1258=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER 1464=head2 WATCHER STATES
1259 1465
1260Each watcher has, by default, a member C<void *data> that you can change 1466There are various watcher states mentioned throughout this manual -
1261and read at any time: libev will completely ignore it. This can be used 1467active, pending and so on. In this section these states and the rules to
1262to associate arbitrary data with your watcher. If you need more data and 1468transition between them will be described in more detail - and while these
1263don't want to allocate memory and store a pointer to it in that data 1469rules might look complicated, they usually do "the right thing".
1264member, you can also "subclass" the watcher type and provide your own
1265data:
1266 1470
1267 struct my_io 1471=over 4
1268 {
1269 ev_io io;
1270 int otherfd;
1271 void *somedata;
1272 struct whatever *mostinteresting;
1273 };
1274 1472
1275 ... 1473=item initialised
1276 struct my_io w;
1277 ev_io_init (&w.io, my_cb, fd, EV_READ);
1278 1474
1279And since your callback will be called with a pointer to the watcher, you 1475Before a watcher can be registered with the event loop it has to be
1280can cast it back to your own type: 1476initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
1477C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
1281 1478
1282 static void my_cb (struct ev_loop *loop, ev_io *w_, int revents) 1479In this state it is simply some block of memory that is suitable for
1283 { 1480use in an event loop. It can be moved around, freed, reused etc. at
1284 struct my_io *w = (struct my_io *)w_; 1481will - as long as you either keep the memory contents intact, or call
1285 ... 1482C<ev_TYPE_init> again.
1286 }
1287 1483
1288More interesting and less C-conformant ways of casting your callback type 1484=item started/running/active
1289instead have been omitted.
1290 1485
1291Another common scenario is to use some data structure with multiple 1486Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
1292embedded watchers: 1487property of the event loop, and is actively waiting for events. While in
1488this state it cannot be accessed (except in a few documented ways), moved,
1489freed or anything else - the only legal thing is to keep a pointer to it,
1490and call libev functions on it that are documented to work on active watchers.
1293 1491
1294 struct my_biggy 1492=item pending
1295 {
1296 int some_data;
1297 ev_timer t1;
1298 ev_timer t2;
1299 }
1300 1493
1301In this case getting the pointer to C<my_biggy> is a bit more 1494If a watcher is active and libev determines that an event it is interested
1302complicated: Either you store the address of your C<my_biggy> struct 1495in has occurred (such as a timer expiring), it will become pending. It will
1303in the C<data> member of the watcher (for woozies), or you need to use 1496stay in this pending state until either it is stopped or its callback is
1304some pointer arithmetic using C<offsetof> inside your watchers (for real 1497about to be invoked, so it is not normally pending inside the watcher
1305programmers): 1498callback.
1306 1499
1307 #include <stddef.h> 1500The watcher might or might not be active while it is pending (for example,
1501an expired non-repeating timer can be pending but no longer active). If it
1502is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
1503but it is still property of the event loop at this time, so cannot be
1504moved, freed or reused. And if it is active the rules described in the
1505previous item still apply.
1308 1506
1309 static void 1507It is also possible to feed an event on a watcher that is not active (e.g.
1310 t1_cb (EV_P_ ev_timer *w, int revents) 1508via C<ev_feed_event>), in which case it becomes pending without being
1311 { 1509active.
1312 struct my_biggy big = (struct my_biggy *)
1313 (((char *)w) - offsetof (struct my_biggy, t1));
1314 }
1315 1510
1316 static void 1511=item stopped
1317 t2_cb (EV_P_ ev_timer *w, int revents) 1512
1318 { 1513A watcher can be stopped implicitly by libev (in which case it might still
1319 struct my_biggy big = (struct my_biggy *) 1514be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
1320 (((char *)w) - offsetof (struct my_biggy, t2)); 1515latter will clear any pending state the watcher might be in, regardless
1321 } 1516of whether it was active or not, so stopping a watcher explicitly before
1517freeing it is often a good idea.
1518
1519While stopped (and not pending) the watcher is essentially in the
1520initialised state, that is, it can be reused, moved, modified in any way
1521you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
1522it again).
1523
1524=back
1322 1525
1323=head2 WATCHER PRIORITY MODELS 1526=head2 WATCHER PRIORITY MODELS
1324 1527
1325Many event loops support I<watcher priorities>, which are usually small 1528Many event loops support I<watcher priorities>, which are usually small
1326integers that influence the ordering of event callback invocation 1529integers that influence the ordering of event callback invocation
1369 1572
1370For example, to emulate how many other event libraries handle priorities, 1573For example, to emulate how many other event libraries handle priorities,
1371you can associate an C<ev_idle> watcher to each such watcher, and in 1574you can associate an C<ev_idle> watcher to each such watcher, and in
1372the normal watcher callback, you just start the idle watcher. The real 1575the normal watcher callback, you just start the idle watcher. The real
1373processing is done in the idle watcher callback. This causes libev to 1576processing is done in the idle watcher callback. This causes libev to
1374continously poll and process kernel event data for the watcher, but when 1577continuously poll and process kernel event data for the watcher, but when
1375the lock-out case is known to be rare (which in turn is rare :), this is 1578the lock-out case is known to be rare (which in turn is rare :), this is
1376workable. 1579workable.
1377 1580
1378Usually, however, the lock-out model implemented that way will perform 1581Usually, however, the lock-out model implemented that way will perform
1379miserably under the type of load it was designed to handle. In that case, 1582miserably under the type of load it was designed to handle. In that case,
1393 { 1596 {
1394 // stop the I/O watcher, we received the event, but 1597 // stop the I/O watcher, we received the event, but
1395 // are not yet ready to handle it. 1598 // are not yet ready to handle it.
1396 ev_io_stop (EV_A_ w); 1599 ev_io_stop (EV_A_ w);
1397 1600
1398 // start the idle watcher to ahndle the actual event. 1601 // start the idle watcher to handle the actual event.
1399 // it will not be executed as long as other watchers 1602 // it will not be executed as long as other watchers
1400 // with the default priority are receiving events. 1603 // with the default priority are receiving events.
1401 ev_idle_start (EV_A_ &idle); 1604 ev_idle_start (EV_A_ &idle);
1402 } 1605 }
1403 1606
1453In general you can register as many read and/or write event watchers per 1656In general you can register as many read and/or write event watchers per
1454fd as you want (as long as you don't confuse yourself). Setting all file 1657fd as you want (as long as you don't confuse yourself). Setting all file
1455descriptors to non-blocking mode is also usually a good idea (but not 1658descriptors to non-blocking mode is also usually a good idea (but not
1456required if you know what you are doing). 1659required if you know what you are doing).
1457 1660
1458If you cannot use non-blocking mode, then force the use of a
1459known-to-be-good backend (at the time of this writing, this includes only
1460C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
1461descriptors for which non-blocking operation makes no sense (such as
1462files) - libev doesn't guarentee any specific behaviour in that case.
1463
1464Another thing you have to watch out for is that it is quite easy to 1661Another thing you have to watch out for is that it is quite easy to
1465receive "spurious" readiness notifications, that is your callback might 1662receive "spurious" readiness notifications, that is, your callback might
1466be called with C<EV_READ> but a subsequent C<read>(2) will actually block 1663be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1467because there is no data. Not only are some backends known to create a 1664because there is no data. It is very easy to get into this situation even
1468lot of those (for example Solaris ports), it is very easy to get into 1665with a relatively standard program structure. Thus it is best to always
1469this situation even with a relatively standard program structure. Thus 1666use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
1470it is best to always use non-blocking I/O: An extra C<read>(2) returning
1471C<EAGAIN> is far preferable to a program hanging until some data arrives. 1667preferable to a program hanging until some data arrives.
1472 1668
1473If you cannot run the fd in non-blocking mode (for example you should 1669If you cannot run the fd in non-blocking mode (for example you should
1474not play around with an Xlib connection), then you have to separately 1670not play around with an Xlib connection), then you have to separately
1475re-test whether a file descriptor is really ready with a known-to-be good 1671re-test whether a file descriptor is really ready with a known-to-be good
1476interface such as poll (fortunately in our Xlib example, Xlib already 1672interface such as poll (fortunately in the case of Xlib, it already does
1477does this on its own, so its quite safe to use). Some people additionally 1673this on its own, so its quite safe to use). Some people additionally
1478use C<SIGALRM> and an interval timer, just to be sure you won't block 1674use C<SIGALRM> and an interval timer, just to be sure you won't block
1479indefinitely. 1675indefinitely.
1480 1676
1481But really, best use non-blocking mode. 1677But really, best use non-blocking mode.
1482 1678
1483=head3 The special problem of disappearing file descriptors 1679=head3 The special problem of disappearing file descriptors
1484 1680
1485Some backends (e.g. kqueue, epoll) need to be told about closing a file 1681Some backends (e.g. kqueue, epoll, linuxaio) need to be told about closing
1486descriptor (either due to calling C<close> explicitly or any other means, 1682a file descriptor (either due to calling C<close> explicitly or any other
1487such as C<dup2>). The reason is that you register interest in some file 1683means, such as C<dup2>). The reason is that you register interest in some
1488descriptor, but when it goes away, the operating system will silently drop 1684file descriptor, but when it goes away, the operating system will silently
1489this interest. If another file descriptor with the same number then is 1685drop this interest. If another file descriptor with the same number then
1490registered with libev, there is no efficient way to see that this is, in 1686is registered with libev, there is no efficient way to see that this is,
1491fact, a different file descriptor. 1687in fact, a different file descriptor.
1492 1688
1493To avoid having to explicitly tell libev about such cases, libev follows 1689To avoid having to explicitly tell libev about such cases, libev follows
1494the following policy: Each time C<ev_io_set> is being called, libev 1690the following policy: Each time C<ev_io_set> is being called, libev
1495will assume that this is potentially a new file descriptor, otherwise 1691will assume that this is potentially a new file descriptor, otherwise
1496it is assumed that the file descriptor stays the same. That means that 1692it is assumed that the file descriptor stays the same. That means that
1510 1706
1511There is no workaround possible except not registering events 1707There is no workaround possible except not registering events
1512for potentially C<dup ()>'ed file descriptors, or to resort to 1708for potentially C<dup ()>'ed file descriptors, or to resort to
1513C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>. 1709C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1514 1710
1711=head3 The special problem of files
1712
1713Many people try to use C<select> (or libev) on file descriptors
1714representing files, and expect it to become ready when their program
1715doesn't block on disk accesses (which can take a long time on their own).
1716
1717However, this cannot ever work in the "expected" way - you get a readiness
1718notification as soon as the kernel knows whether and how much data is
1719there, and in the case of open files, that's always the case, so you
1720always get a readiness notification instantly, and your read (or possibly
1721write) will still block on the disk I/O.
1722
1723Another way to view it is that in the case of sockets, pipes, character
1724devices and so on, there is another party (the sender) that delivers data
1725on its own, but in the case of files, there is no such thing: the disk
1726will not send data on its own, simply because it doesn't know what you
1727wish to read - you would first have to request some data.
1728
1729Since files are typically not-so-well supported by advanced notification
1730mechanism, libev tries hard to emulate POSIX behaviour with respect
1731to files, even though you should not use it. The reason for this is
1732convenience: sometimes you want to watch STDIN or STDOUT, which is
1733usually a tty, often a pipe, but also sometimes files or special devices
1734(for example, C<epoll> on Linux works with F</dev/random> but not with
1735F</dev/urandom>), and even though the file might better be served with
1736asynchronous I/O instead of with non-blocking I/O, it is still useful when
1737it "just works" instead of freezing.
1738
1739So avoid file descriptors pointing to files when you know it (e.g. use
1740libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
1741when you rarely read from a file instead of from a socket, and want to
1742reuse the same code path.
1743
1515=head3 The special problem of fork 1744=head3 The special problem of fork
1516 1745
1517Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit 1746Some backends (epoll, kqueue, probably linuxaio) do not support C<fork ()>
1518useless behaviour. Libev fully supports fork, but needs to be told about 1747at all or exhibit useless behaviour. Libev fully supports fork, but needs
1519it in the child. 1748to be told about it in the child if you want to continue to use it in the
1749child.
1520 1750
1521To support fork in your programs, you either have to call 1751To support fork in your child processes, you have to call C<ev_loop_fork
1522C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child, 1752()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
1523enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or 1753C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1524C<EVBACKEND_POLL>.
1525 1754
1526=head3 The special problem of SIGPIPE 1755=head3 The special problem of SIGPIPE
1527 1756
1528While not really specific to libev, it is easy to forget about C<SIGPIPE>: 1757While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1529when writing to a pipe whose other end has been closed, your program gets 1758when writing to a pipe whose other end has been closed, your program gets
1532 1761
1533So when you encounter spurious, unexplained daemon exits, make sure you 1762So when you encounter spurious, unexplained daemon exits, make sure you
1534ignore SIGPIPE (and maybe make sure you log the exit status of your daemon 1763ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
1535somewhere, as that would have given you a big clue). 1764somewhere, as that would have given you a big clue).
1536 1765
1766=head3 The special problem of accept()ing when you can't
1767
1768Many implementations of the POSIX C<accept> function (for example,
1769found in post-2004 Linux) have the peculiar behaviour of not removing a
1770connection from the pending queue in all error cases.
1771
1772For example, larger servers often run out of file descriptors (because
1773of resource limits), causing C<accept> to fail with C<ENFILE> but not
1774rejecting the connection, leading to libev signalling readiness on
1775the next iteration again (the connection still exists after all), and
1776typically causing the program to loop at 100% CPU usage.
1777
1778Unfortunately, the set of errors that cause this issue differs between
1779operating systems, there is usually little the app can do to remedy the
1780situation, and no known thread-safe method of removing the connection to
1781cope with overload is known (to me).
1782
1783One of the easiest ways to handle this situation is to just ignore it
1784- when the program encounters an overload, it will just loop until the
1785situation is over. While this is a form of busy waiting, no OS offers an
1786event-based way to handle this situation, so it's the best one can do.
1787
1788A better way to handle the situation is to log any errors other than
1789C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
1790messages, and continue as usual, which at least gives the user an idea of
1791what could be wrong ("raise the ulimit!"). For extra points one could stop
1792the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
1793usage.
1794
1795If your program is single-threaded, then you could also keep a dummy file
1796descriptor for overload situations (e.g. by opening F</dev/null>), and
1797when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
1798close that fd, and create a new dummy fd. This will gracefully refuse
1799clients under typical overload conditions.
1800
1801The last way to handle it is to simply log the error and C<exit>, as
1802is often done with C<malloc> failures, but this results in an easy
1803opportunity for a DoS attack.
1537 1804
1538=head3 Watcher-Specific Functions 1805=head3 Watcher-Specific Functions
1539 1806
1540=over 4 1807=over 4
1541 1808
1573 ... 1840 ...
1574 struct ev_loop *loop = ev_default_init (0); 1841 struct ev_loop *loop = ev_default_init (0);
1575 ev_io stdin_readable; 1842 ev_io stdin_readable;
1576 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ); 1843 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1577 ev_io_start (loop, &stdin_readable); 1844 ev_io_start (loop, &stdin_readable);
1578 ev_loop (loop, 0); 1845 ev_run (loop, 0);
1579 1846
1580 1847
1581=head2 C<ev_timer> - relative and optionally repeating timeouts 1848=head2 C<ev_timer> - relative and optionally repeating timeouts
1582 1849
1583Timer watchers are simple relative timers that generate an event after a 1850Timer watchers are simple relative timers that generate an event after a
1589detecting time jumps is hard, and some inaccuracies are unavoidable (the 1856detecting time jumps is hard, and some inaccuracies are unavoidable (the
1590monotonic clock option helps a lot here). 1857monotonic clock option helps a lot here).
1591 1858
1592The callback is guaranteed to be invoked only I<after> its timeout has 1859The callback is guaranteed to be invoked only I<after> its timeout has
1593passed (not I<at>, so on systems with very low-resolution clocks this 1860passed (not I<at>, so on systems with very low-resolution clocks this
1594might introduce a small delay). If multiple timers become ready during the 1861might introduce a small delay, see "the special problem of being too
1862early", below). If multiple timers become ready during the same loop
1595same loop iteration then the ones with earlier time-out values are invoked 1863iteration then the ones with earlier time-out values are invoked before
1596before ones of the same priority with later time-out values (but this is 1864ones of the same priority with later time-out values (but this is no
1597no longer true when a callback calls C<ev_loop> recursively). 1865longer true when a callback calls C<ev_run> recursively).
1598 1866
1599=head3 Be smart about timeouts 1867=head3 Be smart about timeouts
1600 1868
1601Many real-world problems involve some kind of timeout, usually for error 1869Many real-world problems involve some kind of timeout, usually for error
1602recovery. A typical example is an HTTP request - if the other side hangs, 1870recovery. A typical example is an HTTP request - if the other side hangs,
1677 1945
1678In this case, it would be more efficient to leave the C<ev_timer> alone, 1946In this case, it would be more efficient to leave the C<ev_timer> alone,
1679but remember the time of last activity, and check for a real timeout only 1947but remember the time of last activity, and check for a real timeout only
1680within the callback: 1948within the callback:
1681 1949
1950 ev_tstamp timeout = 60.;
1682 ev_tstamp last_activity; // time of last activity 1951 ev_tstamp last_activity; // time of last activity
1952 ev_timer timer;
1683 1953
1684 static void 1954 static void
1685 callback (EV_P_ ev_timer *w, int revents) 1955 callback (EV_P_ ev_timer *w, int revents)
1686 { 1956 {
1687 ev_tstamp now = ev_now (EV_A); 1957 // calculate when the timeout would happen
1688 ev_tstamp timeout = last_activity + 60.; 1958 ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
1689 1959
1690 // if last_activity + 60. is older than now, we did time out 1960 // if negative, it means we the timeout already occurred
1691 if (timeout < now) 1961 if (after < 0.)
1692 { 1962 {
1693 // timeout occured, take action 1963 // timeout occurred, take action
1694 } 1964 }
1695 else 1965 else
1696 { 1966 {
1697 // callback was invoked, but there was some activity, re-arm 1967 // callback was invoked, but there was some recent
1698 // the watcher to fire in last_activity + 60, which is 1968 // activity. simply restart the timer to time out
1699 // guaranteed to be in the future, so "again" is positive: 1969 // after "after" seconds, which is the earliest time
1700 w->repeat = timeout - now; 1970 // the timeout can occur.
1971 ev_timer_set (w, after, 0.);
1701 ev_timer_again (EV_A_ w); 1972 ev_timer_start (EV_A_ w);
1702 } 1973 }
1703 } 1974 }
1704 1975
1705To summarise the callback: first calculate the real timeout (defined 1976To summarise the callback: first calculate in how many seconds the
1706as "60 seconds after the last activity"), then check if that time has 1977timeout will occur (by calculating the absolute time when it would occur,
1707been reached, which means something I<did>, in fact, time out. Otherwise 1978C<last_activity + timeout>, and subtracting the current time, C<ev_now
1708the callback was invoked too early (C<timeout> is in the future), so 1979(EV_A)> from that).
1709re-schedule the timer to fire at that future time, to see if maybe we have
1710a timeout then.
1711 1980
1712Note how C<ev_timer_again> is used, taking advantage of the 1981If this value is negative, then we are already past the timeout, i.e. we
1713C<ev_timer_again> optimisation when the timer is already running. 1982timed out, and need to do whatever is needed in this case.
1983
1984Otherwise, we now the earliest time at which the timeout would trigger,
1985and simply start the timer with this timeout value.
1986
1987In other words, each time the callback is invoked it will check whether
1988the timeout occurred. If not, it will simply reschedule itself to check
1989again at the earliest time it could time out. Rinse. Repeat.
1714 1990
1715This scheme causes more callback invocations (about one every 60 seconds 1991This scheme causes more callback invocations (about one every 60 seconds
1716minus half the average time between activity), but virtually no calls to 1992minus half the average time between activity), but virtually no calls to
1717libev to change the timeout. 1993libev to change the timeout.
1718 1994
1719To start the timer, simply initialise the watcher and set C<last_activity> 1995To start the machinery, simply initialise the watcher and set
1720to the current time (meaning we just have some activity :), then call the 1996C<last_activity> to the current time (meaning there was some activity just
1721callback, which will "do the right thing" and start the timer: 1997now), then call the callback, which will "do the right thing" and start
1998the timer:
1722 1999
2000 last_activity = ev_now (EV_A);
1723 ev_init (timer, callback); 2001 ev_init (&timer, callback);
1724 last_activity = ev_now (loop); 2002 callback (EV_A_ &timer, 0);
1725 callback (loop, timer, EV_TIMEOUT);
1726 2003
1727And when there is some activity, simply store the current time in 2004When there is some activity, simply store the current time in
1728C<last_activity>, no libev calls at all: 2005C<last_activity>, no libev calls at all:
1729 2006
2007 if (activity detected)
1730 last_actiivty = ev_now (loop); 2008 last_activity = ev_now (EV_A);
2009
2010When your timeout value changes, then the timeout can be changed by simply
2011providing a new value, stopping the timer and calling the callback, which
2012will again do the right thing (for example, time out immediately :).
2013
2014 timeout = new_value;
2015 ev_timer_stop (EV_A_ &timer);
2016 callback (EV_A_ &timer, 0);
1731 2017
1732This technique is slightly more complex, but in most cases where the 2018This technique is slightly more complex, but in most cases where the
1733time-out is unlikely to be triggered, much more efficient. 2019time-out is unlikely to be triggered, much more efficient.
1734
1735Changing the timeout is trivial as well (if it isn't hard-coded in the
1736callback :) - just change the timeout and invoke the callback, which will
1737fix things for you.
1738 2020
1739=item 4. Wee, just use a double-linked list for your timeouts. 2021=item 4. Wee, just use a double-linked list for your timeouts.
1740 2022
1741If there is not one request, but many thousands (millions...), all 2023If there is not one request, but many thousands (millions...), all
1742employing some kind of timeout with the same timeout value, then one can 2024employing some kind of timeout with the same timeout value, then one can
1769Method #1 is almost always a bad idea, and buys you nothing. Method #4 is 2051Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
1770rather complicated, but extremely efficient, something that really pays 2052rather complicated, but extremely efficient, something that really pays
1771off after the first million or so of active timers, i.e. it's usually 2053off after the first million or so of active timers, i.e. it's usually
1772overkill :) 2054overkill :)
1773 2055
2056=head3 The special problem of being too early
2057
2058If you ask a timer to call your callback after three seconds, then
2059you expect it to be invoked after three seconds - but of course, this
2060cannot be guaranteed to infinite precision. Less obviously, it cannot be
2061guaranteed to any precision by libev - imagine somebody suspending the
2062process with a STOP signal for a few hours for example.
2063
2064So, libev tries to invoke your callback as soon as possible I<after> the
2065delay has occurred, but cannot guarantee this.
2066
2067A less obvious failure mode is calling your callback too early: many event
2068loops compare timestamps with a "elapsed delay >= requested delay", but
2069this can cause your callback to be invoked much earlier than you would
2070expect.
2071
2072To see why, imagine a system with a clock that only offers full second
2073resolution (think windows if you can't come up with a broken enough OS
2074yourself). If you schedule a one-second timer at the time 500.9, then the
2075event loop will schedule your timeout to elapse at a system time of 500
2076(500.9 truncated to the resolution) + 1, or 501.
2077
2078If an event library looks at the timeout 0.1s later, it will see "501 >=
2079501" and invoke the callback 0.1s after it was started, even though a
2080one-second delay was requested - this is being "too early", despite best
2081intentions.
2082
2083This is the reason why libev will never invoke the callback if the elapsed
2084delay equals the requested delay, but only when the elapsed delay is
2085larger than the requested delay. In the example above, libev would only invoke
2086the callback at system time 502, or 1.1s after the timer was started.
2087
2088So, while libev cannot guarantee that your callback will be invoked
2089exactly when requested, it I<can> and I<does> guarantee that the requested
2090delay has actually elapsed, or in other words, it always errs on the "too
2091late" side of things.
2092
1774=head3 The special problem of time updates 2093=head3 The special problem of time updates
1775 2094
1776Establishing the current time is a costly operation (it usually takes at 2095Establishing the current time is a costly operation (it usually takes
1777least two system calls): EV therefore updates its idea of the current 2096at least one system call): EV therefore updates its idea of the current
1778time only before and after C<ev_loop> collects new events, which causes a 2097time only before and after C<ev_run> collects new events, which causes a
1779growing difference between C<ev_now ()> and C<ev_time ()> when handling 2098growing difference between C<ev_now ()> and C<ev_time ()> when handling
1780lots of events in one iteration. 2099lots of events in one iteration.
1781 2100
1782The relative timeouts are calculated relative to the C<ev_now ()> 2101The relative timeouts are calculated relative to the C<ev_now ()>
1783time. This is usually the right thing as this timestamp refers to the time 2102time. This is usually the right thing as this timestamp refers to the time
1784of the event triggering whatever timeout you are modifying/starting. If 2103of the event triggering whatever timeout you are modifying/starting. If
1785you suspect event processing to be delayed and you I<need> to base the 2104you suspect event processing to be delayed and you I<need> to base the
1786timeout on the current time, use something like this to adjust for this: 2105timeout on the current time, use something like the following to adjust
2106for it:
1787 2107
1788 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.); 2108 ev_timer_set (&timer, after + (ev_time () - ev_now ()), 0.);
1789 2109
1790If the event loop is suspended for a long time, you can also force an 2110If the event loop is suspended for a long time, you can also force an
1791update of the time returned by C<ev_now ()> by calling C<ev_now_update 2111update of the time returned by C<ev_now ()> by calling C<ev_now_update
1792()>. 2112()>, although that will push the event time of all outstanding events
2113further into the future.
2114
2115=head3 The special problem of unsynchronised clocks
2116
2117Modern systems have a variety of clocks - libev itself uses the normal
2118"wall clock" clock and, if available, the monotonic clock (to avoid time
2119jumps).
2120
2121Neither of these clocks is synchronised with each other or any other clock
2122on the system, so C<ev_time ()> might return a considerably different time
2123than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
2124a call to C<gettimeofday> might return a second count that is one higher
2125than a directly following call to C<time>.
2126
2127The moral of this is to only compare libev-related timestamps with
2128C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
2129a second or so.
2130
2131One more problem arises due to this lack of synchronisation: if libev uses
2132the system monotonic clock and you compare timestamps from C<ev_time>
2133or C<ev_now> from when you started your timer and when your callback is
2134invoked, you will find that sometimes the callback is a bit "early".
2135
2136This is because C<ev_timer>s work in real time, not wall clock time, so
2137libev makes sure your callback is not invoked before the delay happened,
2138I<measured according to the real time>, not the system clock.
2139
2140If your timeouts are based on a physical timescale (e.g. "time out this
2141connection after 100 seconds") then this shouldn't bother you as it is
2142exactly the right behaviour.
2143
2144If you want to compare wall clock/system timestamps to your timers, then
2145you need to use C<ev_periodic>s, as these are based on the wall clock
2146time, where your comparisons will always generate correct results.
1793 2147
1794=head3 The special problems of suspended animation 2148=head3 The special problems of suspended animation
1795 2149
1796When you leave the server world it is quite customary to hit machines that 2150When you leave the server world it is quite customary to hit machines that
1797can suspend/hibernate - what happens to the clocks during such a suspend? 2151can suspend/hibernate - what happens to the clocks during such a suspend?
1827 2181
1828=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat) 2182=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1829 2183
1830=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat) 2184=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1831 2185
1832Configure the timer to trigger after C<after> seconds. If C<repeat> 2186Configure the timer to trigger after C<after> seconds (fractional and
1833is C<0.>, then it will automatically be stopped once the timeout is 2187negative values are supported). If C<repeat> is C<0.>, then it will
1834reached. If it is positive, then the timer will automatically be 2188automatically be stopped once the timeout is reached. If it is positive,
1835configured to trigger again C<repeat> seconds later, again, and again, 2189then the timer will automatically be configured to trigger again C<repeat>
1836until stopped manually. 2190seconds later, again, and again, until stopped manually.
1837 2191
1838The timer itself will do a best-effort at avoiding drift, that is, if 2192The timer itself will do a best-effort at avoiding drift, that is, if
1839you configure a timer to trigger every 10 seconds, then it will normally 2193you configure a timer to trigger every 10 seconds, then it will normally
1840trigger at exactly 10 second intervals. If, however, your program cannot 2194trigger at exactly 10 second intervals. If, however, your program cannot
1841keep up with the timer (because it takes longer than those 10 seconds to 2195keep up with the timer (because it takes longer than those 10 seconds to
1842do stuff) the timer will not fire more than once per event loop iteration. 2196do stuff) the timer will not fire more than once per event loop iteration.
1843 2197
1844=item ev_timer_again (loop, ev_timer *) 2198=item ev_timer_again (loop, ev_timer *)
1845 2199
1846This will act as if the timer timed out and restart it again if it is 2200This will act as if the timer timed out, and restarts it again if it is
1847repeating. The exact semantics are: 2201repeating. It basically works like calling C<ev_timer_stop>, updating the
2202timeout to the C<repeat> value and calling C<ev_timer_start>.
1848 2203
2204The exact semantics are as in the following rules, all of which will be
2205applied to the watcher:
2206
2207=over 4
2208
1849If the timer is pending, its pending status is cleared. 2209=item If the timer is pending, the pending status is always cleared.
1850 2210
1851If the timer is started but non-repeating, stop it (as if it timed out). 2211=item If the timer is started but non-repeating, stop it (as if it timed
2212out, without invoking it).
1852 2213
1853If the timer is repeating, either start it if necessary (with the 2214=item If the timer is repeating, make the C<repeat> value the new timeout
1854C<repeat> value), or reset the running timer to the C<repeat> value. 2215and start the timer, if necessary.
1855 2216
2217=back
2218
1856This sounds a bit complicated, see L<Be smart about timeouts>, above, for a 2219This sounds a bit complicated, see L</Be smart about timeouts>, above, for a
1857usage example. 2220usage example.
1858 2221
1859=item ev_tstamp ev_timer_remaining (loop, ev_timer *) 2222=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
1860 2223
1861Returns the remaining time until a timer fires. If the timer is active, 2224Returns the remaining time until a timer fires. If the timer is active,
1862then this time is relative to the current event loop time, otherwise it's 2225then this time is relative to the current event loop time, otherwise it's
1863the timeout value currently configured. 2226the timeout value currently configured.
1864 2227
1865That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns 2228That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
1866C<5>. When the timer is started and one second passes, C<ev_timer_remain> 2229C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
1867will return C<4>. When the timer expires and is restarted, it will return 2230will return C<4>. When the timer expires and is restarted, it will return
1868roughly C<7> (likely slightly less as callback invocation takes some time, 2231roughly C<7> (likely slightly less as callback invocation takes some time,
1869too), and so on. 2232too), and so on.
1870 2233
1871=item ev_tstamp repeat [read-write] 2234=item ev_tstamp repeat [read-write]
1900 } 2263 }
1901 2264
1902 ev_timer mytimer; 2265 ev_timer mytimer;
1903 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */ 2266 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1904 ev_timer_again (&mytimer); /* start timer */ 2267 ev_timer_again (&mytimer); /* start timer */
1905 ev_loop (loop, 0); 2268 ev_run (loop, 0);
1906 2269
1907 // and in some piece of code that gets executed on any "activity": 2270 // and in some piece of code that gets executed on any "activity":
1908 // reset the timeout to start ticking again at 10 seconds 2271 // reset the timeout to start ticking again at 10 seconds
1909 ev_timer_again (&mytimer); 2272 ev_timer_again (&mytimer);
1910 2273
1914Periodic watchers are also timers of a kind, but they are very versatile 2277Periodic watchers are also timers of a kind, but they are very versatile
1915(and unfortunately a bit complex). 2278(and unfortunately a bit complex).
1916 2279
1917Unlike C<ev_timer>, periodic watchers are not based on real time (or 2280Unlike C<ev_timer>, periodic watchers are not based on real time (or
1918relative time, the physical time that passes) but on wall clock time 2281relative time, the physical time that passes) but on wall clock time
1919(absolute time, the thing you can read on your calender or clock). The 2282(absolute time, the thing you can read on your calendar or clock). The
1920difference is that wall clock time can run faster or slower than real 2283difference is that wall clock time can run faster or slower than real
1921time, and time jumps are not uncommon (e.g. when you adjust your 2284time, and time jumps are not uncommon (e.g. when you adjust your
1922wrist-watch). 2285wrist-watch).
1923 2286
1924You can tell a periodic watcher to trigger after some specific point 2287You can tell a periodic watcher to trigger after some specific point
1929C<ev_timer>, which would still trigger roughly 10 seconds after starting 2292C<ev_timer>, which would still trigger roughly 10 seconds after starting
1930it, as it uses a relative timeout). 2293it, as it uses a relative timeout).
1931 2294
1932C<ev_periodic> watchers can also be used to implement vastly more complex 2295C<ev_periodic> watchers can also be used to implement vastly more complex
1933timers, such as triggering an event on each "midnight, local time", or 2296timers, such as triggering an event on each "midnight, local time", or
1934other complicated rules. This cannot be done with C<ev_timer> watchers, as 2297other complicated rules. This cannot easily be done with C<ev_timer>
1935those cannot react to time jumps. 2298watchers, as those cannot react to time jumps.
1936 2299
1937As with timers, the callback is guaranteed to be invoked only when the 2300As with timers, the callback is guaranteed to be invoked only when the
1938point in time where it is supposed to trigger has passed. If multiple 2301point in time where it is supposed to trigger has passed. If multiple
1939timers become ready during the same loop iteration then the ones with 2302timers become ready during the same loop iteration then the ones with
1940earlier time-out values are invoked before ones with later time-out values 2303earlier time-out values are invoked before ones with later time-out values
1941(but this is no longer true when a callback calls C<ev_loop> recursively). 2304(but this is no longer true when a callback calls C<ev_run> recursively).
1942 2305
1943=head3 Watcher-Specific Functions and Data Members 2306=head3 Watcher-Specific Functions and Data Members
1944 2307
1945=over 4 2308=over 4
1946 2309
1981 2344
1982Another way to think about it (for the mathematically inclined) is that 2345Another way to think about it (for the mathematically inclined) is that
1983C<ev_periodic> will try to run the callback in this mode at the next possible 2346C<ev_periodic> will try to run the callback in this mode at the next possible
1984time where C<time = offset (mod interval)>, regardless of any time jumps. 2347time where C<time = offset (mod interval)>, regardless of any time jumps.
1985 2348
1986For numerical stability it is preferable that the C<offset> value is near 2349The C<interval> I<MUST> be positive, and for numerical stability, the
1987C<ev_now ()> (the current time), but there is no range requirement for 2350interval value should be higher than C<1/8192> (which is around 100
1988this value, and in fact is often specified as zero. 2351microseconds) and C<offset> should be higher than C<0> and should have
2352at most a similar magnitude as the current time (say, within a factor of
2353ten). Typical values for offset are, in fact, C<0> or something between
2354C<0> and C<interval>, which is also the recommended range.
1989 2355
1990Note also that there is an upper limit to how often a timer can fire (CPU 2356Note also that there is an upper limit to how often a timer can fire (CPU
1991speed for example), so if C<interval> is very small then timing stability 2357speed for example), so if C<interval> is very small then timing stability
1992will of course deteriorate. Libev itself tries to be exact to be about one 2358will of course deteriorate. Libev itself tries to be exact to be about one
1993millisecond (if the OS supports it and the machine is fast enough). 2359millisecond (if the OS supports it and the machine is fast enough).
2023 2389
2024NOTE: I<< This callback must always return a time that is higher than or 2390NOTE: I<< This callback must always return a time that is higher than or
2025equal to the passed C<now> value >>. 2391equal to the passed C<now> value >>.
2026 2392
2027This can be used to create very complex timers, such as a timer that 2393This can be used to create very complex timers, such as a timer that
2028triggers on "next midnight, local time". To do this, you would calculate the 2394triggers on "next midnight, local time". To do this, you would calculate
2029next midnight after C<now> and return the timestamp value for this. How 2395the next midnight after C<now> and return the timestamp value for
2030you do this is, again, up to you (but it is not trivial, which is the main 2396this. Here is a (completely untested, no error checking) example on how to
2031reason I omitted it as an example). 2397do this:
2398
2399 #include <time.h>
2400
2401 static ev_tstamp
2402 my_rescheduler (ev_periodic *w, ev_tstamp now)
2403 {
2404 time_t tnow = (time_t)now;
2405 struct tm tm;
2406 localtime_r (&tnow, &tm);
2407
2408 tm.tm_sec = tm.tm_min = tm.tm_hour = 0; // midnight current day
2409 ++tm.tm_mday; // midnight next day
2410
2411 return mktime (&tm);
2412 }
2413
2414Note: this code might run into trouble on days that have more then two
2415midnights (beginning and end).
2032 2416
2033=back 2417=back
2034 2418
2035=item ev_periodic_again (loop, ev_periodic *) 2419=item ev_periodic_again (loop, ev_periodic *)
2036 2420
2074Example: Call a callback every hour, or, more precisely, whenever the 2458Example: Call a callback every hour, or, more precisely, whenever the
2075system time is divisible by 3600. The callback invocation times have 2459system time is divisible by 3600. The callback invocation times have
2076potentially a lot of jitter, but good long-term stability. 2460potentially a lot of jitter, but good long-term stability.
2077 2461
2078 static void 2462 static void
2079 clock_cb (struct ev_loop *loop, ev_io *w, int revents) 2463 clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
2080 { 2464 {
2081 ... its now a full hour (UTC, or TAI or whatever your clock follows) 2465 ... its now a full hour (UTC, or TAI or whatever your clock follows)
2082 } 2466 }
2083 2467
2084 ev_periodic hourly_tick; 2468 ev_periodic hourly_tick;
2101 2485
2102 ev_periodic hourly_tick; 2486 ev_periodic hourly_tick;
2103 ev_periodic_init (&hourly_tick, clock_cb, 2487 ev_periodic_init (&hourly_tick, clock_cb,
2104 fmod (ev_now (loop), 3600.), 3600., 0); 2488 fmod (ev_now (loop), 3600.), 3600., 0);
2105 ev_periodic_start (loop, &hourly_tick); 2489 ev_periodic_start (loop, &hourly_tick);
2106 2490
2107 2491
2108=head2 C<ev_signal> - signal me when a signal gets signalled! 2492=head2 C<ev_signal> - signal me when a signal gets signalled!
2109 2493
2110Signal watchers will trigger an event when the process receives a specific 2494Signal watchers will trigger an event when the process receives a specific
2111signal one or more times. Even though signals are very asynchronous, libev 2495signal one or more times. Even though signals are very asynchronous, libev
2112will try it's best to deliver signals synchronously, i.e. as part of the 2496will try its best to deliver signals synchronously, i.e. as part of the
2113normal event processing, like any other event. 2497normal event processing, like any other event.
2114 2498
2115If you want signals to be delivered truly asynchronously, just use 2499If you want signals to be delivered truly asynchronously, just use
2116C<sigaction> as you would do without libev and forget about sharing 2500C<sigaction> as you would do without libev and forget about sharing
2117the signal. You can even use C<ev_async> from a signal handler to 2501the signal. You can even use C<ev_async> from a signal handler to
2121only within the same loop, i.e. you can watch for C<SIGINT> in your 2505only within the same loop, i.e. you can watch for C<SIGINT> in your
2122default loop and for C<SIGIO> in another loop, but you cannot watch for 2506default loop and for C<SIGIO> in another loop, but you cannot watch for
2123C<SIGINT> in both the default loop and another loop at the same time. At 2507C<SIGINT> in both the default loop and another loop at the same time. At
2124the moment, C<SIGCHLD> is permanently tied to the default loop. 2508the moment, C<SIGCHLD> is permanently tied to the default loop.
2125 2509
2126When the first watcher gets started will libev actually register something 2510Only after the first watcher for a signal is started will libev actually
2127with the kernel (thus it coexists with your own signal handlers as long as 2511register something with the kernel. It thus coexists with your own signal
2128you don't register any with libev for the same signal). 2512handlers as long as you don't register any with libev for the same signal.
2129 2513
2130If possible and supported, libev will install its handlers with 2514If possible and supported, libev will install its handlers with
2131C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should 2515C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
2132not be unduly interrupted. If you have a problem with system calls getting 2516not be unduly interrupted. If you have a problem with system calls getting
2133interrupted by signals you can block all signals in an C<ev_check> watcher 2517interrupted by signals you can block all signals in an C<ev_check> watcher
2134and unblock them in an C<ev_prepare> watcher. 2518and unblock them in an C<ev_prepare> watcher.
2135 2519
2136=head3 The special problem of inheritance over execve 2520=head3 The special problem of inheritance over fork/execve/pthread_create
2137 2521
2138Both the signal mask (C<sigprocmask>) and the signal disposition 2522Both the signal mask (C<sigprocmask>) and the signal disposition
2139(C<sigaction>) are unspecified after starting a signal watcher (and after 2523(C<sigaction>) are unspecified after starting a signal watcher (and after
2140stopping it again), that is, libev might or might not block the signal, 2524stopping it again), that is, libev might or might not block the signal,
2141and might or might not set or restore the installed signal handler. 2525and might or might not set or restore the installed signal handler (but
2526see C<EVFLAG_NOSIGMASK>).
2142 2527
2143While this does not matter for the signal disposition (libev never 2528While this does not matter for the signal disposition (libev never
2144sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on 2529sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
2145C<execve>), this matters for the signal mask: many programs do not expect 2530C<execve>), this matters for the signal mask: many programs do not expect
2146certain signals to be blocked. 2531certain signals to be blocked.
2151 2536
2152The simplest way to ensure that the signal mask is reset in the child is 2537The simplest way to ensure that the signal mask is reset in the child is
2153to install a fork handler with C<pthread_atfork> that resets it. That will 2538to install a fork handler with C<pthread_atfork> that resets it. That will
2154catch fork calls done by libraries (such as the libc) as well. 2539catch fork calls done by libraries (such as the libc) as well.
2155 2540
2156In current versions of libev, you can also ensure that the signal mask is 2541In current versions of libev, the signal will not be blocked indefinitely
2157not blocking any signals (except temporarily, so thread users watch out) 2542unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
2158by specifying the C<EVFLAG_NOSIGFD> when creating the event loop. This 2543the window of opportunity for problems, it will not go away, as libev
2159is not guaranteed for future versions, however. 2544I<has> to modify the signal mask, at least temporarily.
2545
2546So I can't stress this enough: I<If you do not reset your signal mask when
2547you expect it to be empty, you have a race condition in your code>. This
2548is not a libev-specific thing, this is true for most event libraries.
2549
2550=head3 The special problem of threads signal handling
2551
2552POSIX threads has problematic signal handling semantics, specifically,
2553a lot of functionality (sigfd, sigwait etc.) only really works if all
2554threads in a process block signals, which is hard to achieve.
2555
2556When you want to use sigwait (or mix libev signal handling with your own
2557for the same signals), you can tackle this problem by globally blocking
2558all signals before creating any threads (or creating them with a fully set
2559sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
2560loops. Then designate one thread as "signal receiver thread" which handles
2561these signals. You can pass on any signals that libev might be interested
2562in by calling C<ev_feed_signal>.
2160 2563
2161=head3 Watcher-Specific Functions and Data Members 2564=head3 Watcher-Specific Functions and Data Members
2162 2565
2163=over 4 2566=over 4
2164 2567
2180Example: Try to exit cleanly on SIGINT. 2583Example: Try to exit cleanly on SIGINT.
2181 2584
2182 static void 2585 static void
2183 sigint_cb (struct ev_loop *loop, ev_signal *w, int revents) 2586 sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
2184 { 2587 {
2185 ev_unloop (loop, EVUNLOOP_ALL); 2588 ev_break (loop, EVBREAK_ALL);
2186 } 2589 }
2187 2590
2188 ev_signal signal_watcher; 2591 ev_signal signal_watcher;
2189 ev_signal_init (&signal_watcher, sigint_cb, SIGINT); 2592 ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
2190 ev_signal_start (loop, &signal_watcher); 2593 ev_signal_start (loop, &signal_watcher);
2299 2702
2300=head2 C<ev_stat> - did the file attributes just change? 2703=head2 C<ev_stat> - did the file attributes just change?
2301 2704
2302This watches a file system path for attribute changes. That is, it calls 2705This watches a file system path for attribute changes. That is, it calls
2303C<stat> on that path in regular intervals (or when the OS says it changed) 2706C<stat> on that path in regular intervals (or when the OS says it changed)
2304and sees if it changed compared to the last time, invoking the callback if 2707and sees if it changed compared to the last time, invoking the callback
2305it did. 2708if it did. Starting the watcher C<stat>'s the file, so only changes that
2709happen after the watcher has been started will be reported.
2306 2710
2307The path does not need to exist: changing from "path exists" to "path does 2711The path does not need to exist: changing from "path exists" to "path does
2308not exist" is a status change like any other. The condition "path does not 2712not exist" is a status change like any other. The condition "path does not
2309exist" (or more correctly "path cannot be stat'ed") is signified by the 2713exist" (or more correctly "path cannot be stat'ed") is signified by the
2310C<st_nlink> field being zero (which is otherwise always forced to be at 2714C<st_nlink> field being zero (which is otherwise always forced to be at
2540Apart from keeping your process non-blocking (which is a useful 2944Apart from keeping your process non-blocking (which is a useful
2541effect on its own sometimes), idle watchers are a good place to do 2945effect on its own sometimes), idle watchers are a good place to do
2542"pseudo-background processing", or delay processing stuff to after the 2946"pseudo-background processing", or delay processing stuff to after the
2543event loop has handled all outstanding events. 2947event loop has handled all outstanding events.
2544 2948
2949=head3 Abusing an C<ev_idle> watcher for its side-effect
2950
2951As long as there is at least one active idle watcher, libev will never
2952sleep unnecessarily. Or in other words, it will loop as fast as possible.
2953For this to work, the idle watcher doesn't need to be invoked at all - the
2954lowest priority will do.
2955
2956This mode of operation can be useful together with an C<ev_check> watcher,
2957to do something on each event loop iteration - for example to balance load
2958between different connections.
2959
2960See L</Abusing an ev_check watcher for its side-effect> for a longer
2961example.
2962
2545=head3 Watcher-Specific Functions and Data Members 2963=head3 Watcher-Specific Functions and Data Members
2546 2964
2547=over 4 2965=over 4
2548 2966
2549=item ev_idle_init (ev_idle *, callback) 2967=item ev_idle_init (ev_idle *, callback)
2560callback, free it. Also, use no error checking, as usual. 2978callback, free it. Also, use no error checking, as usual.
2561 2979
2562 static void 2980 static void
2563 idle_cb (struct ev_loop *loop, ev_idle *w, int revents) 2981 idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
2564 { 2982 {
2983 // stop the watcher
2984 ev_idle_stop (loop, w);
2985
2986 // now we can free it
2565 free (w); 2987 free (w);
2988
2566 // now do something you wanted to do when the program has 2989 // now do something you wanted to do when the program has
2567 // no longer anything immediate to do. 2990 // no longer anything immediate to do.
2568 } 2991 }
2569 2992
2570 ev_idle *idle_watcher = malloc (sizeof (ev_idle)); 2993 ev_idle *idle_watcher = malloc (sizeof (ev_idle));
2572 ev_idle_start (loop, idle_watcher); 2995 ev_idle_start (loop, idle_watcher);
2573 2996
2574 2997
2575=head2 C<ev_prepare> and C<ev_check> - customise your event loop! 2998=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
2576 2999
2577Prepare and check watchers are usually (but not always) used in pairs: 3000Prepare and check watchers are often (but not always) used in pairs:
2578prepare watchers get invoked before the process blocks and check watchers 3001prepare watchers get invoked before the process blocks and check watchers
2579afterwards. 3002afterwards.
2580 3003
2581You I<must not> call C<ev_loop> or similar functions that enter 3004You I<must not> call C<ev_run> (or similar functions that enter the
2582the current event loop from either C<ev_prepare> or C<ev_check> 3005current event loop) or C<ev_loop_fork> from either C<ev_prepare> or
2583watchers. Other loops than the current one are fine, however. The 3006C<ev_check> watchers. Other loops than the current one are fine,
2584rationale behind this is that you do not need to check for recursion in 3007however. The rationale behind this is that you do not need to check
2585those watchers, i.e. the sequence will always be C<ev_prepare>, blocking, 3008for recursion in those watchers, i.e. the sequence will always be
2586C<ev_check> so if you have one watcher of each kind they will always be 3009C<ev_prepare>, blocking, C<ev_check> so if you have one watcher of each
2587called in pairs bracketing the blocking call. 3010kind they will always be called in pairs bracketing the blocking call.
2588 3011
2589Their main purpose is to integrate other event mechanisms into libev and 3012Their main purpose is to integrate other event mechanisms into libev and
2590their use is somewhat advanced. They could be used, for example, to track 3013their use is somewhat advanced. They could be used, for example, to track
2591variable changes, implement your own watchers, integrate net-snmp or a 3014variable changes, implement your own watchers, integrate net-snmp or a
2592coroutine library and lots more. They are also occasionally useful if 3015coroutine library and lots more. They are also occasionally useful if
2610with priority higher than or equal to the event loop and one coroutine 3033with priority higher than or equal to the event loop and one coroutine
2611of lower priority, but only once, using idle watchers to keep the event 3034of lower priority, but only once, using idle watchers to keep the event
2612loop from blocking if lower-priority coroutines are active, thus mapping 3035loop from blocking if lower-priority coroutines are active, thus mapping
2613low-priority coroutines to idle/background tasks). 3036low-priority coroutines to idle/background tasks).
2614 3037
2615It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>) 3038When used for this purpose, it is recommended to give C<ev_check> watchers
2616priority, to ensure that they are being run before any other watchers 3039highest (C<EV_MAXPRI>) priority, to ensure that they are being run before
2617after the poll (this doesn't matter for C<ev_prepare> watchers). 3040any other watchers after the poll (this doesn't matter for C<ev_prepare>
3041watchers).
2618 3042
2619Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not 3043Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
2620activate ("feed") events into libev. While libev fully supports this, they 3044activate ("feed") events into libev. While libev fully supports this, they
2621might get executed before other C<ev_check> watchers did their job. As 3045might get executed before other C<ev_check> watchers did their job. As
2622C<ev_check> watchers are often used to embed other (non-libev) event 3046C<ev_check> watchers are often used to embed other (non-libev) event
2623loops those other event loops might be in an unusable state until their 3047loops those other event loops might be in an unusable state until their
2624C<ev_check> watcher ran (always remind yourself to coexist peacefully with 3048C<ev_check> watcher ran (always remind yourself to coexist peacefully with
2625others). 3049others).
3050
3051=head3 Abusing an C<ev_check> watcher for its side-effect
3052
3053C<ev_check> (and less often also C<ev_prepare>) watchers can also be
3054useful because they are called once per event loop iteration. For
3055example, if you want to handle a large number of connections fairly, you
3056normally only do a bit of work for each active connection, and if there
3057is more work to do, you wait for the next event loop iteration, so other
3058connections have a chance of making progress.
3059
3060Using an C<ev_check> watcher is almost enough: it will be called on the
3061next event loop iteration. However, that isn't as soon as possible -
3062without external events, your C<ev_check> watcher will not be invoked.
3063
3064This is where C<ev_idle> watchers come in handy - all you need is a
3065single global idle watcher that is active as long as you have one active
3066C<ev_check> watcher. The C<ev_idle> watcher makes sure the event loop
3067will not sleep, and the C<ev_check> watcher makes sure a callback gets
3068invoked. Neither watcher alone can do that.
2626 3069
2627=head3 Watcher-Specific Functions and Data Members 3070=head3 Watcher-Specific Functions and Data Members
2628 3071
2629=over 4 3072=over 4
2630 3073
2754 3197
2755 if (timeout >= 0) 3198 if (timeout >= 0)
2756 // create/start timer 3199 // create/start timer
2757 3200
2758 // poll 3201 // poll
2759 ev_loop (EV_A_ 0); 3202 ev_run (EV_A_ 0);
2760 3203
2761 // stop timer again 3204 // stop timer again
2762 if (timeout >= 0) 3205 if (timeout >= 0)
2763 ev_timer_stop (EV_A_ &to); 3206 ev_timer_stop (EV_A_ &to);
2764 3207
2831 3274
2832=over 4 3275=over 4
2833 3276
2834=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop) 3277=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
2835 3278
2836=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop) 3279=item ev_embed_set (ev_embed *, struct ev_loop *embedded_loop)
2837 3280
2838Configures the watcher to embed the given loop, which must be 3281Configures the watcher to embed the given loop, which must be
2839embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be 3282embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
2840invoked automatically, otherwise it is the responsibility of the callback 3283invoked automatically, otherwise it is the responsibility of the callback
2841to invoke it (it will continue to be called until the sweep has been done, 3284to invoke it (it will continue to be called until the sweep has been done,
2842if you do not want that, you need to temporarily stop the embed watcher). 3285if you do not want that, you need to temporarily stop the embed watcher).
2843 3286
2844=item ev_embed_sweep (loop, ev_embed *) 3287=item ev_embed_sweep (loop, ev_embed *)
2845 3288
2846Make a single, non-blocking sweep over the embedded loop. This works 3289Make a single, non-blocking sweep over the embedded loop. This works
2847similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most 3290similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
2848appropriate way for embedded loops. 3291appropriate way for embedded loops.
2849 3292
2850=item struct ev_loop *other [read-only] 3293=item struct ev_loop *other [read-only]
2851 3294
2852The embedded event loop. 3295The embedded event loop.
2862used). 3305used).
2863 3306
2864 struct ev_loop *loop_hi = ev_default_init (0); 3307 struct ev_loop *loop_hi = ev_default_init (0);
2865 struct ev_loop *loop_lo = 0; 3308 struct ev_loop *loop_lo = 0;
2866 ev_embed embed; 3309 ev_embed embed;
2867 3310
2868 // see if there is a chance of getting one that works 3311 // see if there is a chance of getting one that works
2869 // (remember that a flags value of 0 means autodetection) 3312 // (remember that a flags value of 0 means autodetection)
2870 loop_lo = ev_embeddable_backends () & ev_recommended_backends () 3313 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
2871 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ()) 3314 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
2872 : 0; 3315 : 0;
2886C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too). 3329C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
2887 3330
2888 struct ev_loop *loop = ev_default_init (0); 3331 struct ev_loop *loop = ev_default_init (0);
2889 struct ev_loop *loop_socket = 0; 3332 struct ev_loop *loop_socket = 0;
2890 ev_embed embed; 3333 ev_embed embed;
2891 3334
2892 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE) 3335 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2893 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE)) 3336 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2894 { 3337 {
2895 ev_embed_init (&embed, 0, loop_socket); 3338 ev_embed_init (&embed, 0, loop_socket);
2896 ev_embed_start (loop, &embed); 3339 ev_embed_start (loop, &embed);
2904 3347
2905=head2 C<ev_fork> - the audacity to resume the event loop after a fork 3348=head2 C<ev_fork> - the audacity to resume the event loop after a fork
2906 3349
2907Fork watchers are called when a C<fork ()> was detected (usually because 3350Fork watchers are called when a C<fork ()> was detected (usually because
2908whoever is a good citizen cared to tell libev about it by calling 3351whoever is a good citizen cared to tell libev about it by calling
2909C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the 3352C<ev_loop_fork>). The invocation is done before the event loop blocks next
2910event loop blocks next and before C<ev_check> watchers are being called, 3353and before C<ev_check> watchers are being called, and only in the child
2911and only in the child after the fork. If whoever good citizen calling 3354after the fork. If whoever good citizen calling C<ev_default_fork> cheats
2912C<ev_default_fork> cheats and calls it in the wrong process, the fork 3355and calls it in the wrong process, the fork handlers will be invoked, too,
2913handlers will be invoked, too, of course. 3356of course.
2914 3357
2915=head3 The special problem of life after fork - how is it possible? 3358=head3 The special problem of life after fork - how is it possible?
2916 3359
2917Most uses of C<fork()> consist of forking, then some simple calls to ste 3360Most uses of C<fork ()> consist of forking, then some simple calls to set
2918up/change the process environment, followed by a call to C<exec()>. This 3361up/change the process environment, followed by a call to C<exec()>. This
2919sequence should be handled by libev without any problems. 3362sequence should be handled by libev without any problems.
2920 3363
2921This changes when the application actually wants to do event handling 3364This changes when the application actually wants to do event handling
2922in the child, or both parent in child, in effect "continuing" after the 3365in the child, or both parent in child, in effect "continuing" after the
2938disadvantage of having to use multiple event loops (which do not support 3381disadvantage of having to use multiple event loops (which do not support
2939signal watchers). 3382signal watchers).
2940 3383
2941When this is not possible, or you want to use the default loop for 3384When this is not possible, or you want to use the default loop for
2942other reasons, then in the process that wants to start "fresh", call 3385other reasons, then in the process that wants to start "fresh", call
2943C<ev_default_destroy ()> followed by C<ev_default_loop (...)>. Destroying 3386C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
2944the default loop will "orphan" (not stop) all registered watchers, so you 3387Destroying the default loop will "orphan" (not stop) all registered
2945have to be careful not to execute code that modifies those watchers. Note 3388watchers, so you have to be careful not to execute code that modifies
2946also that in that case, you have to re-register any signal watchers. 3389those watchers. Note also that in that case, you have to re-register any
3390signal watchers.
2947 3391
2948=head3 Watcher-Specific Functions and Data Members 3392=head3 Watcher-Specific Functions and Data Members
2949 3393
2950=over 4 3394=over 4
2951 3395
2952=item ev_fork_init (ev_signal *, callback) 3396=item ev_fork_init (ev_fork *, callback)
2953 3397
2954Initialises and configures the fork watcher - it has no parameters of any 3398Initialises and configures the fork watcher - it has no parameters of any
2955kind. There is a C<ev_fork_set> macro, but using it is utterly pointless, 3399kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
2956believe me. 3400really.
2957 3401
2958=back 3402=back
2959 3403
2960 3404
3405=head2 C<ev_cleanup> - even the best things end
3406
3407Cleanup watchers are called just before the event loop is being destroyed
3408by a call to C<ev_loop_destroy>.
3409
3410While there is no guarantee that the event loop gets destroyed, cleanup
3411watchers provide a convenient method to install cleanup hooks for your
3412program, worker threads and so on - you just to make sure to destroy the
3413loop when you want them to be invoked.
3414
3415Cleanup watchers are invoked in the same way as any other watcher. Unlike
3416all other watchers, they do not keep a reference to the event loop (which
3417makes a lot of sense if you think about it). Like all other watchers, you
3418can call libev functions in the callback, except C<ev_cleanup_start>.
3419
3420=head3 Watcher-Specific Functions and Data Members
3421
3422=over 4
3423
3424=item ev_cleanup_init (ev_cleanup *, callback)
3425
3426Initialises and configures the cleanup watcher - it has no parameters of
3427any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
3428pointless, I assure you.
3429
3430=back
3431
3432Example: Register an atexit handler to destroy the default loop, so any
3433cleanup functions are called.
3434
3435 static void
3436 program_exits (void)
3437 {
3438 ev_loop_destroy (EV_DEFAULT_UC);
3439 }
3440
3441 ...
3442 atexit (program_exits);
3443
3444
2961=head2 C<ev_async> - how to wake up another event loop 3445=head2 C<ev_async> - how to wake up an event loop
2962 3446
2963In general, you cannot use an C<ev_loop> from multiple threads or other 3447In general, you cannot use an C<ev_loop> from multiple threads or other
2964asynchronous sources such as signal handlers (as opposed to multiple event 3448asynchronous sources such as signal handlers (as opposed to multiple event
2965loops - those are of course safe to use in different threads). 3449loops - those are of course safe to use in different threads).
2966 3450
2967Sometimes, however, you need to wake up another event loop you do not 3451Sometimes, however, you need to wake up an event loop you do not control,
2968control, for example because it belongs to another thread. This is what 3452for example because it belongs to another thread. This is what C<ev_async>
2969C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you 3453watchers do: as long as the C<ev_async> watcher is active, you can signal
2970can signal it by calling C<ev_async_send>, which is thread- and signal 3454it by calling C<ev_async_send>, which is thread- and signal safe.
2971safe.
2972 3455
2973This functionality is very similar to C<ev_signal> watchers, as signals, 3456This functionality is very similar to C<ev_signal> watchers, as signals,
2974too, are asynchronous in nature, and signals, too, will be compressed 3457too, are asynchronous in nature, and signals, too, will be compressed
2975(i.e. the number of callback invocations may be less than the number of 3458(i.e. the number of callback invocations may be less than the number of
2976C<ev_async_sent> calls). 3459C<ev_async_send> calls). In fact, you could use signal watchers as a kind
2977 3460of "global async watchers" by using a watcher on an otherwise unused
2978Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not 3461signal, and C<ev_feed_signal> to signal this watcher from another thread,
2979just the default loop. 3462even without knowing which loop owns the signal.
2980 3463
2981=head3 Queueing 3464=head3 Queueing
2982 3465
2983C<ev_async> does not support queueing of data in any way. The reason 3466C<ev_async> does not support queueing of data in any way. The reason
2984is that the author does not know of a simple (or any) algorithm for a 3467is that the author does not know of a simple (or any) algorithm for a
3076trust me. 3559trust me.
3077 3560
3078=item ev_async_send (loop, ev_async *) 3561=item ev_async_send (loop, ev_async *)
3079 3562
3080Sends/signals/activates the given C<ev_async> watcher, that is, feeds 3563Sends/signals/activates the given C<ev_async> watcher, that is, feeds
3081an C<EV_ASYNC> event on the watcher into the event loop. Unlike 3564an C<EV_ASYNC> event on the watcher into the event loop, and instantly
3565returns.
3566
3082C<ev_feed_event>, this call is safe to do from other threads, signal or 3567Unlike C<ev_feed_event>, this call is safe to do from other threads,
3083similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding 3568signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
3084section below on what exactly this means). 3569embedding section below on what exactly this means).
3085 3570
3086Note that, as with other watchers in libev, multiple events might get 3571Note that, as with other watchers in libev, multiple events might get
3087compressed into a single callback invocation (another way to look at this 3572compressed into a single callback invocation (another way to look at
3088is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>, 3573this is that C<ev_async> watchers are level-triggered: they are set on
3089reset when the event loop detects that). 3574C<ev_async_send>, reset when the event loop detects that).
3090 3575
3091This call incurs the overhead of a system call only once per event loop 3576This call incurs the overhead of at most one extra system call per event
3092iteration, so while the overhead might be noticeable, it doesn't apply to 3577loop iteration, if the event loop is blocked, and no syscall at all if
3093repeated calls to C<ev_async_send> for the same event loop. 3578the event loop (or your program) is processing events. That means that
3579repeated calls are basically free (there is no need to avoid calls for
3580performance reasons) and that the overhead becomes smaller (typically
3581zero) under load.
3094 3582
3095=item bool = ev_async_pending (ev_async *) 3583=item bool = ev_async_pending (ev_async *)
3096 3584
3097Returns a non-zero value when C<ev_async_send> has been called on the 3585Returns a non-zero value when C<ev_async_send> has been called on the
3098watcher but the event has not yet been processed (or even noted) by the 3586watcher but the event has not yet been processed (or even noted) by the
3115 3603
3116There are some other functions of possible interest. Described. Here. Now. 3604There are some other functions of possible interest. Described. Here. Now.
3117 3605
3118=over 4 3606=over 4
3119 3607
3120=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback) 3608=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback, arg)
3121 3609
3122This function combines a simple timer and an I/O watcher, calls your 3610This function combines a simple timer and an I/O watcher, calls your
3123callback on whichever event happens first and automatically stops both 3611callback on whichever event happens first and automatically stops both
3124watchers. This is useful if you want to wait for a single event on an fd 3612watchers. This is useful if you want to wait for a single event on an fd
3125or timeout without having to allocate/configure/start/stop/free one or 3613or timeout without having to allocate/configure/start/stop/free one or
3131 3619
3132If C<timeout> is less than 0, then no timeout watcher will be 3620If C<timeout> is less than 0, then no timeout watcher will be
3133started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and 3621started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
3134repeat = 0) will be started. C<0> is a valid timeout. 3622repeat = 0) will be started. C<0> is a valid timeout.
3135 3623
3136The callback has the type C<void (*cb)(int revents, void *arg)> and gets 3624The callback has the type C<void (*cb)(int revents, void *arg)> and is
3137passed an C<revents> set like normal event callbacks (a combination of 3625passed an C<revents> set like normal event callbacks (a combination of
3138C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg> 3626C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
3139value passed to C<ev_once>. Note that it is possible to receive I<both> 3627value passed to C<ev_once>. Note that it is possible to receive I<both>
3140a timeout and an io event at the same time - you probably should give io 3628a timeout and an io event at the same time - you probably should give io
3141events precedence. 3629events precedence.
3142 3630
3143Example: wait up to ten seconds for data to appear on STDIN_FILENO. 3631Example: wait up to ten seconds for data to appear on STDIN_FILENO.
3144 3632
3145 static void stdin_ready (int revents, void *arg) 3633 static void stdin_ready (int revents, void *arg)
3146 { 3634 {
3147 if (revents & EV_READ) 3635 if (revents & EV_READ)
3148 /* stdin might have data for us, joy! */; 3636 /* stdin might have data for us, joy! */;
3149 else if (revents & EV_TIMEOUT) 3637 else if (revents & EV_TIMER)
3150 /* doh, nothing entered */; 3638 /* doh, nothing entered */;
3151 } 3639 }
3152 3640
3153 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0); 3641 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
3154 3642
3155=item ev_feed_fd_event (loop, int fd, int revents) 3643=item ev_feed_fd_event (loop, int fd, int revents)
3156 3644
3157Feed an event on the given fd, as if a file descriptor backend detected 3645Feed an event on the given fd, as if a file descriptor backend detected
3158the given events it. 3646the given events.
3159 3647
3160=item ev_feed_signal_event (loop, int signum) 3648=item ev_feed_signal_event (loop, int signum)
3161 3649
3162Feed an event as if the given signal occurred (C<loop> must be the default 3650Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
3163loop!). 3651which is async-safe.
3164 3652
3165=back 3653=back
3654
3655
3656=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
3657
3658This section explains some common idioms that are not immediately
3659obvious. Note that examples are sprinkled over the whole manual, and this
3660section only contains stuff that wouldn't fit anywhere else.
3661
3662=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
3663
3664Each watcher has, by default, a C<void *data> member that you can read
3665or modify at any time: libev will completely ignore it. This can be used
3666to associate arbitrary data with your watcher. If you need more data and
3667don't want to allocate memory separately and store a pointer to it in that
3668data member, you can also "subclass" the watcher type and provide your own
3669data:
3670
3671 struct my_io
3672 {
3673 ev_io io;
3674 int otherfd;
3675 void *somedata;
3676 struct whatever *mostinteresting;
3677 };
3678
3679 ...
3680 struct my_io w;
3681 ev_io_init (&w.io, my_cb, fd, EV_READ);
3682
3683And since your callback will be called with a pointer to the watcher, you
3684can cast it back to your own type:
3685
3686 static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
3687 {
3688 struct my_io *w = (struct my_io *)w_;
3689 ...
3690 }
3691
3692More interesting and less C-conformant ways of casting your callback
3693function type instead have been omitted.
3694
3695=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
3696
3697Another common scenario is to use some data structure with multiple
3698embedded watchers, in effect creating your own watcher that combines
3699multiple libev event sources into one "super-watcher":
3700
3701 struct my_biggy
3702 {
3703 int some_data;
3704 ev_timer t1;
3705 ev_timer t2;
3706 }
3707
3708In this case getting the pointer to C<my_biggy> is a bit more
3709complicated: Either you store the address of your C<my_biggy> struct in
3710the C<data> member of the watcher (for woozies or C++ coders), or you need
3711to use some pointer arithmetic using C<offsetof> inside your watchers (for
3712real programmers):
3713
3714 #include <stddef.h>
3715
3716 static void
3717 t1_cb (EV_P_ ev_timer *w, int revents)
3718 {
3719 struct my_biggy big = (struct my_biggy *)
3720 (((char *)w) - offsetof (struct my_biggy, t1));
3721 }
3722
3723 static void
3724 t2_cb (EV_P_ ev_timer *w, int revents)
3725 {
3726 struct my_biggy big = (struct my_biggy *)
3727 (((char *)w) - offsetof (struct my_biggy, t2));
3728 }
3729
3730=head2 AVOIDING FINISHING BEFORE RETURNING
3731
3732Often you have structures like this in event-based programs:
3733
3734 callback ()
3735 {
3736 free (request);
3737 }
3738
3739 request = start_new_request (..., callback);
3740
3741The intent is to start some "lengthy" operation. The C<request> could be
3742used to cancel the operation, or do other things with it.
3743
3744It's not uncommon to have code paths in C<start_new_request> that
3745immediately invoke the callback, for example, to report errors. Or you add
3746some caching layer that finds that it can skip the lengthy aspects of the
3747operation and simply invoke the callback with the result.
3748
3749The problem here is that this will happen I<before> C<start_new_request>
3750has returned, so C<request> is not set.
3751
3752Even if you pass the request by some safer means to the callback, you
3753might want to do something to the request after starting it, such as
3754canceling it, which probably isn't working so well when the callback has
3755already been invoked.
3756
3757A common way around all these issues is to make sure that
3758C<start_new_request> I<always> returns before the callback is invoked. If
3759C<start_new_request> immediately knows the result, it can artificially
3760delay invoking the callback by using a C<prepare> or C<idle> watcher for
3761example, or more sneakily, by reusing an existing (stopped) watcher and
3762pushing it into the pending queue:
3763
3764 ev_set_cb (watcher, callback);
3765 ev_feed_event (EV_A_ watcher, 0);
3766
3767This way, C<start_new_request> can safely return before the callback is
3768invoked, while not delaying callback invocation too much.
3769
3770=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
3771
3772Often (especially in GUI toolkits) there are places where you have
3773I<modal> interaction, which is most easily implemented by recursively
3774invoking C<ev_run>.
3775
3776This brings the problem of exiting - a callback might want to finish the
3777main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
3778a modal "Are you sure?" dialog is still waiting), or just the nested one
3779and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
3780other combination: In these cases, a simple C<ev_break> will not work.
3781
3782The solution is to maintain "break this loop" variable for each C<ev_run>
3783invocation, and use a loop around C<ev_run> until the condition is
3784triggered, using C<EVRUN_ONCE>:
3785
3786 // main loop
3787 int exit_main_loop = 0;
3788
3789 while (!exit_main_loop)
3790 ev_run (EV_DEFAULT_ EVRUN_ONCE);
3791
3792 // in a modal watcher
3793 int exit_nested_loop = 0;
3794
3795 while (!exit_nested_loop)
3796 ev_run (EV_A_ EVRUN_ONCE);
3797
3798To exit from any of these loops, just set the corresponding exit variable:
3799
3800 // exit modal loop
3801 exit_nested_loop = 1;
3802
3803 // exit main program, after modal loop is finished
3804 exit_main_loop = 1;
3805
3806 // exit both
3807 exit_main_loop = exit_nested_loop = 1;
3808
3809=head2 THREAD LOCKING EXAMPLE
3810
3811Here is a fictitious example of how to run an event loop in a different
3812thread from where callbacks are being invoked and watchers are
3813created/added/removed.
3814
3815For a real-world example, see the C<EV::Loop::Async> perl module,
3816which uses exactly this technique (which is suited for many high-level
3817languages).
3818
3819The example uses a pthread mutex to protect the loop data, a condition
3820variable to wait for callback invocations, an async watcher to notify the
3821event loop thread and an unspecified mechanism to wake up the main thread.
3822
3823First, you need to associate some data with the event loop:
3824
3825 typedef struct {
3826 mutex_t lock; /* global loop lock */
3827 ev_async async_w;
3828 thread_t tid;
3829 cond_t invoke_cv;
3830 } userdata;
3831
3832 void prepare_loop (EV_P)
3833 {
3834 // for simplicity, we use a static userdata struct.
3835 static userdata u;
3836
3837 ev_async_init (&u->async_w, async_cb);
3838 ev_async_start (EV_A_ &u->async_w);
3839
3840 pthread_mutex_init (&u->lock, 0);
3841 pthread_cond_init (&u->invoke_cv, 0);
3842
3843 // now associate this with the loop
3844 ev_set_userdata (EV_A_ u);
3845 ev_set_invoke_pending_cb (EV_A_ l_invoke);
3846 ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
3847
3848 // then create the thread running ev_run
3849 pthread_create (&u->tid, 0, l_run, EV_A);
3850 }
3851
3852The callback for the C<ev_async> watcher does nothing: the watcher is used
3853solely to wake up the event loop so it takes notice of any new watchers
3854that might have been added:
3855
3856 static void
3857 async_cb (EV_P_ ev_async *w, int revents)
3858 {
3859 // just used for the side effects
3860 }
3861
3862The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
3863protecting the loop data, respectively.
3864
3865 static void
3866 l_release (EV_P)
3867 {
3868 userdata *u = ev_userdata (EV_A);
3869 pthread_mutex_unlock (&u->lock);
3870 }
3871
3872 static void
3873 l_acquire (EV_P)
3874 {
3875 userdata *u = ev_userdata (EV_A);
3876 pthread_mutex_lock (&u->lock);
3877 }
3878
3879The event loop thread first acquires the mutex, and then jumps straight
3880into C<ev_run>:
3881
3882 void *
3883 l_run (void *thr_arg)
3884 {
3885 struct ev_loop *loop = (struct ev_loop *)thr_arg;
3886
3887 l_acquire (EV_A);
3888 pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
3889 ev_run (EV_A_ 0);
3890 l_release (EV_A);
3891
3892 return 0;
3893 }
3894
3895Instead of invoking all pending watchers, the C<l_invoke> callback will
3896signal the main thread via some unspecified mechanism (signals? pipe
3897writes? C<Async::Interrupt>?) and then waits until all pending watchers
3898have been called (in a while loop because a) spurious wakeups are possible
3899and b) skipping inter-thread-communication when there are no pending
3900watchers is very beneficial):
3901
3902 static void
3903 l_invoke (EV_P)
3904 {
3905 userdata *u = ev_userdata (EV_A);
3906
3907 while (ev_pending_count (EV_A))
3908 {
3909 wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
3910 pthread_cond_wait (&u->invoke_cv, &u->lock);
3911 }
3912 }
3913
3914Now, whenever the main thread gets told to invoke pending watchers, it
3915will grab the lock, call C<ev_invoke_pending> and then signal the loop
3916thread to continue:
3917
3918 static void
3919 real_invoke_pending (EV_P)
3920 {
3921 userdata *u = ev_userdata (EV_A);
3922
3923 pthread_mutex_lock (&u->lock);
3924 ev_invoke_pending (EV_A);
3925 pthread_cond_signal (&u->invoke_cv);
3926 pthread_mutex_unlock (&u->lock);
3927 }
3928
3929Whenever you want to start/stop a watcher or do other modifications to an
3930event loop, you will now have to lock:
3931
3932 ev_timer timeout_watcher;
3933 userdata *u = ev_userdata (EV_A);
3934
3935 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
3936
3937 pthread_mutex_lock (&u->lock);
3938 ev_timer_start (EV_A_ &timeout_watcher);
3939 ev_async_send (EV_A_ &u->async_w);
3940 pthread_mutex_unlock (&u->lock);
3941
3942Note that sending the C<ev_async> watcher is required because otherwise
3943an event loop currently blocking in the kernel will have no knowledge
3944about the newly added timer. By waking up the loop it will pick up any new
3945watchers in the next event loop iteration.
3946
3947=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
3948
3949While the overhead of a callback that e.g. schedules a thread is small, it
3950is still an overhead. If you embed libev, and your main usage is with some
3951kind of threads or coroutines, you might want to customise libev so that
3952doesn't need callbacks anymore.
3953
3954Imagine you have coroutines that you can switch to using a function
3955C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
3956and that due to some magic, the currently active coroutine is stored in a
3957global called C<current_coro>. Then you can build your own "wait for libev
3958event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
3959the differing C<;> conventions):
3960
3961 #define EV_CB_DECLARE(type) struct my_coro *cb;
3962 #define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
3963
3964That means instead of having a C callback function, you store the
3965coroutine to switch to in each watcher, and instead of having libev call
3966your callback, you instead have it switch to that coroutine.
3967
3968A coroutine might now wait for an event with a function called
3969C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
3970matter when, or whether the watcher is active or not when this function is
3971called):
3972
3973 void
3974 wait_for_event (ev_watcher *w)
3975 {
3976 ev_set_cb (w, current_coro);
3977 switch_to (libev_coro);
3978 }
3979
3980That basically suspends the coroutine inside C<wait_for_event> and
3981continues the libev coroutine, which, when appropriate, switches back to
3982this or any other coroutine.
3983
3984You can do similar tricks if you have, say, threads with an event queue -
3985instead of storing a coroutine, you store the queue object and instead of
3986switching to a coroutine, you push the watcher onto the queue and notify
3987any waiters.
3988
3989To embed libev, see L</EMBEDDING>, but in short, it's easiest to create two
3990files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
3991
3992 // my_ev.h
3993 #define EV_CB_DECLARE(type) struct my_coro *cb;
3994 #define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
3995 #include "../libev/ev.h"
3996
3997 // my_ev.c
3998 #define EV_H "my_ev.h"
3999 #include "../libev/ev.c"
4000
4001And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
4002F<my_ev.c> into your project. When properly specifying include paths, you
4003can even use F<ev.h> as header file name directly.
3166 4004
3167 4005
3168=head1 LIBEVENT EMULATION 4006=head1 LIBEVENT EMULATION
3169 4007
3170Libev offers a compatibility emulation layer for libevent. It cannot 4008Libev offers a compatibility emulation layer for libevent. It cannot
3171emulate the internals of libevent, so here are some usage hints: 4009emulate the internals of libevent, so here are some usage hints:
3172 4010
3173=over 4 4011=over 4
4012
4013=item * Only the libevent-1.4.1-beta API is being emulated.
4014
4015This was the newest libevent version available when libev was implemented,
4016and is still mostly unchanged in 2010.
3174 4017
3175=item * Use it by including <event.h>, as usual. 4018=item * Use it by including <event.h>, as usual.
3176 4019
3177=item * The following members are fully supported: ev_base, ev_callback, 4020=item * The following members are fully supported: ev_base, ev_callback,
3178ev_arg, ev_fd, ev_res, ev_events. 4021ev_arg, ev_fd, ev_res, ev_events.
3184=item * Priorities are not currently supported. Initialising priorities 4027=item * Priorities are not currently supported. Initialising priorities
3185will fail and all watchers will have the same priority, even though there 4028will fail and all watchers will have the same priority, even though there
3186is an ev_pri field. 4029is an ev_pri field.
3187 4030
3188=item * In libevent, the last base created gets the signals, in libev, the 4031=item * In libevent, the last base created gets the signals, in libev, the
3189first base created (== the default loop) gets the signals. 4032base that registered the signal gets the signals.
3190 4033
3191=item * Other members are not supported. 4034=item * Other members are not supported.
3192 4035
3193=item * The libev emulation is I<not> ABI compatible to libevent, you need 4036=item * The libev emulation is I<not> ABI compatible to libevent, you need
3194to use the libev header file and library. 4037to use the libev header file and library.
3195 4038
3196=back 4039=back
3197 4040
3198=head1 C++ SUPPORT 4041=head1 C++ SUPPORT
4042
4043=head2 C API
4044
4045The normal C API should work fine when used from C++: both ev.h and the
4046libev sources can be compiled as C++. Therefore, code that uses the C API
4047will work fine.
4048
4049Proper exception specifications might have to be added to callbacks passed
4050to libev: exceptions may be thrown only from watcher callbacks, all other
4051callbacks (allocator, syserr, loop acquire/release and periodic reschedule
4052callbacks) must not throw exceptions, and might need a C<noexcept>
4053specification. If you have code that needs to be compiled as both C and
4054C++ you can use the C<EV_NOEXCEPT> macro for this:
4055
4056 static void
4057 fatal_error (const char *msg) EV_NOEXCEPT
4058 {
4059 perror (msg);
4060 abort ();
4061 }
4062
4063 ...
4064 ev_set_syserr_cb (fatal_error);
4065
4066The only API functions that can currently throw exceptions are C<ev_run>,
4067C<ev_invoke>, C<ev_invoke_pending> and C<ev_loop_destroy> (the latter
4068because it runs cleanup watchers).
4069
4070Throwing exceptions in watcher callbacks is only supported if libev itself
4071is compiled with a C++ compiler or your C and C++ environments allow
4072throwing exceptions through C libraries (most do).
4073
4074=head2 C++ API
3199 4075
3200Libev comes with some simplistic wrapper classes for C++ that mainly allow 4076Libev comes with some simplistic wrapper classes for C++ that mainly allow
3201you to use some convenience methods to start/stop watchers and also change 4077you to use some convenience methods to start/stop watchers and also change
3202the callback model to a model using method callbacks on objects. 4078the callback model to a model using method callbacks on objects.
3203 4079
3204To use it, 4080To use it,
3205 4081
3206 #include <ev++.h> 4082 #include <ev++.h>
3207 4083
3208This automatically includes F<ev.h> and puts all of its definitions (many 4084This automatically includes F<ev.h> and puts all of its definitions (many
3209of them macros) into the global namespace. All C++ specific things are 4085of them macros) into the global namespace. All C++ specific things are
3210put into the C<ev> namespace. It should support all the same embedding 4086put into the C<ev> namespace. It should support all the same embedding
3213Care has been taken to keep the overhead low. The only data member the C++ 4089Care has been taken to keep the overhead low. The only data member the C++
3214classes add (compared to plain C-style watchers) is the event loop pointer 4090classes add (compared to plain C-style watchers) is the event loop pointer
3215that the watcher is associated with (or no additional members at all if 4091that the watcher is associated with (or no additional members at all if
3216you disable C<EV_MULTIPLICITY> when embedding libev). 4092you disable C<EV_MULTIPLICITY> when embedding libev).
3217 4093
3218Currently, functions, and static and non-static member functions can be 4094Currently, functions, static and non-static member functions and classes
3219used as callbacks. Other types should be easy to add as long as they only 4095with C<operator ()> can be used as callbacks. Other types should be easy
3220need one additional pointer for context. If you need support for other 4096to add as long as they only need one additional pointer for context. If
3221types of functors please contact the author (preferably after implementing 4097you need support for other types of functors please contact the author
3222it). 4098(preferably after implementing it).
4099
4100For all this to work, your C++ compiler either has to use the same calling
4101conventions as your C compiler (for static member functions), or you have
4102to embed libev and compile libev itself as C++.
3223 4103
3224Here is a list of things available in the C<ev> namespace: 4104Here is a list of things available in the C<ev> namespace:
3225 4105
3226=over 4 4106=over 4
3227 4107
3237=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc. 4117=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
3238 4118
3239For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of 4119For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
3240the same name in the C<ev> namespace, with the exception of C<ev_signal> 4120the same name in the C<ev> namespace, with the exception of C<ev_signal>
3241which is called C<ev::sig> to avoid clashes with the C<signal> macro 4121which is called C<ev::sig> to avoid clashes with the C<signal> macro
3242defines by many implementations. 4122defined by many implementations.
3243 4123
3244All of those classes have these methods: 4124All of those classes have these methods:
3245 4125
3246=over 4 4126=over 4
3247 4127
3288 myclass obj; 4168 myclass obj;
3289 ev::io iow; 4169 ev::io iow;
3290 iow.set <myclass, &myclass::io_cb> (&obj); 4170 iow.set <myclass, &myclass::io_cb> (&obj);
3291 4171
3292=item w->set (object *) 4172=item w->set (object *)
3293
3294This is an B<experimental> feature that might go away in a future version.
3295 4173
3296This is a variation of a method callback - leaving out the method to call 4174This is a variation of a method callback - leaving out the method to call
3297will default the method to C<operator ()>, which makes it possible to use 4175will default the method to C<operator ()>, which makes it possible to use
3298functor objects without having to manually specify the C<operator ()> all 4176functor objects without having to manually specify the C<operator ()> all
3299the time. Incidentally, you can then also leave out the template argument 4177the time. Incidentally, you can then also leave out the template argument
3311 void operator() (ev::io &w, int revents) 4189 void operator() (ev::io &w, int revents)
3312 { 4190 {
3313 ... 4191 ...
3314 } 4192 }
3315 } 4193 }
3316 4194
3317 myfunctor f; 4195 myfunctor f;
3318 4196
3319 ev::io w; 4197 ev::io w;
3320 w.set (&f); 4198 w.set (&f);
3321 4199
3339Associates a different C<struct ev_loop> with this watcher. You can only 4217Associates a different C<struct ev_loop> with this watcher. You can only
3340do this when the watcher is inactive (and not pending either). 4218do this when the watcher is inactive (and not pending either).
3341 4219
3342=item w->set ([arguments]) 4220=item w->set ([arguments])
3343 4221
3344Basically the same as C<ev_TYPE_set>, with the same arguments. Must be 4222Basically the same as C<ev_TYPE_set> (except for C<ev::embed> watchers>),
4223with the same arguments. Either this method or a suitable start method
3345called at least once. Unlike the C counterpart, an active watcher gets 4224must be called at least once. Unlike the C counterpart, an active watcher
3346automatically stopped and restarted when reconfiguring it with this 4225gets automatically stopped and restarted when reconfiguring it with this
3347method. 4226method.
4227
4228For C<ev::embed> watchers this method is called C<set_embed>, to avoid
4229clashing with the C<set (loop)> method.
3348 4230
3349=item w->start () 4231=item w->start ()
3350 4232
3351Starts the watcher. Note that there is no C<loop> argument, as the 4233Starts the watcher. Note that there is no C<loop> argument, as the
3352constructor already stores the event loop. 4234constructor already stores the event loop.
3353 4235
4236=item w->start ([arguments])
4237
4238Instead of calling C<set> and C<start> methods separately, it is often
4239convenient to wrap them in one call. Uses the same type of arguments as
4240the configure C<set> method of the watcher.
4241
3354=item w->stop () 4242=item w->stop ()
3355 4243
3356Stops the watcher if it is active. Again, no C<loop> argument. 4244Stops the watcher if it is active. Again, no C<loop> argument.
3357 4245
3358=item w->again () (C<ev::timer>, C<ev::periodic> only) 4246=item w->again () (C<ev::timer>, C<ev::periodic> only)
3370 4258
3371=back 4259=back
3372 4260
3373=back 4261=back
3374 4262
3375Example: Define a class with an IO and idle watcher, start one of them in 4263Example: Define a class with two I/O and idle watchers, start the I/O
3376the constructor. 4264watchers in the constructor.
3377 4265
3378 class myclass 4266 class myclass
3379 { 4267 {
3380 ev::io io ; void io_cb (ev::io &w, int revents); 4268 ev::io io ; void io_cb (ev::io &w, int revents);
4269 ev::io io2 ; void io2_cb (ev::io &w, int revents);
3381 ev::idle idle; void idle_cb (ev::idle &w, int revents); 4270 ev::idle idle; void idle_cb (ev::idle &w, int revents);
3382 4271
3383 myclass (int fd) 4272 myclass (int fd)
3384 { 4273 {
3385 io .set <myclass, &myclass::io_cb > (this); 4274 io .set <myclass, &myclass::io_cb > (this);
4275 io2 .set <myclass, &myclass::io2_cb > (this);
3386 idle.set <myclass, &myclass::idle_cb> (this); 4276 idle.set <myclass, &myclass::idle_cb> (this);
3387 4277
3388 io.start (fd, ev::READ); 4278 io.set (fd, ev::WRITE); // configure the watcher
4279 io.start (); // start it whenever convenient
4280
4281 io2.start (fd, ev::READ); // set + start in one call
3389 } 4282 }
3390 }; 4283 };
3391 4284
3392 4285
3393=head1 OTHER LANGUAGE BINDINGS 4286=head1 OTHER LANGUAGE BINDINGS
3432L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>. 4325L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
3433 4326
3434=item D 4327=item D
3435 4328
3436Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to 4329Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
3437be found at L<http://proj.llucax.com.ar/wiki/evd>. 4330be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
3438 4331
3439=item Ocaml 4332=item Ocaml
3440 4333
3441Erkki Seppala has written Ocaml bindings for libev, to be found at 4334Erkki Seppala has written Ocaml bindings for libev, to be found at
3442L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>. 4335L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
3443 4336
3444=item Lua 4337=item Lua
3445 4338
3446Brian Maher has written a partial interface to libev 4339Brian Maher has written a partial interface to libev for lua (at the
3447for lua (only C<ev_io> and C<ev_timer>), to be found at 4340time of this writing, only C<ev_io> and C<ev_timer>), to be found at
3448L<http://github.com/brimworks/lua-ev>. 4341L<http://github.com/brimworks/lua-ev>.
4342
4343=item Javascript
4344
4345Node.js (L<http://nodejs.org>) uses libev as the underlying event library.
4346
4347=item Others
4348
4349There are others, and I stopped counting.
3449 4350
3450=back 4351=back
3451 4352
3452 4353
3453=head1 MACRO MAGIC 4354=head1 MACRO MAGIC
3467loop argument"). The C<EV_A> form is used when this is the sole argument, 4368loop argument"). The C<EV_A> form is used when this is the sole argument,
3468C<EV_A_> is used when other arguments are following. Example: 4369C<EV_A_> is used when other arguments are following. Example:
3469 4370
3470 ev_unref (EV_A); 4371 ev_unref (EV_A);
3471 ev_timer_add (EV_A_ watcher); 4372 ev_timer_add (EV_A_ watcher);
3472 ev_loop (EV_A_ 0); 4373 ev_run (EV_A_ 0);
3473 4374
3474It assumes the variable C<loop> of type C<struct ev_loop *> is in scope, 4375It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
3475which is often provided by the following macro. 4376which is often provided by the following macro.
3476 4377
3477=item C<EV_P>, C<EV_P_> 4378=item C<EV_P>, C<EV_P_>
3490suitable for use with C<EV_A>. 4391suitable for use with C<EV_A>.
3491 4392
3492=item C<EV_DEFAULT>, C<EV_DEFAULT_> 4393=item C<EV_DEFAULT>, C<EV_DEFAULT_>
3493 4394
3494Similar to the other two macros, this gives you the value of the default 4395Similar to the other two macros, this gives you the value of the default
3495loop, if multiple loops are supported ("ev loop default"). 4396loop, if multiple loops are supported ("ev loop default"). The default loop
4397will be initialised if it isn't already initialised.
4398
4399For non-multiplicity builds, these macros do nothing, so you always have
4400to initialise the loop somewhere.
3496 4401
3497=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_> 4402=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
3498 4403
3499Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the 4404Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
3500default loop has been initialised (C<UC> == unchecked). Their behaviour 4405default loop has been initialised (C<UC> == unchecked). Their behaviour
3517 } 4422 }
3518 4423
3519 ev_check check; 4424 ev_check check;
3520 ev_check_init (&check, check_cb); 4425 ev_check_init (&check, check_cb);
3521 ev_check_start (EV_DEFAULT_ &check); 4426 ev_check_start (EV_DEFAULT_ &check);
3522 ev_loop (EV_DEFAULT_ 0); 4427 ev_run (EV_DEFAULT_ 0);
3523 4428
3524=head1 EMBEDDING 4429=head1 EMBEDDING
3525 4430
3526Libev can (and often is) directly embedded into host 4431Libev can (and often is) directly embedded into host
3527applications. Examples of applications that embed it include the Deliantra 4432applications. Examples of applications that embed it include the Deliantra
3567 ev_vars.h 4472 ev_vars.h
3568 ev_wrap.h 4473 ev_wrap.h
3569 4474
3570 ev_win32.c required on win32 platforms only 4475 ev_win32.c required on win32 platforms only
3571 4476
3572 ev_select.c only when select backend is enabled (which is enabled by default) 4477 ev_select.c only when select backend is enabled
3573 ev_poll.c only when poll backend is enabled (disabled by default) 4478 ev_poll.c only when poll backend is enabled
3574 ev_epoll.c only when the epoll backend is enabled (disabled by default) 4479 ev_epoll.c only when the epoll backend is enabled
4480 ev_linuxaio.c only when the linux aio backend is enabled
3575 ev_kqueue.c only when the kqueue backend is enabled (disabled by default) 4481 ev_kqueue.c only when the kqueue backend is enabled
3576 ev_port.c only when the solaris port backend is enabled (disabled by default) 4482 ev_port.c only when the solaris port backend is enabled
3577 4483
3578F<ev.c> includes the backend files directly when enabled, so you only need 4484F<ev.c> includes the backend files directly when enabled, so you only need
3579to compile this single file. 4485to compile this single file.
3580 4486
3581=head3 LIBEVENT COMPATIBILITY API 4487=head3 LIBEVENT COMPATIBILITY API
3607 libev.m4 4513 libev.m4
3608 4514
3609=head2 PREPROCESSOR SYMBOLS/MACROS 4515=head2 PREPROCESSOR SYMBOLS/MACROS
3610 4516
3611Libev can be configured via a variety of preprocessor symbols you have to 4517Libev can be configured via a variety of preprocessor symbols you have to
3612define before including any of its files. The default in the absence of 4518define before including (or compiling) any of its files. The default in
3613autoconf is documented for every option. 4519the absence of autoconf is documented for every option.
4520
4521Symbols marked with "(h)" do not change the ABI, and can have different
4522values when compiling libev vs. including F<ev.h>, so it is permissible
4523to redefine them before including F<ev.h> without breaking compatibility
4524to a compiled library. All other symbols change the ABI, which means all
4525users of libev and the libev code itself must be compiled with compatible
4526settings.
3614 4527
3615=over 4 4528=over 4
3616 4529
4530=item EV_COMPAT3 (h)
4531
4532Backwards compatibility is a major concern for libev. This is why this
4533release of libev comes with wrappers for the functions and symbols that
4534have been renamed between libev version 3 and 4.
4535
4536You can disable these wrappers (to test compatibility with future
4537versions) by defining C<EV_COMPAT3> to C<0> when compiling your
4538sources. This has the additional advantage that you can drop the C<struct>
4539from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
4540typedef in that case.
4541
4542In some future version, the default for C<EV_COMPAT3> will become C<0>,
4543and in some even more future version the compatibility code will be
4544removed completely.
4545
3617=item EV_STANDALONE 4546=item EV_STANDALONE (h)
3618 4547
3619Must always be C<1> if you do not use autoconf configuration, which 4548Must always be C<1> if you do not use autoconf configuration, which
3620keeps libev from including F<config.h>, and it also defines dummy 4549keeps libev from including F<config.h>, and it also defines dummy
3621implementations for some libevent functions (such as logging, which is not 4550implementations for some libevent functions (such as logging, which is not
3622supported). It will also not define any of the structs usually found in 4551supported). It will also not define any of the structs usually found in
3623F<event.h> that are not directly supported by the libev core alone. 4552F<event.h> that are not directly supported by the libev core alone.
3624 4553
3625In standalone mode, libev will still try to automatically deduce the 4554In standalone mode, libev will still try to automatically deduce the
3626configuration, but has to be more conservative. 4555configuration, but has to be more conservative.
4556
4557=item EV_USE_FLOOR
4558
4559If defined to be C<1>, libev will use the C<floor ()> function for its
4560periodic reschedule calculations, otherwise libev will fall back on a
4561portable (slower) implementation. If you enable this, you usually have to
4562link against libm or something equivalent. Enabling this when the C<floor>
4563function is not available will fail, so the safe default is to not enable
4564this.
3627 4565
3628=item EV_USE_MONOTONIC 4566=item EV_USE_MONOTONIC
3629 4567
3630If defined to be C<1>, libev will try to detect the availability of the 4568If defined to be C<1>, libev will try to detect the availability of the
3631monotonic clock option at both compile time and runtime. Otherwise no 4569monotonic clock option at both compile time and runtime. Otherwise no
3717If programs implement their own fd to handle mapping on win32, then this 4655If programs implement their own fd to handle mapping on win32, then this
3718macro can be used to override the C<close> function, useful to unregister 4656macro can be used to override the C<close> function, useful to unregister
3719file descriptors again. Note that the replacement function has to close 4657file descriptors again. Note that the replacement function has to close
3720the underlying OS handle. 4658the underlying OS handle.
3721 4659
4660=item EV_USE_WSASOCKET
4661
4662If defined to be C<1>, libev will use C<WSASocket> to create its internal
4663communication socket, which works better in some environments. Otherwise,
4664the normal C<socket> function will be used, which works better in other
4665environments.
4666
3722=item EV_USE_POLL 4667=item EV_USE_POLL
3723 4668
3724If defined to be C<1>, libev will compile in support for the C<poll>(2) 4669If defined to be C<1>, libev will compile in support for the C<poll>(2)
3725backend. Otherwise it will be enabled on non-win32 platforms. It 4670backend. Otherwise it will be enabled on non-win32 platforms. It
3726takes precedence over select. 4671takes precedence over select.
3730If defined to be C<1>, libev will compile in support for the Linux 4675If defined to be C<1>, libev will compile in support for the Linux
3731C<epoll>(7) backend. Its availability will be detected at runtime, 4676C<epoll>(7) backend. Its availability will be detected at runtime,
3732otherwise another method will be used as fallback. This is the preferred 4677otherwise another method will be used as fallback. This is the preferred
3733backend for GNU/Linux systems. If undefined, it will be enabled if the 4678backend for GNU/Linux systems. If undefined, it will be enabled if the
3734headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled. 4679headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
4680
4681=item EV_USE_LINUXAIO
4682
4683If defined to be C<1>, libev will compile in support for the Linux
4684aio backend. Due to it's currenbt limitations it has to be requested
4685explicitly. If undefined, it will be enabled on linux, otherwise
4686disabled.
3735 4687
3736=item EV_USE_KQUEUE 4688=item EV_USE_KQUEUE
3737 4689
3738If defined to be C<1>, libev will compile in support for the BSD style 4690If defined to be C<1>, libev will compile in support for the BSD style
3739C<kqueue>(2) backend. Its actual availability will be detected at runtime, 4691C<kqueue>(2) backend. Its actual availability will be detected at runtime,
3761If defined to be C<1>, libev will compile in support for the Linux inotify 4713If defined to be C<1>, libev will compile in support for the Linux inotify
3762interface to speed up C<ev_stat> watchers. Its actual availability will 4714interface to speed up C<ev_stat> watchers. Its actual availability will
3763be detected at runtime. If undefined, it will be enabled if the headers 4715be detected at runtime. If undefined, it will be enabled if the headers
3764indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled. 4716indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
3765 4717
4718=item EV_NO_SMP
4719
4720If defined to be C<1>, libev will assume that memory is always coherent
4721between threads, that is, threads can be used, but threads never run on
4722different cpus (or different cpu cores). This reduces dependencies
4723and makes libev faster.
4724
4725=item EV_NO_THREADS
4726
4727If defined to be C<1>, libev will assume that it will never be called from
4728different threads (that includes signal handlers), which is a stronger
4729assumption than C<EV_NO_SMP>, above. This reduces dependencies and makes
4730libev faster.
4731
3766=item EV_ATOMIC_T 4732=item EV_ATOMIC_T
3767 4733
3768Libev requires an integer type (suitable for storing C<0> or C<1>) whose 4734Libev requires an integer type (suitable for storing C<0> or C<1>) whose
3769access is atomic with respect to other threads or signal contexts. No such 4735access is atomic with respect to other threads or signal contexts. No
3770type is easily found in the C language, so you can provide your own type 4736such type is easily found in the C language, so you can provide your own
3771that you know is safe for your purposes. It is used both for signal handler "locking" 4737type that you know is safe for your purposes. It is used both for signal
3772as well as for signal and thread safety in C<ev_async> watchers. 4738handler "locking" as well as for signal and thread safety in C<ev_async>
4739watchers.
3773 4740
3774In the absence of this define, libev will use C<sig_atomic_t volatile> 4741In the absence of this define, libev will use C<sig_atomic_t volatile>
3775(from F<signal.h>), which is usually good enough on most platforms. 4742(from F<signal.h>), which is usually good enough on most platforms.
3776 4743
3777=item EV_H 4744=item EV_H (h)
3778 4745
3779The name of the F<ev.h> header file used to include it. The default if 4746The name of the F<ev.h> header file used to include it. The default if
3780undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be 4747undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
3781used to virtually rename the F<ev.h> header file in case of conflicts. 4748used to virtually rename the F<ev.h> header file in case of conflicts.
3782 4749
3783=item EV_CONFIG_H 4750=item EV_CONFIG_H (h)
3784 4751
3785If C<EV_STANDALONE> isn't C<1>, this variable can be used to override 4752If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
3786F<ev.c>'s idea of where to find the F<config.h> file, similarly to 4753F<ev.c>'s idea of where to find the F<config.h> file, similarly to
3787C<EV_H>, above. 4754C<EV_H>, above.
3788 4755
3789=item EV_EVENT_H 4756=item EV_EVENT_H (h)
3790 4757
3791Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea 4758Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
3792of how the F<event.h> header can be found, the default is C<"event.h">. 4759of how the F<event.h> header can be found, the default is C<"event.h">.
3793 4760
3794=item EV_PROTOTYPES 4761=item EV_PROTOTYPES (h)
3795 4762
3796If defined to be C<0>, then F<ev.h> will not define any function 4763If defined to be C<0>, then F<ev.h> will not define any function
3797prototypes, but still define all the structs and other symbols. This is 4764prototypes, but still define all the structs and other symbols. This is
3798occasionally useful if you want to provide your own wrapper functions 4765occasionally useful if you want to provide your own wrapper functions
3799around libev functions. 4766around libev functions.
3804will have the C<struct ev_loop *> as first argument, and you can create 4771will have the C<struct ev_loop *> as first argument, and you can create
3805additional independent event loops. Otherwise there will be no support 4772additional independent event loops. Otherwise there will be no support
3806for multiple event loops and there is no first event loop pointer 4773for multiple event loops and there is no first event loop pointer
3807argument. Instead, all functions act on the single default loop. 4774argument. Instead, all functions act on the single default loop.
3808 4775
4776Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
4777default loop when multiplicity is switched off - you always have to
4778initialise the loop manually in this case.
4779
3809=item EV_MINPRI 4780=item EV_MINPRI
3810 4781
3811=item EV_MAXPRI 4782=item EV_MAXPRI
3812 4783
3813The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to 4784The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
3821fine. 4792fine.
3822 4793
3823If your embedding application does not need any priorities, defining these 4794If your embedding application does not need any priorities, defining these
3824both to C<0> will save some memory and CPU. 4795both to C<0> will save some memory and CPU.
3825 4796
3826=item EV_PERIODIC_ENABLE 4797=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
4798EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
4799EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
3827 4800
3828If undefined or defined to be C<1>, then periodic timers are supported. If 4801If undefined or defined to be C<1> (and the platform supports it), then
3829defined to be C<0>, then they are not. Disabling them saves a few kB of 4802the respective watcher type is supported. If defined to be C<0>, then it
3830code. 4803is not. Disabling watcher types mainly saves code size.
3831 4804
3832=item EV_IDLE_ENABLE 4805=item EV_FEATURES
3833
3834If undefined or defined to be C<1>, then idle watchers are supported. If
3835defined to be C<0>, then they are not. Disabling them saves a few kB of
3836code.
3837
3838=item EV_EMBED_ENABLE
3839
3840If undefined or defined to be C<1>, then embed watchers are supported. If
3841defined to be C<0>, then they are not. Embed watchers rely on most other
3842watcher types, which therefore must not be disabled.
3843
3844=item EV_STAT_ENABLE
3845
3846If undefined or defined to be C<1>, then stat watchers are supported. If
3847defined to be C<0>, then they are not.
3848
3849=item EV_FORK_ENABLE
3850
3851If undefined or defined to be C<1>, then fork watchers are supported. If
3852defined to be C<0>, then they are not.
3853
3854=item EV_ASYNC_ENABLE
3855
3856If undefined or defined to be C<1>, then async watchers are supported. If
3857defined to be C<0>, then they are not.
3858
3859=item EV_MINIMAL
3860 4806
3861If you need to shave off some kilobytes of code at the expense of some 4807If you need to shave off some kilobytes of code at the expense of some
3862speed (but with the full API), define this symbol to C<1>. Currently this 4808speed (but with the full API), you can define this symbol to request
3863is used to override some inlining decisions, saves roughly 30% code size 4809certain subsets of functionality. The default is to enable all features
3864on amd64. It also selects a much smaller 2-heap for timer management over 4810that can be enabled on the platform.
3865the default 4-heap.
3866 4811
3867You can save even more by disabling watcher types you do not need 4812A typical way to use this symbol is to define it to C<0> (or to a bitset
3868and setting C<EV_MAXPRI> == C<EV_MINPRI>. Also, disabling C<assert> 4813with some broad features you want) and then selectively re-enable
3869(C<-DNDEBUG>) will usually reduce code size a lot. 4814additional parts you want, for example if you want everything minimal,
4815but multiple event loop support, async and child watchers and the poll
4816backend, use this:
3870 4817
3871Defining C<EV_MINIMAL> to C<2> will additionally reduce the core API to 4818 #define EV_FEATURES 0
3872provide a bare-bones event library. See C<ev.h> for details on what parts 4819 #define EV_MULTIPLICITY 1
3873of the API are still available, and do not complain if this subset changes 4820 #define EV_USE_POLL 1
3874over time. 4821 #define EV_CHILD_ENABLE 1
4822 #define EV_ASYNC_ENABLE 1
4823
4824The actual value is a bitset, it can be a combination of the following
4825values (by default, all of these are enabled):
4826
4827=over 4
4828
4829=item C<1> - faster/larger code
4830
4831Use larger code to speed up some operations.
4832
4833Currently this is used to override some inlining decisions (enlarging the
4834code size by roughly 30% on amd64).
4835
4836When optimising for size, use of compiler flags such as C<-Os> with
4837gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
4838assertions.
4839
4840The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
4841(e.g. gcc with C<-Os>).
4842
4843=item C<2> - faster/larger data structures
4844
4845Replaces the small 2-heap for timer management by a faster 4-heap, larger
4846hash table sizes and so on. This will usually further increase code size
4847and can additionally have an effect on the size of data structures at
4848runtime.
4849
4850The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
4851(e.g. gcc with C<-Os>).
4852
4853=item C<4> - full API configuration
4854
4855This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
4856enables multiplicity (C<EV_MULTIPLICITY>=1).
4857
4858=item C<8> - full API
4859
4860This enables a lot of the "lesser used" API functions. See C<ev.h> for
4861details on which parts of the API are still available without this
4862feature, and do not complain if this subset changes over time.
4863
4864=item C<16> - enable all optional watcher types
4865
4866Enables all optional watcher types. If you want to selectively enable
4867only some watcher types other than I/O and timers (e.g. prepare,
4868embed, async, child...) you can enable them manually by defining
4869C<EV_watchertype_ENABLE> to C<1> instead.
4870
4871=item C<32> - enable all backends
4872
4873This enables all backends - without this feature, you need to enable at
4874least one backend manually (C<EV_USE_SELECT> is a good choice).
4875
4876=item C<64> - enable OS-specific "helper" APIs
4877
4878Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
4879default.
4880
4881=back
4882
4883Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
4884reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
4885code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
4886watchers, timers and monotonic clock support.
4887
4888With an intelligent-enough linker (gcc+binutils are intelligent enough
4889when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
4890your program might be left out as well - a binary starting a timer and an
4891I/O watcher then might come out at only 5Kb.
4892
4893=item EV_API_STATIC
4894
4895If this symbol is defined (by default it is not), then all identifiers
4896will have static linkage. This means that libev will not export any
4897identifiers, and you cannot link against libev anymore. This can be useful
4898when you embed libev, only want to use libev functions in a single file,
4899and do not want its identifiers to be visible.
4900
4901To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
4902wants to use libev.
4903
4904This option only works when libev is compiled with a C compiler, as C++
4905doesn't support the required declaration syntax.
4906
4907=item EV_AVOID_STDIO
4908
4909If this is set to C<1> at compiletime, then libev will avoid using stdio
4910functions (printf, scanf, perror etc.). This will increase the code size
4911somewhat, but if your program doesn't otherwise depend on stdio and your
4912libc allows it, this avoids linking in the stdio library which is quite
4913big.
4914
4915Note that error messages might become less precise when this option is
4916enabled.
3875 4917
3876=item EV_NSIG 4918=item EV_NSIG
3877 4919
3878The highest supported signal number, +1 (or, the number of 4920The highest supported signal number, +1 (or, the number of
3879signals): Normally, libev tries to deduce the maximum number of signals 4921signals): Normally, libev tries to deduce the maximum number of signals
3880automatically, but sometimes this fails, in which case it can be 4922automatically, but sometimes this fails, in which case it can be
3881specified. Also, using a lower number than detected (C<32> should be 4923specified. Also, using a lower number than detected (C<32> should be
3882good for about any system in existance) can save some memory, as libev 4924good for about any system in existence) can save some memory, as libev
3883statically allocates some 12-24 bytes per signal number. 4925statically allocates some 12-24 bytes per signal number.
3884 4926
3885=item EV_PID_HASHSIZE 4927=item EV_PID_HASHSIZE
3886 4928
3887C<ev_child> watchers use a small hash table to distribute workload by 4929C<ev_child> watchers use a small hash table to distribute workload by
3888pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more 4930pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
3889than enough. If you need to manage thousands of children you might want to 4931usually more than enough. If you need to manage thousands of children you
3890increase this value (I<must> be a power of two). 4932might want to increase this value (I<must> be a power of two).
3891 4933
3892=item EV_INOTIFY_HASHSIZE 4934=item EV_INOTIFY_HASHSIZE
3893 4935
3894C<ev_stat> watchers use a small hash table to distribute workload by 4936C<ev_stat> watchers use a small hash table to distribute workload by
3895inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>), 4937inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
3896usually more than enough. If you need to manage thousands of C<ev_stat> 4938disabled), usually more than enough. If you need to manage thousands of
3897watchers you might want to increase this value (I<must> be a power of 4939C<ev_stat> watchers you might want to increase this value (I<must> be a
3898two). 4940power of two).
3899 4941
3900=item EV_USE_4HEAP 4942=item EV_USE_4HEAP
3901 4943
3902Heaps are not very cache-efficient. To improve the cache-efficiency of the 4944Heaps are not very cache-efficient. To improve the cache-efficiency of the
3903timer and periodics heaps, libev uses a 4-heap when this symbol is defined 4945timer and periodics heaps, libev uses a 4-heap when this symbol is defined
3904to C<1>. The 4-heap uses more complicated (longer) code but has noticeably 4946to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
3905faster performance with many (thousands) of watchers. 4947faster performance with many (thousands) of watchers.
3906 4948
3907The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0> 4949The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
3908(disabled). 4950will be C<0>.
3909 4951
3910=item EV_HEAP_CACHE_AT 4952=item EV_HEAP_CACHE_AT
3911 4953
3912Heaps are not very cache-efficient. To improve the cache-efficiency of the 4954Heaps are not very cache-efficient. To improve the cache-efficiency of the
3913timer and periodics heaps, libev can cache the timestamp (I<at>) within 4955timer and periodics heaps, libev can cache the timestamp (I<at>) within
3914the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>), 4956the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
3915which uses 8-12 bytes more per watcher and a few hundred bytes more code, 4957which uses 8-12 bytes more per watcher and a few hundred bytes more code,
3916but avoids random read accesses on heap changes. This improves performance 4958but avoids random read accesses on heap changes. This improves performance
3917noticeably with many (hundreds) of watchers. 4959noticeably with many (hundreds) of watchers.
3918 4960
3919The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0> 4961The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
3920(disabled). 4962will be C<0>.
3921 4963
3922=item EV_VERIFY 4964=item EV_VERIFY
3923 4965
3924Controls how much internal verification (see C<ev_loop_verify ()>) will 4966Controls how much internal verification (see C<ev_verify ()>) will
3925be done: If set to C<0>, no internal verification code will be compiled 4967be done: If set to C<0>, no internal verification code will be compiled
3926in. If set to C<1>, then verification code will be compiled in, but not 4968in. If set to C<1>, then verification code will be compiled in, but not
3927called. If set to C<2>, then the internal verification code will be 4969called. If set to C<2>, then the internal verification code will be
3928called once per loop, which can slow down libev. If set to C<3>, then the 4970called once per loop, which can slow down libev. If set to C<3>, then the
3929verification code will be called very frequently, which will slow down 4971verification code will be called very frequently, which will slow down
3930libev considerably. 4972libev considerably.
3931 4973
3932The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be 4974The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
3933C<0>. 4975will be C<0>.
3934 4976
3935=item EV_COMMON 4977=item EV_COMMON
3936 4978
3937By default, all watchers have a C<void *data> member. By redefining 4979By default, all watchers have a C<void *data> member. By redefining
3938this macro to a something else you can include more and other types of 4980this macro to something else you can include more and other types of
3939members. You have to define it each time you include one of the files, 4981members. You have to define it each time you include one of the files,
3940though, and it must be identical each time. 4982though, and it must be identical each time.
3941 4983
3942For example, the perl EV module uses something like this: 4984For example, the perl EV module uses something like this:
3943 4985
3996file. 5038file.
3997 5039
3998The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file 5040The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
3999that everybody includes and which overrides some configure choices: 5041that everybody includes and which overrides some configure choices:
4000 5042
4001 #define EV_MINIMAL 1 5043 #define EV_FEATURES 8
4002 #define EV_USE_POLL 0 5044 #define EV_USE_SELECT 1
4003 #define EV_MULTIPLICITY 0
4004 #define EV_PERIODIC_ENABLE 0 5045 #define EV_PREPARE_ENABLE 1
5046 #define EV_IDLE_ENABLE 1
4005 #define EV_STAT_ENABLE 0 5047 #define EV_SIGNAL_ENABLE 1
4006 #define EV_FORK_ENABLE 0 5048 #define EV_CHILD_ENABLE 1
5049 #define EV_USE_STDEXCEPT 0
4007 #define EV_CONFIG_H <config.h> 5050 #define EV_CONFIG_H <config.h>
4008 #define EV_MINPRI 0
4009 #define EV_MAXPRI 0
4010 5051
4011 #include "ev++.h" 5052 #include "ev++.h"
4012 5053
4013And a F<ev_cpp.C> implementation file that contains libev proper and is compiled: 5054And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
4014 5055
4015 #include "ev_cpp.h" 5056 #include "ev_cpp.h"
4016 #include "ev.c" 5057 #include "ev.c"
4017 5058
4018=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES 5059=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
4019 5060
4020=head2 THREADS AND COROUTINES 5061=head2 THREADS AND COROUTINES
4021 5062
4022=head3 THREADS 5063=head3 THREADS
4023 5064
4074default loop and triggering an C<ev_async> watcher from the default loop 5115default loop and triggering an C<ev_async> watcher from the default loop
4075watcher callback into the event loop interested in the signal. 5116watcher callback into the event loop interested in the signal.
4076 5117
4077=back 5118=back
4078 5119
4079=head4 THREAD LOCKING EXAMPLE 5120See also L</THREAD LOCKING EXAMPLE>.
4080
4081Here is a fictitious example of how to run an event loop in a different
4082thread than where callbacks are being invoked and watchers are
4083created/added/removed.
4084
4085For a real-world example, see the C<EV::Loop::Async> perl module,
4086which uses exactly this technique (which is suited for many high-level
4087languages).
4088
4089The example uses a pthread mutex to protect the loop data, a condition
4090variable to wait for callback invocations, an async watcher to notify the
4091event loop thread and an unspecified mechanism to wake up the main thread.
4092
4093First, you need to associate some data with the event loop:
4094
4095 typedef struct {
4096 mutex_t lock; /* global loop lock */
4097 ev_async async_w;
4098 thread_t tid;
4099 cond_t invoke_cv;
4100 } userdata;
4101
4102 void prepare_loop (EV_P)
4103 {
4104 // for simplicity, we use a static userdata struct.
4105 static userdata u;
4106
4107 ev_async_init (&u->async_w, async_cb);
4108 ev_async_start (EV_A_ &u->async_w);
4109
4110 pthread_mutex_init (&u->lock, 0);
4111 pthread_cond_init (&u->invoke_cv, 0);
4112
4113 // now associate this with the loop
4114 ev_set_userdata (EV_A_ u);
4115 ev_set_invoke_pending_cb (EV_A_ l_invoke);
4116 ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
4117
4118 // then create the thread running ev_loop
4119 pthread_create (&u->tid, 0, l_run, EV_A);
4120 }
4121
4122The callback for the C<ev_async> watcher does nothing: the watcher is used
4123solely to wake up the event loop so it takes notice of any new watchers
4124that might have been added:
4125
4126 static void
4127 async_cb (EV_P_ ev_async *w, int revents)
4128 {
4129 // just used for the side effects
4130 }
4131
4132The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
4133protecting the loop data, respectively.
4134
4135 static void
4136 l_release (EV_P)
4137 {
4138 userdata *u = ev_userdata (EV_A);
4139 pthread_mutex_unlock (&u->lock);
4140 }
4141
4142 static void
4143 l_acquire (EV_P)
4144 {
4145 userdata *u = ev_userdata (EV_A);
4146 pthread_mutex_lock (&u->lock);
4147 }
4148
4149The event loop thread first acquires the mutex, and then jumps straight
4150into C<ev_loop>:
4151
4152 void *
4153 l_run (void *thr_arg)
4154 {
4155 struct ev_loop *loop = (struct ev_loop *)thr_arg;
4156
4157 l_acquire (EV_A);
4158 pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
4159 ev_loop (EV_A_ 0);
4160 l_release (EV_A);
4161
4162 return 0;
4163 }
4164
4165Instead of invoking all pending watchers, the C<l_invoke> callback will
4166signal the main thread via some unspecified mechanism (signals? pipe
4167writes? C<Async::Interrupt>?) and then waits until all pending watchers
4168have been called (in a while loop because a) spurious wakeups are possible
4169and b) skipping inter-thread-communication when there are no pending
4170watchers is very beneficial):
4171
4172 static void
4173 l_invoke (EV_P)
4174 {
4175 userdata *u = ev_userdata (EV_A);
4176
4177 while (ev_pending_count (EV_A))
4178 {
4179 wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
4180 pthread_cond_wait (&u->invoke_cv, &u->lock);
4181 }
4182 }
4183
4184Now, whenever the main thread gets told to invoke pending watchers, it
4185will grab the lock, call C<ev_invoke_pending> and then signal the loop
4186thread to continue:
4187
4188 static void
4189 real_invoke_pending (EV_P)
4190 {
4191 userdata *u = ev_userdata (EV_A);
4192
4193 pthread_mutex_lock (&u->lock);
4194 ev_invoke_pending (EV_A);
4195 pthread_cond_signal (&u->invoke_cv);
4196 pthread_mutex_unlock (&u->lock);
4197 }
4198
4199Whenever you want to start/stop a watcher or do other modifications to an
4200event loop, you will now have to lock:
4201
4202 ev_timer timeout_watcher;
4203 userdata *u = ev_userdata (EV_A);
4204
4205 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
4206
4207 pthread_mutex_lock (&u->lock);
4208 ev_timer_start (EV_A_ &timeout_watcher);
4209 ev_async_send (EV_A_ &u->async_w);
4210 pthread_mutex_unlock (&u->lock);
4211
4212Note that sending the C<ev_async> watcher is required because otherwise
4213an event loop currently blocking in the kernel will have no knowledge
4214about the newly added timer. By waking up the loop it will pick up any new
4215watchers in the next event loop iteration.
4216 5121
4217=head3 COROUTINES 5122=head3 COROUTINES
4218 5123
4219Libev is very accommodating to coroutines ("cooperative threads"): 5124Libev is very accommodating to coroutines ("cooperative threads"):
4220libev fully supports nesting calls to its functions from different 5125libev fully supports nesting calls to its functions from different
4221coroutines (e.g. you can call C<ev_loop> on the same loop from two 5126coroutines (e.g. you can call C<ev_run> on the same loop from two
4222different coroutines, and switch freely between both coroutines running 5127different coroutines, and switch freely between both coroutines running
4223the loop, as long as you don't confuse yourself). The only exception is 5128the loop, as long as you don't confuse yourself). The only exception is
4224that you must not do this from C<ev_periodic> reschedule callbacks. 5129that you must not do this from C<ev_periodic> reschedule callbacks.
4225 5130
4226Care has been taken to ensure that libev does not keep local state inside 5131Care has been taken to ensure that libev does not keep local state inside
4227C<ev_loop>, and other calls do not usually allow for coroutine switches as 5132C<ev_run>, and other calls do not usually allow for coroutine switches as
4228they do not call any callbacks. 5133they do not call any callbacks.
4229 5134
4230=head2 COMPILER WARNINGS 5135=head2 COMPILER WARNINGS
4231 5136
4232Depending on your compiler and compiler settings, you might get no or a 5137Depending on your compiler and compiler settings, you might get no or a
4243maintainable. 5148maintainable.
4244 5149
4245And of course, some compiler warnings are just plain stupid, or simply 5150And of course, some compiler warnings are just plain stupid, or simply
4246wrong (because they don't actually warn about the condition their message 5151wrong (because they don't actually warn about the condition their message
4247seems to warn about). For example, certain older gcc versions had some 5152seems to warn about). For example, certain older gcc versions had some
4248warnings that resulted an extreme number of false positives. These have 5153warnings that resulted in an extreme number of false positives. These have
4249been fixed, but some people still insist on making code warn-free with 5154been fixed, but some people still insist on making code warn-free with
4250such buggy versions. 5155such buggy versions.
4251 5156
4252While libev is written to generate as few warnings as possible, 5157While libev is written to generate as few warnings as possible,
4253"warn-free" code is not a goal, and it is recommended not to build libev 5158"warn-free" code is not a goal, and it is recommended not to build libev
4289I suggest using suppression lists. 5194I suggest using suppression lists.
4290 5195
4291 5196
4292=head1 PORTABILITY NOTES 5197=head1 PORTABILITY NOTES
4293 5198
5199=head2 GNU/LINUX 32 BIT LIMITATIONS
5200
5201GNU/Linux is the only common platform that supports 64 bit file/large file
5202interfaces but I<disables> them by default.
5203
5204That means that libev compiled in the default environment doesn't support
5205files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
5206
5207Unfortunately, many programs try to work around this GNU/Linux issue
5208by enabling the large file API, which makes them incompatible with the
5209standard libev compiled for their system.
5210
5211Likewise, libev cannot enable the large file API itself as this would
5212suddenly make it incompatible to the default compile time environment,
5213i.e. all programs not using special compile switches.
5214
5215=head2 OS/X AND DARWIN BUGS
5216
5217The whole thing is a bug if you ask me - basically any system interface
5218you touch is broken, whether it is locales, poll, kqueue or even the
5219OpenGL drivers.
5220
5221=head3 C<kqueue> is buggy
5222
5223The kqueue syscall is broken in all known versions - most versions support
5224only sockets, many support pipes.
5225
5226Libev tries to work around this by not using C<kqueue> by default on this
5227rotten platform, but of course you can still ask for it when creating a
5228loop - embedding a socket-only kqueue loop into a select-based one is
5229probably going to work well.
5230
5231=head3 C<poll> is buggy
5232
5233Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
5234implementation by something calling C<kqueue> internally around the 10.5.6
5235release, so now C<kqueue> I<and> C<poll> are broken.
5236
5237Libev tries to work around this by not using C<poll> by default on
5238this rotten platform, but of course you can still ask for it when creating
5239a loop.
5240
5241=head3 C<select> is buggy
5242
5243All that's left is C<select>, and of course Apple found a way to fuck this
5244one up as well: On OS/X, C<select> actively limits the number of file
5245descriptors you can pass in to 1024 - your program suddenly crashes when
5246you use more.
5247
5248There is an undocumented "workaround" for this - defining
5249C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
5250work on OS/X.
5251
5252=head2 SOLARIS PROBLEMS AND WORKAROUNDS
5253
5254=head3 C<errno> reentrancy
5255
5256The default compile environment on Solaris is unfortunately so
5257thread-unsafe that you can't even use components/libraries compiled
5258without C<-D_REENTRANT> in a threaded program, which, of course, isn't
5259defined by default. A valid, if stupid, implementation choice.
5260
5261If you want to use libev in threaded environments you have to make sure
5262it's compiled with C<_REENTRANT> defined.
5263
5264=head3 Event port backend
5265
5266The scalable event interface for Solaris is called "event
5267ports". Unfortunately, this mechanism is very buggy in all major
5268releases. If you run into high CPU usage, your program freezes or you get
5269a large number of spurious wakeups, make sure you have all the relevant
5270and latest kernel patches applied. No, I don't know which ones, but there
5271are multiple ones to apply, and afterwards, event ports actually work
5272great.
5273
5274If you can't get it to work, you can try running the program by setting
5275the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
5276C<select> backends.
5277
5278=head2 AIX POLL BUG
5279
5280AIX unfortunately has a broken C<poll.h> header. Libev works around
5281this by trying to avoid the poll backend altogether (i.e. it's not even
5282compiled in), which normally isn't a big problem as C<select> works fine
5283with large bitsets on AIX, and AIX is dead anyway.
5284
4294=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS 5285=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
5286
5287=head3 General issues
4295 5288
4296Win32 doesn't support any of the standards (e.g. POSIX) that libev 5289Win32 doesn't support any of the standards (e.g. POSIX) that libev
4297requires, and its I/O model is fundamentally incompatible with the POSIX 5290requires, and its I/O model is fundamentally incompatible with the POSIX
4298model. Libev still offers limited functionality on this platform in 5291model. Libev still offers limited functionality on this platform in
4299the form of the C<EVBACKEND_SELECT> backend, and only supports socket 5292the form of the C<EVBACKEND_SELECT> backend, and only supports socket
4300descriptors. This only applies when using Win32 natively, not when using 5293descriptors. This only applies when using Win32 natively, not when using
4301e.g. cygwin. 5294e.g. cygwin. Actually, it only applies to the microsofts own compilers,
5295as every compiler comes with a slightly differently broken/incompatible
5296environment.
4302 5297
4303Lifting these limitations would basically require the full 5298Lifting these limitations would basically require the full
4304re-implementation of the I/O system. If you are into these kinds of 5299re-implementation of the I/O system. If you are into this kind of thing,
4305things, then note that glib does exactly that for you in a very portable 5300then note that glib does exactly that for you in a very portable way (note
4306way (note also that glib is the slowest event library known to man). 5301also that glib is the slowest event library known to man).
4307 5302
4308There is no supported compilation method available on windows except 5303There is no supported compilation method available on windows except
4309embedding it into other applications. 5304embedding it into other applications.
4310 5305
4311Sensible signal handling is officially unsupported by Microsoft - libev 5306Sensible signal handling is officially unsupported by Microsoft - libev
4339you do I<not> compile the F<ev.c> or any other embedded source files!): 5334you do I<not> compile the F<ev.c> or any other embedded source files!):
4340 5335
4341 #include "evwrap.h" 5336 #include "evwrap.h"
4342 #include "ev.c" 5337 #include "ev.c"
4343 5338
4344=over 4
4345
4346=item The winsocket select function 5339=head3 The winsocket C<select> function
4347 5340
4348The winsocket C<select> function doesn't follow POSIX in that it 5341The winsocket C<select> function doesn't follow POSIX in that it
4349requires socket I<handles> and not socket I<file descriptors> (it is 5342requires socket I<handles> and not socket I<file descriptors> (it is
4350also extremely buggy). This makes select very inefficient, and also 5343also extremely buggy). This makes select very inefficient, and also
4351requires a mapping from file descriptors to socket handles (the Microsoft 5344requires a mapping from file descriptors to socket handles (the Microsoft
4360 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */ 5353 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
4361 5354
4362Note that winsockets handling of fd sets is O(n), so you can easily get a 5355Note that winsockets handling of fd sets is O(n), so you can easily get a
4363complexity in the O(n²) range when using win32. 5356complexity in the O(n²) range when using win32.
4364 5357
4365=item Limited number of file descriptors 5358=head3 Limited number of file descriptors
4366 5359
4367Windows has numerous arbitrary (and low) limits on things. 5360Windows has numerous arbitrary (and low) limits on things.
4368 5361
4369Early versions of winsocket's select only supported waiting for a maximum 5362Early versions of winsocket's select only supported waiting for a maximum
4370of C<64> handles (probably owning to the fact that all windows kernels 5363of C<64> handles (probably owning to the fact that all windows kernels
4385runtime libraries. This might get you to about C<512> or C<2048> sockets 5378runtime libraries. This might get you to about C<512> or C<2048> sockets
4386(depending on windows version and/or the phase of the moon). To get more, 5379(depending on windows version and/or the phase of the moon). To get more,
4387you need to wrap all I/O functions and provide your own fd management, but 5380you need to wrap all I/O functions and provide your own fd management, but
4388the cost of calling select (O(n²)) will likely make this unworkable. 5381the cost of calling select (O(n²)) will likely make this unworkable.
4389 5382
4390=back
4391
4392=head2 PORTABILITY REQUIREMENTS 5383=head2 PORTABILITY REQUIREMENTS
4393 5384
4394In addition to a working ISO-C implementation and of course the 5385In addition to a working ISO-C implementation and of course the
4395backend-specific APIs, libev relies on a few additional extensions: 5386backend-specific APIs, libev relies on a few additional extensions:
4396 5387
4402Libev assumes not only that all watcher pointers have the same internal 5393Libev assumes not only that all watcher pointers have the same internal
4403structure (guaranteed by POSIX but not by ISO C for example), but it also 5394structure (guaranteed by POSIX but not by ISO C for example), but it also
4404assumes that the same (machine) code can be used to call any watcher 5395assumes that the same (machine) code can be used to call any watcher
4405callback: The watcher callbacks have different type signatures, but libev 5396callback: The watcher callbacks have different type signatures, but libev
4406calls them using an C<ev_watcher *> internally. 5397calls them using an C<ev_watcher *> internally.
5398
5399=item null pointers and integer zero are represented by 0 bytes
5400
5401Libev uses C<memset> to initialise structs and arrays to C<0> bytes, and
5402relies on this setting pointers and integers to null.
5403
5404=item pointer accesses must be thread-atomic
5405
5406Accessing a pointer value must be atomic, it must both be readable and
5407writable in one piece - this is the case on all current architectures.
4407 5408
4408=item C<sig_atomic_t volatile> must be thread-atomic as well 5409=item C<sig_atomic_t volatile> must be thread-atomic as well
4409 5410
4410The type C<sig_atomic_t volatile> (or whatever is defined as 5411The type C<sig_atomic_t volatile> (or whatever is defined as
4411C<EV_ATOMIC_T>) must be atomic with respect to accesses from different 5412C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
4420thread" or will block signals process-wide, both behaviours would 5421thread" or will block signals process-wide, both behaviours would
4421be compatible with libev. Interaction between C<sigprocmask> and 5422be compatible with libev. Interaction between C<sigprocmask> and
4422C<pthread_sigmask> could complicate things, however. 5423C<pthread_sigmask> could complicate things, however.
4423 5424
4424The most portable way to handle signals is to block signals in all threads 5425The most portable way to handle signals is to block signals in all threads
4425except the initial one, and run the default loop in the initial thread as 5426except the initial one, and run the signal handling loop in the initial
4426well. 5427thread as well.
4427 5428
4428=item C<long> must be large enough for common memory allocation sizes 5429=item C<long> must be large enough for common memory allocation sizes
4429 5430
4430To improve portability and simplify its API, libev uses C<long> internally 5431To improve portability and simplify its API, libev uses C<long> internally
4431instead of C<size_t> when allocating its data structures. On non-POSIX 5432instead of C<size_t> when allocating its data structures. On non-POSIX
4434watchers. 5435watchers.
4435 5436
4436=item C<double> must hold a time value in seconds with enough accuracy 5437=item C<double> must hold a time value in seconds with enough accuracy
4437 5438
4438The type C<double> is used to represent timestamps. It is required to 5439The type C<double> is used to represent timestamps. It is required to
4439have at least 51 bits of mantissa (and 9 bits of exponent), which is good 5440have at least 51 bits of mantissa (and 9 bits of exponent), which is
4440enough for at least into the year 4000. This requirement is fulfilled by 5441good enough for at least into the year 4000 with millisecond accuracy
5442(the design goal for libev). This requirement is overfulfilled by
4441implementations implementing IEEE 754, which is basically all existing 5443implementations using IEEE 754, which is basically all existing ones.
5444
4442ones. With IEEE 754 doubles, you get microsecond accuracy until at least 5445With IEEE 754 doubles, you get microsecond accuracy until at least the
44432200. 5446year 2255 (and millisecond accuracy till the year 287396 - by then, libev
5447is either obsolete or somebody patched it to use C<long double> or
5448something like that, just kidding).
4444 5449
4445=back 5450=back
4446 5451
4447If you know of other additional requirements drop me a note. 5452If you know of other additional requirements drop me a note.
4448 5453
4510=item Processing ev_async_send: O(number_of_async_watchers) 5515=item Processing ev_async_send: O(number_of_async_watchers)
4511 5516
4512=item Processing signals: O(max_signal_number) 5517=item Processing signals: O(max_signal_number)
4513 5518
4514Sending involves a system call I<iff> there were no other C<ev_async_send> 5519Sending involves a system call I<iff> there were no other C<ev_async_send>
4515calls in the current loop iteration. Checking for async and signal events 5520calls in the current loop iteration and the loop is currently
5521blocked. Checking for async and signal events involves iterating over all
4516involves iterating over all running async watchers or all signal numbers. 5522running async watchers or all signal numbers.
4517 5523
4518=back 5524=back
4519 5525
4520 5526
5527=head1 PORTING FROM LIBEV 3.X TO 4.X
5528
5529The major version 4 introduced some incompatible changes to the API.
5530
5531At the moment, the C<ev.h> header file provides compatibility definitions
5532for all changes, so most programs should still compile. The compatibility
5533layer might be removed in later versions of libev, so better update to the
5534new API early than late.
5535
5536=over 4
5537
5538=item C<EV_COMPAT3> backwards compatibility mechanism
5539
5540The backward compatibility mechanism can be controlled by
5541C<EV_COMPAT3>. See L</"PREPROCESSOR SYMBOLS/MACROS"> in the L</EMBEDDING>
5542section.
5543
5544=item C<ev_default_destroy> and C<ev_default_fork> have been removed
5545
5546These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
5547
5548 ev_loop_destroy (EV_DEFAULT_UC);
5549 ev_loop_fork (EV_DEFAULT);
5550
5551=item function/symbol renames
5552
5553A number of functions and symbols have been renamed:
5554
5555 ev_loop => ev_run
5556 EVLOOP_NONBLOCK => EVRUN_NOWAIT
5557 EVLOOP_ONESHOT => EVRUN_ONCE
5558
5559 ev_unloop => ev_break
5560 EVUNLOOP_CANCEL => EVBREAK_CANCEL
5561 EVUNLOOP_ONE => EVBREAK_ONE
5562 EVUNLOOP_ALL => EVBREAK_ALL
5563
5564 EV_TIMEOUT => EV_TIMER
5565
5566 ev_loop_count => ev_iteration
5567 ev_loop_depth => ev_depth
5568 ev_loop_verify => ev_verify
5569
5570Most functions working on C<struct ev_loop> objects don't have an
5571C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
5572associated constants have been renamed to not collide with the C<struct
5573ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
5574as all other watcher types. Note that C<ev_loop_fork> is still called
5575C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
5576typedef.
5577
5578=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
5579
5580The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
5581mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
5582and work, but the library code will of course be larger.
5583
5584=back
5585
5586
4521=head1 GLOSSARY 5587=head1 GLOSSARY
4522 5588
4523=over 4 5589=over 4
4524 5590
4525=item active 5591=item active
4526 5592
4527A watcher is active as long as it has been started (has been attached to 5593A watcher is active as long as it has been started and not yet stopped.
4528an event loop) but not yet stopped (disassociated from the event loop). 5594See L</WATCHER STATES> for details.
4529 5595
4530=item application 5596=item application
4531 5597
4532In this document, an application is whatever is using libev. 5598In this document, an application is whatever is using libev.
5599
5600=item backend
5601
5602The part of the code dealing with the operating system interfaces.
4533 5603
4534=item callback 5604=item callback
4535 5605
4536The address of a function that is called when some event has been 5606The address of a function that is called when some event has been
4537detected. Callbacks are being passed the event loop, the watcher that 5607detected. Callbacks are being passed the event loop, the watcher that
4538received the event, and the actual event bitset. 5608received the event, and the actual event bitset.
4539 5609
4540=item callback invocation 5610=item callback/watcher invocation
4541 5611
4542The act of calling the callback associated with a watcher. 5612The act of calling the callback associated with a watcher.
4543 5613
4544=item event 5614=item event
4545 5615
4546A change of state of some external event, such as data now being available 5616A change of state of some external event, such as data now being available
4547for reading on a file descriptor, time having passed or simply not having 5617for reading on a file descriptor, time having passed or simply not having
4548any other events happening anymore. 5618any other events happening anymore.
4549 5619
4550In libev, events are represented as single bits (such as C<EV_READ> or 5620In libev, events are represented as single bits (such as C<EV_READ> or
4551C<EV_TIMEOUT>). 5621C<EV_TIMER>).
4552 5622
4553=item event library 5623=item event library
4554 5624
4555A software package implementing an event model and loop. 5625A software package implementing an event model and loop.
4556 5626
4564The model used to describe how an event loop handles and processes 5634The model used to describe how an event loop handles and processes
4565watchers and events. 5635watchers and events.
4566 5636
4567=item pending 5637=item pending
4568 5638
4569A watcher is pending as soon as the corresponding event has been detected, 5639A watcher is pending as soon as the corresponding event has been
4570and stops being pending as soon as the watcher will be invoked or its 5640detected. See L</WATCHER STATES> for details.
4571pending status is explicitly cleared by the application.
4572
4573A watcher can be pending, but not active. Stopping a watcher also clears
4574its pending status.
4575 5641
4576=item real time 5642=item real time
4577 5643
4578The physical time that is observed. It is apparently strictly monotonic :) 5644The physical time that is observed. It is apparently strictly monotonic :)
4579 5645
4580=item wall-clock time 5646=item wall-clock time
4581 5647
4582The time and date as shown on clocks. Unlike real time, it can actually 5648The time and date as shown on clocks. Unlike real time, it can actually
4583be wrong and jump forwards and backwards, e.g. when the you adjust your 5649be wrong and jump forwards and backwards, e.g. when you adjust your
4584clock. 5650clock.
4585 5651
4586=item watcher 5652=item watcher
4587 5653
4588A data structure that describes interest in certain events. Watchers need 5654A data structure that describes interest in certain events. Watchers need
4589to be started (attached to an event loop) before they can receive events. 5655to be started (attached to an event loop) before they can receive events.
4590 5656
4591=item watcher invocation
4592
4593The act of calling the callback associated with a watcher.
4594
4595=back 5657=back
4596 5658
4597=head1 AUTHOR 5659=head1 AUTHOR
4598 5660
4599Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson. 5661Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
5662Magnusson and Emanuele Giaquinta, and minor corrections by many others.
4600 5663

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines