1 <?xml version="1.0" encoding="UTF-8"?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
3 <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
6 <meta name="description" content="Pod documentation for libev" />
7 <meta name="inputfile" content="<standard input>" />
8 <meta name="outputfile" content="<standard output>" />
9 <meta name="created" content="Wed Dec 12 18:55:04 2007" />
10 <meta name="generator" content="Pod::Xhtml 1.57" />
11 <link rel="stylesheet" href="http://res.tst.eu/pod.css"/></head>
15 <h3 id="TOP">Index</h3>
17 <ul><li><a href="#NAME">NAME</a></li>
18 <li><a href="#SYNOPSIS">SYNOPSIS</a></li>
19 <li><a href="#EXAMPLE_PROGRAM">EXAMPLE PROGRAM</a></li>
20 <li><a href="#DESCRIPTION">DESCRIPTION</a></li>
21 <li><a href="#FEATURES">FEATURES</a></li>
22 <li><a href="#CONVENTIONS">CONVENTIONS</a></li>
23 <li><a href="#TIME_REPRESENTATION">TIME REPRESENTATION</a></li>
24 <li><a href="#GLOBAL_FUNCTIONS">GLOBAL FUNCTIONS</a></li>
25 <li><a href="#FUNCTIONS_CONTROLLING_THE_EVENT_LOOP">FUNCTIONS CONTROLLING THE EVENT LOOP</a></li>
26 <li><a href="#ANATOMY_OF_A_WATCHER">ANATOMY OF A WATCHER</a>
27 <ul><li><a href="#GENERIC_WATCHER_FUNCTIONS">GENERIC WATCHER FUNCTIONS</a></li>
28 <li><a href="#ASSOCIATING_CUSTOM_DATA_WITH_A_WATCH">ASSOCIATING CUSTOM DATA WITH A WATCHER</a></li>
31 <li><a href="#WATCHER_TYPES">WATCHER TYPES</a>
32 <ul><li><a href="#code_ev_io_code_is_this_file_descrip"><code>ev_io</code> - is this file descriptor readable or writable?</a>
33 <ul><li><a href="#The_special_problem_of_disappearing_">The special problem of disappearing file descriptors</a></li>
34 <li><a href="#Watcher_Specific_Functions">Watcher-Specific Functions</a></li>
37 <li><a href="#code_ev_timer_code_relative_and_opti"><code>ev_timer</code> - relative and optionally repeating timeouts</a>
38 <ul><li><a href="#Watcher_Specific_Functions_and_Data_">Watcher-Specific Functions and Data Members</a></li>
41 <li><a href="#code_ev_periodic_code_to_cron_or_not"><code>ev_periodic</code> - to cron or not to cron?</a>
42 <ul><li><a href="#Watcher_Specific_Functions_and_Data_-3">Watcher-Specific Functions and Data Members</a></li>
45 <li><a href="#code_ev_signal_code_signal_me_when_a"><code>ev_signal</code> - signal me when a signal gets signalled!</a>
46 <ul><li><a href="#Watcher_Specific_Functions_and_Data_-4">Watcher-Specific Functions and Data Members</a></li>
49 <li><a href="#code_ev_child_code_watch_out_for_pro"><code>ev_child</code> - watch out for process status changes</a>
50 <ul><li><a href="#Watcher_Specific_Functions_and_Data_-5">Watcher-Specific Functions and Data Members</a></li>
53 <li><a href="#code_ev_stat_code_did_the_file_attri"><code>ev_stat</code> - did the file attributes just change?</a>
54 <ul><li><a href="#Watcher_Specific_Functions_and_Data_-6">Watcher-Specific Functions and Data Members</a></li>
57 <li><a href="#code_ev_idle_code_when_you_ve_got_no"><code>ev_idle</code> - when you've got nothing better to do...</a>
58 <ul><li><a href="#Watcher_Specific_Functions_and_Data_-7">Watcher-Specific Functions and Data Members</a></li>
61 <li><a href="#code_ev_prepare_code_and_code_ev_che"><code>ev_prepare</code> and <code>ev_check</code> - customise your event loop!</a>
62 <ul><li><a href="#Watcher_Specific_Functions_and_Data_-8">Watcher-Specific Functions and Data Members</a></li>
65 <li><a href="#code_ev_embed_code_when_one_backend_"><code>ev_embed</code> - when one backend isn't enough...</a>
66 <ul><li><a href="#Watcher_Specific_Functions_and_Data_-9">Watcher-Specific Functions and Data Members</a></li>
69 <li><a href="#code_ev_fork_code_the_audacity_to_re"><code>ev_fork</code> - the audacity to resume the event loop after a fork</a></li>
72 <li><a href="#OTHER_FUNCTIONS">OTHER FUNCTIONS</a></li>
73 <li><a href="#LIBEVENT_EMULATION">LIBEVENT EMULATION</a></li>
74 <li><a href="#C_SUPPORT">C++ SUPPORT</a></li>
75 <li><a href="#MACRO_MAGIC">MACRO MAGIC</a></li>
76 <li><a href="#EMBEDDING">EMBEDDING</a>
77 <ul><li><a href="#FILESETS">FILESETS</a>
78 <ul><li><a href="#CORE_EVENT_LOOP">CORE EVENT LOOP</a></li>
79 <li><a href="#LIBEVENT_COMPATIBILITY_API">LIBEVENT COMPATIBILITY API</a></li>
80 <li><a href="#AUTOCONF_SUPPORT">AUTOCONF SUPPORT</a></li>
83 <li><a href="#PREPROCESSOR_SYMBOLS_MACROS">PREPROCESSOR SYMBOLS/MACROS</a></li>
84 <li><a href="#EXAMPLES">EXAMPLES</a></li>
87 <li><a href="#COMPLEXITIES">COMPLEXITIES</a></li>
88 <li><a href="#AUTHOR">AUTHOR</a>
93 <h1 id="NAME">NAME</h1>
94 <div id="NAME_CONTENT">
95 <p>libev - a high performance full-featured event loop written in C</p>
98 <h1 id="SYNOPSIS">SYNOPSIS</h1>
99 <div id="SYNOPSIS_CONTENT">
100 <pre> #include <ev.h>
105 <h1 id="EXAMPLE_PROGRAM">EXAMPLE PROGRAM</h1>
106 <div id="EXAMPLE_PROGRAM_CONTENT">
107 <pre> #include <ev.h>
110 ev_timer timeout_watcher;
112 /* called when data readable on stdin */
114 stdin_cb (EV_P_ struct ev_io *w, int revents)
116 /* puts ("stdin ready"); */
117 ev_io_stop (EV_A_ w); /* just a syntax example */
118 ev_unloop (EV_A_ EVUNLOOP_ALL); /* leave all loop calls */
122 timeout_cb (EV_P_ struct ev_timer *w, int revents)
124 /* puts ("timeout"); */
125 ev_unloop (EV_A_ EVUNLOOP_ONE); /* leave one loop call */
131 struct ev_loop *loop = ev_default_loop (0);
133 /* initialise an io watcher, then start it */
134 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
135 ev_io_start (loop, &stdin_watcher);
137 /* simple non-repeating 5.5 second timeout */
138 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
139 ev_timer_start (loop, &timeout_watcher);
141 /* loop till timeout or data ready */
150 <h1 id="DESCRIPTION">DESCRIPTION</h1>
151 <div id="DESCRIPTION_CONTENT">
152 <p>The newest version of this document is also available as a html-formatted
153 web page you might find easier to navigate when reading it for the first
154 time: <a href="http://cvs.schmorp.de/libev/ev.html">http://cvs.schmorp.de/libev/ev.html</a>.</p>
155 <p>Libev is an event loop: you register interest in certain events (such as a
156 file descriptor being readable or a timeout occuring), and it will manage
157 these event sources and provide your program with events.</p>
158 <p>To do this, it must take more or less complete control over your process
159 (or thread) by executing the <i>event loop</i> handler, and will then
160 communicate events via a callback mechanism.</p>
161 <p>You register interest in certain events by registering so-called <i>event
162 watchers</i>, which are relatively small C structures you initialise with the
163 details of the event, and then hand it over to libev by <i>starting</i> the
167 <h1 id="FEATURES">FEATURES</h1>
168 <div id="FEATURES_CONTENT">
169 <p>Libev supports <code>select</code>, <code>poll</code>, the Linux-specific <code>epoll</code>, the
170 BSD-specific <code>kqueue</code> and the Solaris-specific event port mechanisms
171 for file descriptor events (<code>ev_io</code>), the Linux <code>inotify</code> interface
172 (for <code>ev_stat</code>), relative timers (<code>ev_timer</code>), absolute timers
173 with customised rescheduling (<code>ev_periodic</code>), synchronous signals
174 (<code>ev_signal</code>), process status change events (<code>ev_child</code>), and event
175 watchers dealing with the event loop mechanism itself (<code>ev_idle</code>,
176 <code>ev_embed</code>, <code>ev_prepare</code> and <code>ev_check</code> watchers) as well as
177 file watchers (<code>ev_stat</code>) and even limited support for fork events
178 (<code>ev_fork</code>).</p>
179 <p>It also is quite fast (see this
180 <a href="http://libev.schmorp.de/bench.html">benchmark</a> comparing it to libevent
184 <h1 id="CONVENTIONS">CONVENTIONS</h1>
185 <div id="CONVENTIONS_CONTENT">
186 <p>Libev is very configurable. In this manual the default configuration will
187 be described, which supports multiple event loops. For more info about
188 various configuration options please have a look at <strong>EMBED</strong> section in
189 this manual. If libev was configured without support for multiple event
190 loops, then all functions taking an initial argument of name <code>loop</code>
191 (which is always of type <code>struct ev_loop *</code>) will not have this argument.</p>
194 <h1 id="TIME_REPRESENTATION">TIME REPRESENTATION</h1>
195 <div id="TIME_REPRESENTATION_CONTENT">
196 <p>Libev represents time as a single floating point number, representing the
197 (fractional) number of seconds since the (POSIX) epoch (somewhere near
198 the beginning of 1970, details are complicated, don't ask). This type is
199 called <code>ev_tstamp</code>, which is what you should use too. It usually aliases
200 to the <code>double</code> type in C, and when you need to do any calculations on
201 it, you should treat it as such.</p>
204 <h1 id="GLOBAL_FUNCTIONS">GLOBAL FUNCTIONS</h1>
205 <div id="GLOBAL_FUNCTIONS_CONTENT">
206 <p>These functions can be called anytime, even before initialising the
207 library in any way.</p>
209 <dt>ev_tstamp ev_time ()</dt>
211 <p>Returns the current time as libev would use it. Please note that the
212 <code>ev_now</code> function is usually faster and also often returns the timestamp
213 you actually want to know.</p>
215 <dt>int ev_version_major ()</dt>
216 <dt>int ev_version_minor ()</dt>
218 <p>You can find out the major and minor ABI version numbers of the library
219 you linked against by calling the functions <code>ev_version_major</code> and
220 <code>ev_version_minor</code>. If you want, you can compare against the global
221 symbols <code>EV_VERSION_MAJOR</code> and <code>EV_VERSION_MINOR</code>, which specify the
222 version of the library your program was compiled against.</p>
223 <p>These version numbers refer to the ABI version of the library, not the
225 <p>Usually, it's a good idea to terminate if the major versions mismatch,
226 as this indicates an incompatible change. Minor versions are usually
227 compatible to older versions, so a larger minor version alone is usually
229 <p>Example: Make sure we haven't accidentally been linked against the wrong
231 <pre> assert (("libev version mismatch",
232 ev_version_major () == EV_VERSION_MAJOR
233 && ev_version_minor () >= EV_VERSION_MINOR));
237 <dt>unsigned int ev_supported_backends ()</dt>
239 <p>Return the set of all backends (i.e. their corresponding <code>EV_BACKEND_*</code>
240 value) compiled into this binary of libev (independent of their
241 availability on the system you are running on). See <code>ev_default_loop</code> for
242 a description of the set values.</p>
243 <p>Example: make sure we have the epoll method, because yeah this is cool and
244 a must have and can we have a torrent of it please!!!11</p>
245 <pre> assert (("sorry, no epoll, no sex",
246 ev_supported_backends () & EVBACKEND_EPOLL));
250 <dt>unsigned int ev_recommended_backends ()</dt>
252 <p>Return the set of all backends compiled into this binary of libev and also
253 recommended for this platform. This set is often smaller than the one
254 returned by <code>ev_supported_backends</code>, as for example kqueue is broken on
255 most BSDs and will not be autodetected unless you explicitly request it
256 (assuming you know what you are doing). This is the set of backends that
257 libev will probe for if you specify no backends explicitly.</p>
259 <dt>unsigned int ev_embeddable_backends ()</dt>
261 <p>Returns the set of backends that are embeddable in other event loops. This
262 is the theoretical, all-platform, value. To find which backends
263 might be supported on the current system, you would need to look at
264 <code>ev_embeddable_backends () & ev_supported_backends ()</code>, likewise for
265 recommended ones.</p>
266 <p>See the description of <code>ev_embed</code> watchers for more info.</p>
268 <dt>ev_set_allocator (void *(*cb)(void *ptr, long size))</dt>
270 <p>Sets the allocation function to use (the prototype is similar - the
271 semantics is identical - to the realloc C function). It is used to
272 allocate and free memory (no surprises here). If it returns zero when
273 memory needs to be allocated, the library might abort or take some
274 potentially destructive action. The default is your system realloc
276 <p>You could override this function in high-availability programs to, say,
277 free some memory if it cannot allocate memory, to use a special allocator,
278 or even to sleep a while and retry until some memory is available.</p>
279 <p>Example: Replace the libev allocator with one that waits a bit and then
282 persistent_realloc (void *ptr, size_t size)
286 void *newptr = realloc (ptr, size);
296 ev_set_allocator (persistent_realloc);
300 <dt>ev_set_syserr_cb (void (*cb)(const char *msg));</dt>
302 <p>Set the callback function to call on a retryable syscall error (such
303 as failed select, poll, epoll_wait). The message is a printable string
304 indicating the system call or subsystem causing the problem. If this
305 callback is set, then libev will expect it to remedy the sitution, no
306 matter what, when it returns. That is, libev will generally retry the
307 requested operation, or, if the condition doesn't go away, do bad stuff
309 <p>Example: This is basically the same thing that libev does internally, too.</p>
311 fatal_error (const char *msg)
318 ev_set_syserr_cb (fatal_error);
325 <h1 id="FUNCTIONS_CONTROLLING_THE_EVENT_LOOP">FUNCTIONS CONTROLLING THE EVENT LOOP</h1>
326 <div id="FUNCTIONS_CONTROLLING_THE_EVENT_LOOP-2">
327 <p>An event loop is described by a <code>struct ev_loop *</code>. The library knows two
328 types of such loops, the <i>default</i> loop, which supports signals and child
329 events, and dynamically created loops which do not.</p>
330 <p>If you use threads, a common model is to run the default event loop
331 in your main thread (or in a separate thread) and for each thread you
332 create, you also create another event loop. Libev itself does no locking
333 whatsoever, so if you mix calls to the same event loop in different
334 threads, make sure you lock (this is usually a bad idea, though, even if
335 done correctly, because it's hideous and inefficient).</p>
337 <dt>struct ev_loop *ev_default_loop (unsigned int flags)</dt>
339 <p>This will initialise the default event loop if it hasn't been initialised
340 yet and return it. If the default loop could not be initialised, returns
341 false. If it already was initialised it simply returns it (and ignores the
342 flags. If that is troubling you, check <code>ev_backend ()</code> afterwards).</p>
343 <p>If you don't know what event loop to use, use the one returned from this
345 <p>The flags argument can be used to specify special behaviour or specific
346 backends to use, and is usually specified as <code>0</code> (or <code>EVFLAG_AUTO</code>).</p>
347 <p>The following flags are supported:</p>
350 <dt><code>EVFLAG_AUTO</code></dt>
352 <p>The default flags value. Use this if you have no clue (it's the right
353 thing, believe me).</p>
355 <dt><code>EVFLAG_NOENV</code></dt>
357 <p>If this flag bit is ored into the flag value (or the program runs setuid
358 or setgid) then libev will <i>not</i> look at the environment variable
359 <code>LIBEV_FLAGS</code>. Otherwise (the default), this environment variable will
360 override the flags completely if it is found in the environment. This is
361 useful to try out specific backends to test their performance, or to work
364 <dt><code>EVFLAG_FORKCHECK</code></dt>
366 <p>Instead of calling <code>ev_default_fork</code> or <code>ev_loop_fork</code> manually after
367 a fork, you can also make libev check for a fork in each iteration by
368 enabling this flag.</p>
369 <p>This works by calling <code>getpid ()</code> on every iteration of the loop,
370 and thus this might slow down your event loop if you do a lot of loop
371 iterations and little real work, but is usually not noticeable (on my
372 Linux system for example, <code>getpid</code> is actually a simple 5-insn sequence
373 without a syscall and thus <i>very</i> fast, but my Linux system also has
374 <code>pthread_atfork</code> which is even faster).</p>
375 <p>The big advantage of this flag is that you can forget about fork (and
376 forget about forgetting to tell libev about forking) when you use this
378 <p>This flag setting cannot be overriden or specified in the <code>LIBEV_FLAGS</code>
379 environment variable.</p>
381 <dt><code>EVBACKEND_SELECT</code> (value 1, portable select backend)</dt>
383 <p>This is your standard select(2) backend. Not <i>completely</i> standard, as
384 libev tries to roll its own fd_set with no limits on the number of fds,
385 but if that fails, expect a fairly low limit on the number of fds when
386 using this backend. It doesn't scale too well (O(highest_fd)), but its usually
387 the fastest backend for a low number of fds.</p>
389 <dt><code>EVBACKEND_POLL</code> (value 2, poll backend, available everywhere except on windows)</dt>
391 <p>And this is your standard poll(2) backend. It's more complicated than
392 select, but handles sparse fds better and has no artificial limit on the
393 number of fds you can use (except it will slow down considerably with a
394 lot of inactive fds). It scales similarly to select, i.e. O(total_fds).</p>
396 <dt><code>EVBACKEND_EPOLL</code> (value 4, Linux)</dt>
398 <p>For few fds, this backend is a bit little slower than poll and select,
399 but it scales phenomenally better. While poll and select usually scale like
400 O(total_fds) where n is the total number of fds (or the highest fd), epoll scales
401 either O(1) or O(active_fds).</p>
402 <p>While stopping and starting an I/O watcher in the same iteration will
403 result in some caching, there is still a syscall per such incident
404 (because the fd could point to a different file description now), so its
405 best to avoid that. Also, dup()ed file descriptors might not work very
406 well if you register events for both fds.</p>
407 <p>Please note that epoll sometimes generates spurious notifications, so you
408 need to use non-blocking I/O or other means to avoid blocking when no data
409 (or space) is available.</p>
411 <dt><code>EVBACKEND_KQUEUE</code> (value 8, most BSD clones)</dt>
413 <p>Kqueue deserves special mention, as at the time of this writing, it
414 was broken on all BSDs except NetBSD (usually it doesn't work with
415 anything but sockets and pipes, except on Darwin, where of course its
416 completely useless). For this reason its not being "autodetected"
417 unless you explicitly specify it explicitly in the flags (i.e. using
418 <code>EVBACKEND_KQUEUE</code>).</p>
419 <p>It scales in the same way as the epoll backend, but the interface to the
420 kernel is more efficient (which says nothing about its actual speed, of
421 course). While starting and stopping an I/O watcher does not cause an
422 extra syscall as with epoll, it still adds up to four event changes per
423 incident, so its best to avoid that.</p>
425 <dt><code>EVBACKEND_DEVPOLL</code> (value 16, Solaris 8)</dt>
427 <p>This is not implemented yet (and might never be).</p>
429 <dt><code>EVBACKEND_PORT</code> (value 32, Solaris 10)</dt>
431 <p>This uses the Solaris 10 port mechanism. As with everything on Solaris,
432 it's really slow, but it still scales very well (O(active_fds)).</p>
433 <p>Please note that solaris ports can result in a lot of spurious
434 notifications, so you need to use non-blocking I/O or other means to avoid
435 blocking when no data (or space) is available.</p>
437 <dt><code>EVBACKEND_ALL</code></dt>
439 <p>Try all backends (even potentially broken ones that wouldn't be tried
440 with <code>EVFLAG_AUTO</code>). Since this is a mask, you can do stuff such as
441 <code>EVBACKEND_ALL & ~EVBACKEND_KQUEUE</code>.</p>
445 <p>If one or more of these are ored into the flags value, then only these
446 backends will be tried (in the reverse order as given here). If none are
447 specified, most compiled-in backend will be tried, usually in reverse
448 order of their flag values :)</p>
449 <p>The most typical usage is like this:</p>
450 <pre> if (!ev_default_loop (0))
451 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
454 <p>Restrict libev to the select and poll backends, and do not allow
455 environment settings to be taken into account:</p>
456 <pre> ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
459 <p>Use whatever libev has to offer, but make sure that kqueue is used if
460 available (warning, breaks stuff, best use only with your own private
461 event loop and only if you know the OS supports your types of fds):</p>
462 <pre> ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
466 <dt>struct ev_loop *ev_loop_new (unsigned int flags)</dt>
468 <p>Similar to <code>ev_default_loop</code>, but always creates a new event loop that is
469 always distinct from the default loop. Unlike the default loop, it cannot
470 handle signal and child watchers, and attempts to do so will be greeted by
471 undefined behaviour (or a failed assertion if assertions are enabled).</p>
472 <p>Example: Try to create a event loop that uses epoll and nothing else.</p>
473 <pre> struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
475 fatal ("no epoll found here, maybe it hides under your chair");
479 <dt>ev_default_destroy ()</dt>
481 <p>Destroys the default loop again (frees all memory and kernel state
482 etc.). None of the active event watchers will be stopped in the normal
483 sense, so e.g. <code>ev_is_active</code> might still return true. It is your
484 responsibility to either stop all watchers cleanly yoursef <i>before</i>
485 calling this function, or cope with the fact afterwards (which is usually
486 the easiest thing, youc na just ignore the watchers and/or <code>free ()</code> them
489 <dt>ev_loop_destroy (loop)</dt>
491 <p>Like <code>ev_default_destroy</code>, but destroys an event loop created by an
492 earlier call to <code>ev_loop_new</code>.</p>
494 <dt>ev_default_fork ()</dt>
496 <p>This function reinitialises the kernel state for backends that have
497 one. Despite the name, you can call it anytime, but it makes most sense
498 after forking, in either the parent or child process (or both, but that
499 again makes little sense).</p>
500 <p>You <i>must</i> call this function in the child process after forking if and
501 only if you want to use the event library in both processes. If you just
502 fork+exec, you don't have to call it.</p>
503 <p>The function itself is quite fast and it's usually not a problem to call
504 it just in case after a fork. To make this easy, the function will fit in
505 quite nicely into a call to <code>pthread_atfork</code>:</p>
506 <pre> pthread_atfork (0, 0, ev_default_fork);
509 <p>At the moment, <code>EVBACKEND_SELECT</code> and <code>EVBACKEND_POLL</code> are safe to use
510 without calling this function, so if you force one of those backends you
511 do not need to care.</p>
513 <dt>ev_loop_fork (loop)</dt>
515 <p>Like <code>ev_default_fork</code>, but acts on an event loop created by
516 <code>ev_loop_new</code>. Yes, you have to call this on every allocated event loop
517 after fork, and how you do this is entirely your own problem.</p>
519 <dt>unsigned int ev_loop_count (loop)</dt>
521 <p>Returns the count of loop iterations for the loop, which is identical to
522 the number of times libev did poll for new events. It starts at <code>0</code> and
523 happily wraps around with enough iterations.</p>
524 <p>This value can sometimes be useful as a generation counter of sorts (it
525 "ticks" the number of loop iterations), as it roughly corresponds with
526 <code>ev_prepare</code> and <code>ev_check</code> calls.</p>
528 <dt>unsigned int ev_backend (loop)</dt>
530 <p>Returns one of the <code>EVBACKEND_*</code> flags indicating the event backend in
533 <dt>ev_tstamp ev_now (loop)</dt>
535 <p>Returns the current "event loop time", which is the time the event loop
536 received events and started processing them. This timestamp does not
537 change as long as callbacks are being processed, and this is also the base
538 time used for relative timers. You can treat it as the timestamp of the
539 event occuring (or more correctly, libev finding out about it).</p>
541 <dt>ev_loop (loop, int flags)</dt>
543 <p>Finally, this is it, the event handler. This function usually is called
544 after you initialised all your watchers and you want to start handling
546 <p>If the flags argument is specified as <code>0</code>, it will not return until
547 either no event watchers are active anymore or <code>ev_unloop</code> was called.</p>
548 <p>Please note that an explicit <code>ev_unloop</code> is usually better than
549 relying on all watchers to be stopped when deciding when a program has
550 finished (especially in interactive programs), but having a program that
551 automatically loops as long as it has to and no longer by virtue of
552 relying on its watchers stopping correctly is a thing of beauty.</p>
553 <p>A flags value of <code>EVLOOP_NONBLOCK</code> will look for new events, will handle
554 those events and any outstanding ones, but will not block your process in
555 case there are no events and will return after one iteration of the loop.</p>
556 <p>A flags value of <code>EVLOOP_ONESHOT</code> will look for new events (waiting if
557 neccessary) and will handle those and any outstanding ones. It will block
558 your process until at least one new event arrives, and will return after
559 one iteration of the loop. This is useful if you are waiting for some
560 external event in conjunction with something not expressible using other
561 libev watchers. However, a pair of <code>ev_prepare</code>/<code>ev_check</code> watchers is
562 usually a better approach for this kind of thing.</p>
563 <p>Here are the gory details of what <code>ev_loop</code> does:</p>
564 <pre> - Before the first iteration, call any pending watchers.
565 * If there are no active watchers (reference count is zero), return.
566 - Queue all prepare watchers and then call all outstanding watchers.
567 - If we have been forked, recreate the kernel state.
568 - Update the kernel state with all outstanding changes.
569 - Update the "event loop time".
570 - Calculate for how long to block.
571 - Block the process, waiting for any events.
572 - Queue all outstanding I/O (fd) events.
573 - Update the "event loop time" and do time jump handling.
574 - Queue all outstanding timers.
575 - Queue all outstanding periodics.
576 - If no events are pending now, queue all idle watchers.
577 - Queue all check watchers.
578 - Call all queued watchers in reverse order (i.e. check watchers first).
579 Signals and child watchers are implemented as I/O watchers, and will
580 be handled here by queueing them when their watcher gets executed.
581 - If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
582 were used, return, otherwise continue with step *.
585 <p>Example: Queue some jobs and then loop until no events are outsanding
587 <pre> ... queue jobs here, make sure they register event watchers as long
588 ... as they still have work to do (even an idle watcher will do..)
589 ev_loop (my_loop, 0);
594 <dt>ev_unloop (loop, how)</dt>
596 <p>Can be used to make a call to <code>ev_loop</code> return early (but only after it
597 has processed all outstanding events). The <code>how</code> argument must be either
598 <code>EVUNLOOP_ONE</code>, which will make the innermost <code>ev_loop</code> call return, or
599 <code>EVUNLOOP_ALL</code>, which will make all nested <code>ev_loop</code> calls return.</p>
601 <dt>ev_ref (loop)</dt>
602 <dt>ev_unref (loop)</dt>
604 <p>Ref/unref can be used to add or remove a reference count on the event
605 loop: Every watcher keeps one reference, and as long as the reference
606 count is nonzero, <code>ev_loop</code> will not return on its own. If you have
607 a watcher you never unregister that should not keep <code>ev_loop</code> from
608 returning, ev_unref() after starting, and ev_ref() before stopping it. For
609 example, libev itself uses this for its internal signal pipe: It is not
610 visible to the libev user and should not keep <code>ev_loop</code> from exiting if
611 no event watchers registered by it are active. It is also an excellent
612 way to do this for generic recurring timers or from within third-party
613 libraries. Just remember to <i>unref after start</i> and <i>ref before stop</i>.</p>
614 <p>Example: Create a signal watcher, but keep it from keeping <code>ev_loop</code>
615 running when nothing else is active.</p>
616 <pre> struct ev_signal exitsig;
617 ev_signal_init (&exitsig, sig_cb, SIGINT);
618 ev_signal_start (loop, &exitsig);
622 <p>Example: For some weird reason, unregister the above signal handler again.</p>
624 ev_signal_stop (loop, &exitsig);
635 <h1 id="ANATOMY_OF_A_WATCHER">ANATOMY OF A WATCHER</h1>
636 <div id="ANATOMY_OF_A_WATCHER_CONTENT">
637 <p>A watcher is a structure that you create and register to record your
638 interest in some event. For instance, if you want to wait for STDIN to
639 become readable, you would create an <code>ev_io</code> watcher for that:</p>
640 <pre> static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)
643 ev_unloop (loop, EVUNLOOP_ALL);
646 struct ev_loop *loop = ev_default_loop (0);
647 struct ev_io stdin_watcher;
648 ev_init (&stdin_watcher, my_cb);
649 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
650 ev_io_start (loop, &stdin_watcher);
654 <p>As you can see, you are responsible for allocating the memory for your
655 watcher structures (and it is usually a bad idea to do this on the stack,
656 although this can sometimes be quite valid).</p>
657 <p>Each watcher structure must be initialised by a call to <code>ev_init
658 (watcher *, callback)</code>, which expects a callback to be provided. This
659 callback gets invoked each time the event occurs (or, in the case of io
660 watchers, each time the event loop detects that the file descriptor given
661 is readable and/or writable).</p>
662 <p>Each watcher type has its own <code>ev_<type>_set (watcher *, ...)</code> macro
663 with arguments specific to this watcher type. There is also a macro
664 to combine initialisation and setting in one call: <code>ev_<type>_init
665 (watcher *, callback, ...)</code>.</p>
666 <p>To make the watcher actually watch out for events, you have to start it
667 with a watcher-specific start function (<code>ev_<type>_start (loop, watcher
668 *)</code>), and you can stop watching for events at any time by calling the
669 corresponding stop function (<code>ev_<type>_stop (loop, watcher *)</code>.</p>
670 <p>As long as your watcher is active (has been started but not stopped) you
671 must not touch the values stored in it. Most specifically you must never
672 reinitialise it or call its <code>set</code> macro.</p>
673 <p>Each and every callback receives the event loop pointer as first, the
674 registered watcher structure as second, and a bitset of received events as
676 <p>The received events usually include a single bit per event type received
677 (you can receive multiple events at the same time). The possible bit masks
680 <dt><code>EV_READ</code></dt>
681 <dt><code>EV_WRITE</code></dt>
683 <p>The file descriptor in the <code>ev_io</code> watcher has become readable and/or
686 <dt><code>EV_TIMEOUT</code></dt>
688 <p>The <code>ev_timer</code> watcher has timed out.</p>
690 <dt><code>EV_PERIODIC</code></dt>
692 <p>The <code>ev_periodic</code> watcher has timed out.</p>
694 <dt><code>EV_SIGNAL</code></dt>
696 <p>The signal specified in the <code>ev_signal</code> watcher has been received by a thread.</p>
698 <dt><code>EV_CHILD</code></dt>
700 <p>The pid specified in the <code>ev_child</code> watcher has received a status change.</p>
702 <dt><code>EV_STAT</code></dt>
704 <p>The path specified in the <code>ev_stat</code> watcher changed its attributes somehow.</p>
706 <dt><code>EV_IDLE</code></dt>
708 <p>The <code>ev_idle</code> watcher has determined that you have nothing better to do.</p>
710 <dt><code>EV_PREPARE</code></dt>
711 <dt><code>EV_CHECK</code></dt>
713 <p>All <code>ev_prepare</code> watchers are invoked just <i>before</i> <code>ev_loop</code> starts
714 to gather new events, and all <code>ev_check</code> watchers are invoked just after
715 <code>ev_loop</code> has gathered them, but before it invokes any callbacks for any
716 received events. Callbacks of both watcher types can start and stop as
717 many watchers as they want, and all of them will be taken into account
718 (for example, a <code>ev_prepare</code> watcher might start an idle watcher to keep
719 <code>ev_loop</code> from blocking).</p>
721 <dt><code>EV_EMBED</code></dt>
723 <p>The embedded event loop specified in the <code>ev_embed</code> watcher needs attention.</p>
725 <dt><code>EV_FORK</code></dt>
727 <p>The event loop has been resumed in the child process after fork (see
728 <code>ev_fork</code>).</p>
730 <dt><code>EV_ERROR</code></dt>
732 <p>An unspecified error has occured, the watcher has been stopped. This might
733 happen because the watcher could not be properly started because libev
734 ran out of memory, a file descriptor was found to be closed or any other
735 problem. You best act on it by reporting the problem and somehow coping
736 with the watcher being stopped.</p>
737 <p>Libev will usually signal a few "dummy" events together with an error,
738 for example it might indicate that a fd is readable or writable, and if
739 your callbacks is well-written it can just attempt the operation and cope
740 with the error from read() or write(). This will not work in multithreaded
741 programs, though, so beware.</p>
746 <h2 id="GENERIC_WATCHER_FUNCTIONS">GENERIC WATCHER FUNCTIONS</h2>
747 <div id="GENERIC_WATCHER_FUNCTIONS_CONTENT">
748 <p>In the following description, <code>TYPE</code> stands for the watcher type,
749 e.g. <code>timer</code> for <code>ev_timer</code> watchers and <code>io</code> for <code>ev_io</code> watchers.</p>
751 <dt><code>ev_init</code> (ev_TYPE *watcher, callback)</dt>
753 <p>This macro initialises the generic portion of a watcher. The contents
754 of the watcher object can be arbitrary (so <code>malloc</code> will do). Only
755 the generic parts of the watcher are initialised, you <i>need</i> to call
756 the type-specific <code>ev_TYPE_set</code> macro afterwards to initialise the
757 type-specific parts. For each type there is also a <code>ev_TYPE_init</code> macro
758 which rolls both calls into one.</p>
759 <p>You can reinitialise a watcher at any time as long as it has been stopped
760 (or never started) and there are no pending events outstanding.</p>
761 <p>The callback is always of type <code>void (*)(ev_loop *loop, ev_TYPE *watcher,
762 int revents)</code>.</p>
764 <dt><code>ev_TYPE_set</code> (ev_TYPE *, [args])</dt>
766 <p>This macro initialises the type-specific parts of a watcher. You need to
767 call <code>ev_init</code> at least once before you call this macro, but you can
768 call <code>ev_TYPE_set</code> any number of times. You must not, however, call this
769 macro on a watcher that is active (it can be pending, however, which is a
770 difference to the <code>ev_init</code> macro).</p>
771 <p>Although some watcher types do not have type-specific arguments
772 (e.g. <code>ev_prepare</code>) you still need to call its <code>set</code> macro.</p>
774 <dt><code>ev_TYPE_init</code> (ev_TYPE *watcher, callback, [args])</dt>
776 <p>This convinience macro rolls both <code>ev_init</code> and <code>ev_TYPE_set</code> macro
777 calls into a single call. This is the most convinient method to initialise
778 a watcher. The same limitations apply, of course.</p>
780 <dt><code>ev_TYPE_start</code> (loop *, ev_TYPE *watcher)</dt>
782 <p>Starts (activates) the given watcher. Only active watchers will receive
783 events. If the watcher is already active nothing will happen.</p>
785 <dt><code>ev_TYPE_stop</code> (loop *, ev_TYPE *watcher)</dt>
787 <p>Stops the given watcher again (if active) and clears the pending
788 status. It is possible that stopped watchers are pending (for example,
789 non-repeating timers are being stopped when they become pending), but
790 <code>ev_TYPE_stop</code> ensures that the watcher is neither active nor pending. If
791 you want to free or reuse the memory used by the watcher it is therefore a
792 good idea to always call its <code>ev_TYPE_stop</code> function.</p>
794 <dt>bool ev_is_active (ev_TYPE *watcher)</dt>
796 <p>Returns a true value iff the watcher is active (i.e. it has been started
797 and not yet been stopped). As long as a watcher is active you must not modify
800 <dt>bool ev_is_pending (ev_TYPE *watcher)</dt>
802 <p>Returns a true value iff the watcher is pending, (i.e. it has outstanding
803 events but its callback has not yet been invoked). As long as a watcher
804 is pending (but not active) you must not call an init function on it (but
805 <code>ev_TYPE_set</code> is safe), you must not change its priority, and you must
806 make sure the watcher is available to libev (e.g. you cannot <code>free ()</code>
809 <dt>callback ev_cb (ev_TYPE *watcher)</dt>
811 <p>Returns the callback currently set on the watcher.</p>
813 <dt>ev_cb_set (ev_TYPE *watcher, callback)</dt>
815 <p>Change the callback. You can change the callback at virtually any time
816 (modulo threads).</p>
818 <dt>ev_set_priority (ev_TYPE *watcher, priority)</dt>
819 <dt>int ev_priority (ev_TYPE *watcher)</dt>
821 <p>Set and query the priority of the watcher. The priority is a small
822 integer between <code>EV_MAXPRI</code> (default: <code>2</code>) and <code>EV_MINPRI</code>
823 (default: <code>-2</code>). Pending watchers with higher priority will be invoked
824 before watchers with lower priority, but priority will not keep watchers
825 from being executed (except for <code>ev_idle</code> watchers).</p>
826 <p>This means that priorities are <i>only</i> used for ordering callback
827 invocation after new events have been received. This is useful, for
828 example, to reduce latency after idling, or more often, to bind two
829 watchers on the same event and make sure one is called first.</p>
830 <p>If you need to suppress invocation when higher priority events are pending
831 you need to look at <code>ev_idle</code> watchers, which provide this functionality.</p>
832 <p>You <i>must not</i> change the priority of a watcher as long as it is active or
834 <p>The default priority used by watchers when no priority has been set is
835 always <code>0</code>, which is supposed to not be too high and not be too low :).</p>
836 <p>Setting a priority outside the range of <code>EV_MINPRI</code> to <code>EV_MAXPRI</code> is
837 fine, as long as you do not mind that the priority value you query might
838 or might not have been adjusted to be within valid range.</p>
840 <dt>ev_invoke (loop, ev_TYPE *watcher, int revents)</dt>
842 <p>Invoke the <code>watcher</code> with the given <code>loop</code> and <code>revents</code>. Neither
843 <code>loop</code> nor <code>revents</code> need to be valid as long as the watcher callback
844 can deal with that fact.</p>
846 <dt>int ev_clear_pending (loop, ev_TYPE *watcher)</dt>
848 <p>If the watcher is pending, this function returns clears its pending status
849 and returns its <code>revents</code> bitset (as if its callback was invoked). If the
850 watcher isn't pending it does nothing and returns <code>0</code>.</p>
859 <h2 id="ASSOCIATING_CUSTOM_DATA_WITH_A_WATCH">ASSOCIATING CUSTOM DATA WITH A WATCHER</h2>
860 <div id="ASSOCIATING_CUSTOM_DATA_WITH_A_WATCH-2">
861 <p>Each watcher has, by default, a member <code>void *data</code> that you can change
862 and read at any time, libev will completely ignore it. This can be used
863 to associate arbitrary data with your watcher. If you need more data and
864 don't want to allocate memory and store a pointer to it in that data
865 member, you can also "subclass" the watcher type and provide your own
872 struct whatever *mostinteresting;
876 <p>And since your callback will be called with a pointer to the watcher, you
877 can cast it back to your own type:</p>
878 <pre> static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
880 struct my_io *w = (struct my_io *)w_;
885 <p>More interesting and less C-conformant ways of casting your callback type
886 instead have been omitted.</p>
887 <p>Another common scenario is having some data structure with multiple
889 <pre> struct my_biggy
897 <p>In this case getting the pointer to <code>my_biggy</code> is a bit more complicated,
898 you need to use <code>offsetof</code>:</p>
899 <pre> #include <stddef.h>
902 t1_cb (EV_P_ struct ev_timer *w, int revents)
904 struct my_biggy big = (struct my_biggy *
905 (((char *)w) - offsetof (struct my_biggy, t1));
909 t2_cb (EV_P_ struct ev_timer *w, int revents)
911 struct my_biggy big = (struct my_biggy *
912 (((char *)w) - offsetof (struct my_biggy, t2));
921 <h1 id="WATCHER_TYPES">WATCHER TYPES</h1>
922 <div id="WATCHER_TYPES_CONTENT">
923 <p>This section describes each watcher in detail, but will not repeat
924 information given in the last section. Any initialisation/set macros,
925 functions and members specific to the watcher type are explained.</p>
926 <p>Members are additionally marked with either <i>[read-only]</i>, meaning that,
927 while the watcher is active, you can look at the member and expect some
928 sensible content, but you must not modify it (you can modify it while the
929 watcher is stopped to your hearts content), or <i>[read-write]</i>, which
930 means you can expect it to have some sensible content while the watcher
931 is active, but you can also modify it. Modifying it may not do something
932 sensible or take immediate effect (or do anything at all), but libev will
933 not crash or malfunction in any way.</p>
940 <h2 id="code_ev_io_code_is_this_file_descrip"><code>ev_io</code> - is this file descriptor readable or writable?</h2>
941 <div id="code_ev_io_code_is_this_file_descrip-2">
942 <p>I/O watchers check whether a file descriptor is readable or writable
943 in each iteration of the event loop, or, more precisely, when reading
944 would not block the process and writing would at least be able to write
945 some data. This behaviour is called level-triggering because you keep
946 receiving events as long as the condition persists. Remember you can stop
947 the watcher if you don't want to act on the event and neither want to
948 receive future events.</p>
949 <p>In general you can register as many read and/or write event watchers per
950 fd as you want (as long as you don't confuse yourself). Setting all file
951 descriptors to non-blocking mode is also usually a good idea (but not
952 required if you know what you are doing).</p>
953 <p>You have to be careful with dup'ed file descriptors, though. Some backends
954 (the linux epoll backend is a notable example) cannot handle dup'ed file
955 descriptors correctly if you register interest in two or more fds pointing
956 to the same underlying file/socket/etc. description (that is, they share
957 the same underlying "file open").</p>
958 <p>If you must do this, then force the use of a known-to-be-good backend
959 (at the time of this writing, this includes only <code>EVBACKEND_SELECT</code> and
960 <code>EVBACKEND_POLL</code>).</p>
961 <p>Another thing you have to watch out for is that it is quite easy to
962 receive "spurious" readyness notifications, that is your callback might
963 be called with <code>EV_READ</code> but a subsequent <code>read</code>(2) will actually block
964 because there is no data. Not only are some backends known to create a
965 lot of those (for example solaris ports), it is very easy to get into
966 this situation even with a relatively standard program structure. Thus
967 it is best to always use non-blocking I/O: An extra <code>read</code>(2) returning
968 <code>EAGAIN</code> is far preferable to a program hanging until some data arrives.</p>
969 <p>If you cannot run the fd in non-blocking mode (for example you should not
970 play around with an Xlib connection), then you have to seperately re-test
971 whether a file descriptor is really ready with a known-to-be good interface
972 such as poll (fortunately in our Xlib example, Xlib already does this on
973 its own, so its quite safe to use).</p>
976 <h3 id="The_special_problem_of_disappearing_">The special problem of disappearing file descriptors</h3>
977 <div id="The_special_problem_of_disappearing_-2">
978 <p>Some backends (e.g kqueue, epoll) need to be told about closing a file
979 descriptor (either by calling <code>close</code> explicitly or by any other means,
980 such as <code>dup</code>). The reason is that you register interest in some file
981 descriptor, but when it goes away, the operating system will silently drop
982 this interest. If another file descriptor with the same number then is
983 registered with libev, there is no efficient way to see that this is, in
984 fact, a different file descriptor.</p>
985 <p>To avoid having to explicitly tell libev about such cases, libev follows
986 the following policy: Each time <code>ev_io_set</code> is being called, libev
987 will assume that this is potentially a new file descriptor, otherwise
988 it is assumed that the file descriptor stays the same. That means that
989 you <i>have</i> to call <code>ev_io_set</code> (or <code>ev_io_init</code>) when you change the
990 descriptor even if the file descriptor number itself did not change.</p>
991 <p>This is how one would do it normally anyway, the important point is that
992 the libev application should not optimise around libev but should leave
993 optimisations to libev.</p>
1000 <h3 id="Watcher_Specific_Functions">Watcher-Specific Functions</h3>
1001 <div id="Watcher_Specific_Functions_CONTENT">
1003 <dt>ev_io_init (ev_io *, callback, int fd, int events)</dt>
1004 <dt>ev_io_set (ev_io *, int fd, int events)</dt>
1006 <p>Configures an <code>ev_io</code> watcher. The <code>fd</code> is the file descriptor to
1007 rceeive events for and events is either <code>EV_READ</code>, <code>EV_WRITE</code> or
1008 <code>EV_READ | EV_WRITE</code> to receive the given events.</p>
1010 <dt>int fd [read-only]</dt>
1012 <p>The file descriptor being watched.</p>
1014 <dt>int events [read-only]</dt>
1016 <p>The events being watched.</p>
1019 <p>Example: Call <code>stdin_readable_cb</code> when STDIN_FILENO has become, well
1020 readable, but only once. Since it is likely line-buffered, you could
1021 attempt to read a whole line in the callback.</p>
1023 stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1025 ev_io_stop (loop, w);
1026 .. read from stdin here (or from w->fd) and haqndle any I/O errors
1030 struct ev_loop *loop = ev_default_init (0);
1031 struct ev_io stdin_readable;
1032 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1033 ev_io_start (loop, &stdin_readable);
1042 <h2 id="code_ev_timer_code_relative_and_opti"><code>ev_timer</code> - relative and optionally repeating timeouts</h2>
1043 <div id="code_ev_timer_code_relative_and_opti-2">
1044 <p>Timer watchers are simple relative timers that generate an event after a
1045 given time, and optionally repeating in regular intervals after that.</p>
1046 <p>The timers are based on real time, that is, if you register an event that
1047 times out after an hour and you reset your system clock to last years
1048 time, it will still time out after (roughly) and hour. "Roughly" because
1049 detecting time jumps is hard, and some inaccuracies are unavoidable (the
1050 monotonic clock option helps a lot here).</p>
1051 <p>The relative timeouts are calculated relative to the <code>ev_now ()</code>
1052 time. This is usually the right thing as this timestamp refers to the time
1053 of the event triggering whatever timeout you are modifying/starting. If
1054 you suspect event processing to be delayed and you <i>need</i> to base the timeout
1055 on the current time, use something like this to adjust for this:</p>
1056 <pre> ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1059 <p>The callback is guarenteed to be invoked only when its timeout has passed,
1060 but if multiple timers become ready during the same loop iteration then
1061 order of execution is undefined.</p>
1064 <h3 id="Watcher_Specific_Functions_and_Data_">Watcher-Specific Functions and Data Members</h3>
1065 <div id="Watcher_Specific_Functions_and_Data_-2">
1067 <dt>ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)</dt>
1068 <dt>ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)</dt>
1070 <p>Configure the timer to trigger after <code>after</code> seconds. If <code>repeat</code> is
1071 <code>0.</code>, then it will automatically be stopped. If it is positive, then the
1072 timer will automatically be configured to trigger again <code>repeat</code> seconds
1073 later, again, and again, until stopped manually.</p>
1074 <p>The timer itself will do a best-effort at avoiding drift, that is, if you
1075 configure a timer to trigger every 10 seconds, then it will trigger at
1076 exactly 10 second intervals. If, however, your program cannot keep up with
1077 the timer (because it takes longer than those 10 seconds to do stuff) the
1078 timer will not fire more than once per event loop iteration.</p>
1080 <dt>ev_timer_again (loop)</dt>
1082 <p>This will act as if the timer timed out and restart it again if it is
1083 repeating. The exact semantics are:</p>
1084 <p>If the timer is pending, its pending status is cleared.</p>
1085 <p>If the timer is started but nonrepeating, stop it (as if it timed out).</p>
1086 <p>If the timer is repeating, either start it if necessary (with the
1087 <code>repeat</code> value), or reset the running timer to the <code>repeat</code> value.</p>
1088 <p>This sounds a bit complicated, but here is a useful and typical
1089 example: Imagine you have a tcp connection and you want a so-called idle
1090 timeout, that is, you want to be called when there have been, say, 60
1091 seconds of inactivity on the socket. The easiest way to do this is to
1092 configure an <code>ev_timer</code> with a <code>repeat</code> value of <code>60</code> and then call
1093 <code>ev_timer_again</code> each time you successfully read or write some data. If
1094 you go into an idle state where you do not expect data to travel on the
1095 socket, you can <code>ev_timer_stop</code> the timer, and <code>ev_timer_again</code> will
1096 automatically restart it if need be.</p>
1097 <p>That means you can ignore the <code>after</code> value and <code>ev_timer_start</code>
1098 altogether and only ever use the <code>repeat</code> value and <code>ev_timer_again</code>:</p>
1099 <pre> ev_timer_init (timer, callback, 0., 5.);
1100 ev_timer_again (loop, timer);
1102 timer->again = 17.;
1103 ev_timer_again (loop, timer);
1105 timer->again = 10.;
1106 ev_timer_again (loop, timer);
1109 <p>This is more slightly efficient then stopping/starting the timer each time
1110 you want to modify its timeout value.</p>
1112 <dt>ev_tstamp repeat [read-write]</dt>
1114 <p>The current <code>repeat</code> value. Will be used each time the watcher times out
1115 or <code>ev_timer_again</code> is called and determines the next timeout (if any),
1116 which is also when any modifications are taken into account.</p>
1119 <p>Example: Create a timer that fires after 60 seconds.</p>
1121 one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1123 .. one minute over, w is actually stopped right here
1126 struct ev_timer mytimer;
1127 ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1128 ev_timer_start (loop, &mytimer);
1131 <p>Example: Create a timeout timer that times out after 10 seconds of
1134 timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1136 .. ten seconds without any activity
1139 struct ev_timer mytimer;
1140 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1141 ev_timer_again (&mytimer); /* start timer */
1144 // and in some piece of code that gets executed on any "activity":
1145 // reset the timeout to start ticking again at 10 seconds
1146 ev_timer_again (&mytimer);
1154 <h2 id="code_ev_periodic_code_to_cron_or_not"><code>ev_periodic</code> - to cron or not to cron?</h2>
1155 <div id="code_ev_periodic_code_to_cron_or_not-2">
1156 <p>Periodic watchers are also timers of a kind, but they are very versatile
1157 (and unfortunately a bit complex).</p>
1158 <p>Unlike <code>ev_timer</code>'s, they are not based on real time (or relative time)
1159 but on wallclock time (absolute time). You can tell a periodic watcher
1160 to trigger "at" some specific point in time. For example, if you tell a
1161 periodic watcher to trigger in 10 seconds (by specifiying e.g. <code>ev_now ()
1162 + 10.</code>) and then reset your system clock to the last year, then it will
1163 take a year to trigger the event (unlike an <code>ev_timer</code>, which would trigger
1164 roughly 10 seconds later).</p>
1165 <p>They can also be used to implement vastly more complex timers, such as
1166 triggering an event on each midnight, local time or other, complicated,
1168 <p>As with timers, the callback is guarenteed to be invoked only when the
1169 time (<code>at</code>) has been passed, but if multiple periodic timers become ready
1170 during the same loop iteration then order of execution is undefined.</p>
1173 <h3 id="Watcher_Specific_Functions_and_Data_-3">Watcher-Specific Functions and Data Members</h3>
1174 <div id="Watcher_Specific_Functions_and_Data_-2">
1176 <dt>ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)</dt>
1177 <dt>ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)</dt>
1179 <p>Lots of arguments, lets sort it out... There are basically three modes of
1180 operation, and we will explain them from simplest to complex:</p>
1183 <dt>* absolute timer (at = time, interval = reschedule_cb = 0)</dt>
1185 <p>In this configuration the watcher triggers an event at the wallclock time
1186 <code>at</code> and doesn't repeat. It will not adjust when a time jump occurs,
1187 that is, if it is to be run at January 1st 2011 then it will run when the
1188 system time reaches or surpasses this time.</p>
1190 <dt>* non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)</dt>
1192 <p>In this mode the watcher will always be scheduled to time out at the next
1193 <code>at + N * interval</code> time (for some integer N, which can also be negative)
1194 and then repeat, regardless of any time jumps.</p>
1195 <p>This can be used to create timers that do not drift with respect to system
1197 <pre> ev_periodic_set (&periodic, 0., 3600., 0);
1200 <p>This doesn't mean there will always be 3600 seconds in between triggers,
1201 but only that the the callback will be called when the system time shows a
1202 full hour (UTC), or more correctly, when the system time is evenly divisible
1204 <p>Another way to think about it (for the mathematically inclined) is that
1205 <code>ev_periodic</code> will try to run the callback in this mode at the next possible
1206 time where <code>time = at (mod interval)</code>, regardless of any time jumps.</p>
1207 <p>For numerical stability it is preferable that the <code>at</code> value is near
1208 <code>ev_now ()</code> (the current time), but there is no range requirement for
1211 <dt>* manual reschedule mode (at and interval ignored, reschedule_cb = callback)</dt>
1213 <p>In this mode the values for <code>interval</code> and <code>at</code> are both being
1214 ignored. Instead, each time the periodic watcher gets scheduled, the
1215 reschedule callback will be called with the watcher as first, and the
1216 current time as second argument.</p>
1217 <p>NOTE: <i>This callback MUST NOT stop or destroy any periodic watcher,
1218 ever, or make any event loop modifications</i>. If you need to stop it,
1219 return <code>now + 1e30</code> (or so, fudge fudge) and stop it afterwards (e.g. by
1220 starting an <code>ev_prepare</code> watcher, which is legal).</p>
1221 <p>Its prototype is <code>ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
1222 ev_tstamp now)</code>, e.g.:</p>
1223 <pre> static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
1229 <p>It must return the next time to trigger, based on the passed time value
1230 (that is, the lowest time value larger than to the second argument). It
1231 will usually be called just before the callback will be triggered, but
1232 might be called at other times, too.</p>
1233 <p>NOTE: <i>This callback must always return a time that is later than the
1234 passed <code>now</code> value</i>. Not even <code>now</code> itself will do, it <i>must</i> be larger.</p>
1235 <p>This can be used to create very complex timers, such as a timer that
1236 triggers on each midnight, local time. To do this, you would calculate the
1237 next midnight after <code>now</code> and return the timestamp value for this. How
1238 you do this is, again, up to you (but it is not trivial, which is the main
1239 reason I omitted it as an example).</p>
1244 <dt>ev_periodic_again (loop, ev_periodic *)</dt>
1246 <p>Simply stops and restarts the periodic watcher again. This is only useful
1247 when you changed some parameters or the reschedule callback would return
1248 a different time than the last time it was called (e.g. in a crond like
1249 program when the crontabs have changed).</p>
1251 <dt>ev_tstamp offset [read-write]</dt>
1253 <p>When repeating, this contains the offset value, otherwise this is the
1254 absolute point in time (the <code>at</code> value passed to <code>ev_periodic_set</code>).</p>
1255 <p>Can be modified any time, but changes only take effect when the periodic
1256 timer fires or <code>ev_periodic_again</code> is being called.</p>
1258 <dt>ev_tstamp interval [read-write]</dt>
1260 <p>The current interval value. Can be modified any time, but changes only
1261 take effect when the periodic timer fires or <code>ev_periodic_again</code> is being
1264 <dt>ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]</dt>
1266 <p>The current reschedule callback, or <code>0</code>, if this functionality is
1267 switched off. Can be changed any time, but changes only take effect when
1268 the periodic timer fires or <code>ev_periodic_again</code> is being called.</p>
1271 <p>Example: Call a callback every hour, or, more precisely, whenever the
1272 system clock is divisible by 3600. The callback invocation times have
1273 potentially a lot of jittering, but good long-term stability.</p>
1275 clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1277 ... its now a full hour (UTC, or TAI or whatever your clock follows)
1280 struct ev_periodic hourly_tick;
1281 ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1282 ev_periodic_start (loop, &hourly_tick);
1285 <p>Example: The same as above, but use a reschedule callback to do it:</p>
1286 <pre> #include <math.h>
1289 my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)
1291 return fmod (now, 3600.) + 3600.;
1294 ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1297 <p>Example: Call a callback every hour, starting now:</p>
1298 <pre> struct ev_periodic hourly_tick;
1299 ev_periodic_init (&hourly_tick, clock_cb,
1300 fmod (ev_now (loop), 3600.), 3600., 0);
1301 ev_periodic_start (loop, &hourly_tick);
1309 <h2 id="code_ev_signal_code_signal_me_when_a"><code>ev_signal</code> - signal me when a signal gets signalled!</h2>
1310 <div id="code_ev_signal_code_signal_me_when_a-2">
1311 <p>Signal watchers will trigger an event when the process receives a specific
1312 signal one or more times. Even though signals are very asynchronous, libev
1313 will try it's best to deliver signals synchronously, i.e. as part of the
1314 normal event processing, like any other event.</p>
1315 <p>You can configure as many watchers as you like per signal. Only when the
1316 first watcher gets started will libev actually register a signal watcher
1317 with the kernel (thus it coexists with your own signal handlers as long
1318 as you don't register any with libev). Similarly, when the last signal
1319 watcher for a signal is stopped libev will reset the signal handler to
1320 SIG_DFL (regardless of what it was set to before).</p>
1323 <h3 id="Watcher_Specific_Functions_and_Data_-4">Watcher-Specific Functions and Data Members</h3>
1324 <div id="Watcher_Specific_Functions_and_Data_-2-2">
1326 <dt>ev_signal_init (ev_signal *, callback, int signum)</dt>
1327 <dt>ev_signal_set (ev_signal *, int signum)</dt>
1329 <p>Configures the watcher to trigger on the given signal number (usually one
1330 of the <code>SIGxxx</code> constants).</p>
1332 <dt>int signum [read-only]</dt>
1334 <p>The signal the watcher watches out for.</p>
1343 <h2 id="code_ev_child_code_watch_out_for_pro"><code>ev_child</code> - watch out for process status changes</h2>
1344 <div id="code_ev_child_code_watch_out_for_pro-2">
1345 <p>Child watchers trigger when your process receives a SIGCHLD in response to
1346 some child status changes (most typically when a child of yours dies).</p>
1349 <h3 id="Watcher_Specific_Functions_and_Data_-5">Watcher-Specific Functions and Data Members</h3>
1350 <div id="Watcher_Specific_Functions_and_Data_-2-3">
1352 <dt>ev_child_init (ev_child *, callback, int pid)</dt>
1353 <dt>ev_child_set (ev_child *, int pid)</dt>
1355 <p>Configures the watcher to wait for status changes of process <code>pid</code> (or
1356 <i>any</i> process if <code>pid</code> is specified as <code>0</code>). The callback can look
1357 at the <code>rstatus</code> member of the <code>ev_child</code> watcher structure to see
1358 the status word (use the macros from <code>sys/wait.h</code> and see your systems
1359 <code>waitpid</code> documentation). The <code>rpid</code> member contains the pid of the
1360 process causing the status change.</p>
1362 <dt>int pid [read-only]</dt>
1364 <p>The process id this watcher watches out for, or <code>0</code>, meaning any process id.</p>
1366 <dt>int rpid [read-write]</dt>
1368 <p>The process id that detected a status change.</p>
1370 <dt>int rstatus [read-write]</dt>
1372 <p>The process exit/trace status caused by <code>rpid</code> (see your systems
1373 <code>waitpid</code> and <code>sys/wait.h</code> documentation for details).</p>
1376 <p>Example: Try to exit cleanly on SIGINT and SIGTERM.</p>
1378 sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
1380 ev_unloop (loop, EVUNLOOP_ALL);
1383 struct ev_signal signal_watcher;
1384 ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1385 ev_signal_start (loop, &sigint_cb);
1393 <h2 id="code_ev_stat_code_did_the_file_attri"><code>ev_stat</code> - did the file attributes just change?</h2>
1394 <div id="code_ev_stat_code_did_the_file_attri-2">
1395 <p>This watches a filesystem path for attribute changes. That is, it calls
1396 <code>stat</code> regularly (or when the OS says it changed) and sees if it changed
1397 compared to the last time, invoking the callback if it did.</p>
1398 <p>The path does not need to exist: changing from "path exists" to "path does
1399 not exist" is a status change like any other. The condition "path does
1400 not exist" is signified by the <code>st_nlink</code> field being zero (which is
1401 otherwise always forced to be at least one) and all the other fields of
1402 the stat buffer having unspecified contents.</p>
1403 <p>The path <i>should</i> be absolute and <i>must not</i> end in a slash. If it is
1404 relative and your working directory changes, the behaviour is undefined.</p>
1405 <p>Since there is no standard to do this, the portable implementation simply
1406 calls <code>stat (2)</code> regularly on the path to see if it changed somehow. You
1407 can specify a recommended polling interval for this case. If you specify
1408 a polling interval of <code>0</code> (highly recommended!) then a <i>suitable,
1409 unspecified default</i> value will be used (which you can expect to be around
1410 five seconds, although this might change dynamically). Libev will also
1411 impose a minimum interval which is currently around <code>0.1</code>, but thats
1412 usually overkill.</p>
1413 <p>This watcher type is not meant for massive numbers of stat watchers,
1414 as even with OS-supported change notifications, this can be
1415 resource-intensive.</p>
1416 <p>At the time of this writing, only the Linux inotify interface is
1417 implemented (implementing kqueue support is left as an exercise for the
1418 reader). Inotify will be used to give hints only and should not change the
1419 semantics of <code>ev_stat</code> watchers, which means that libev sometimes needs
1420 to fall back to regular polling again even with inotify, but changes are
1421 usually detected immediately, and if the file exists there will be no
1425 <h3 id="Watcher_Specific_Functions_and_Data_-6">Watcher-Specific Functions and Data Members</h3>
1426 <div id="Watcher_Specific_Functions_and_Data_-2-4">
1428 <dt>ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)</dt>
1429 <dt>ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)</dt>
1431 <p>Configures the watcher to wait for status changes of the given
1432 <code>path</code>. The <code>interval</code> is a hint on how quickly a change is expected to
1433 be detected and should normally be specified as <code>0</code> to let libev choose
1434 a suitable value. The memory pointed to by <code>path</code> must point to the same
1435 path for as long as the watcher is active.</p>
1436 <p>The callback will be receive <code>EV_STAT</code> when a change was detected,
1437 relative to the attributes at the time the watcher was started (or the
1438 last change was detected).</p>
1440 <dt>ev_stat_stat (ev_stat *)</dt>
1442 <p>Updates the stat buffer immediately with new values. If you change the
1443 watched path in your callback, you could call this fucntion to avoid
1444 detecting this change (while introducing a race condition). Can also be
1445 useful simply to find out the new values.</p>
1447 <dt>ev_statdata attr [read-only]</dt>
1449 <p>The most-recently detected attributes of the file. Although the type is of
1450 <code>ev_statdata</code>, this is usually the (or one of the) <code>struct stat</code> types
1451 suitable for your system. If the <code>st_nlink</code> member is <code>0</code>, then there
1452 was some error while <code>stat</code>ing the file.</p>
1454 <dt>ev_statdata prev [read-only]</dt>
1456 <p>The previous attributes of the file. The callback gets invoked whenever
1457 <code>prev</code> != <code>attr</code>.</p>
1459 <dt>ev_tstamp interval [read-only]</dt>
1461 <p>The specified interval.</p>
1463 <dt>const char *path [read-only]</dt>
1465 <p>The filesystem path that is being watched.</p>
1468 <p>Example: Watch <code>/etc/passwd</code> for attribute changes.</p>
1470 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
1472 /* /etc/passwd changed in some way */
1473 if (w->attr.st_nlink)
1475 printf ("passwd current size %ld\n", (long)w->attr.st_size);
1476 printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
1477 printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
1480 /* you shalt not abuse printf for puts */
1481 puts ("wow, /etc/passwd is not there, expect problems. "
1482 "if this is windows, they already arrived\n");
1488 ev_stat_init (&passwd, passwd_cb, "/etc/passwd");
1489 ev_stat_start (loop, &passwd);
1497 <h2 id="code_ev_idle_code_when_you_ve_got_no"><code>ev_idle</code> - when you've got nothing better to do...</h2>
1498 <div id="code_ev_idle_code_when_you_ve_got_no-2">
1499 <p>Idle watchers trigger events when no other events of the same or higher
1500 priority are pending (prepare, check and other idle watchers do not
1502 <p>That is, as long as your process is busy handling sockets or timeouts
1503 (or even signals, imagine) of the same or higher priority it will not be
1504 triggered. But when your process is idle (or only lower-priority watchers
1505 are pending), the idle watchers are being called once per event loop
1506 iteration - until stopped, that is, or your process receives more events
1507 and becomes busy again with higher priority stuff.</p>
1508 <p>The most noteworthy effect is that as long as any idle watchers are
1509 active, the process will not block when waiting for new events.</p>
1510 <p>Apart from keeping your process non-blocking (which is a useful
1511 effect on its own sometimes), idle watchers are a good place to do
1512 "pseudo-background processing", or delay processing stuff to after the
1513 event loop has handled all outstanding events.</p>
1516 <h3 id="Watcher_Specific_Functions_and_Data_-7">Watcher-Specific Functions and Data Members</h3>
1517 <div id="Watcher_Specific_Functions_and_Data_-2-5">
1519 <dt>ev_idle_init (ev_signal *, callback)</dt>
1521 <p>Initialises and configures the idle watcher - it has no parameters of any
1522 kind. There is a <code>ev_idle_set</code> macro, but using it is utterly pointless,
1526 <p>Example: Dynamically allocate an <code>ev_idle</code> watcher, start it, and in the
1527 callback, free it. Also, use no error checking, as usual.</p>
1529 idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)
1532 // now do something you wanted to do when the program has
1533 // no longer asnything immediate to do.
1536 struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));
1537 ev_idle_init (idle_watcher, idle_cb);
1538 ev_idle_start (loop, idle_cb);
1546 <h2 id="code_ev_prepare_code_and_code_ev_che"><code>ev_prepare</code> and <code>ev_check</code> - customise your event loop!</h2>
1547 <div id="code_ev_prepare_code_and_code_ev_che-2">
1548 <p>Prepare and check watchers are usually (but not always) used in tandem:
1549 prepare watchers get invoked before the process blocks and check watchers
1551 <p>You <i>must not</i> call <code>ev_loop</code> or similar functions that enter
1552 the current event loop from either <code>ev_prepare</code> or <code>ev_check</code>
1553 watchers. Other loops than the current one are fine, however. The
1554 rationale behind this is that you do not need to check for recursion in
1555 those watchers, i.e. the sequence will always be <code>ev_prepare</code>, blocking,
1556 <code>ev_check</code> so if you have one watcher of each kind they will always be
1557 called in pairs bracketing the blocking call.</p>
1558 <p>Their main purpose is to integrate other event mechanisms into libev and
1559 their use is somewhat advanced. This could be used, for example, to track
1560 variable changes, implement your own watchers, integrate net-snmp or a
1561 coroutine library and lots more. They are also occasionally useful if
1562 you cache some data and want to flush it before blocking (for example,
1563 in X programs you might want to do an <code>XFlush ()</code> in an <code>ev_prepare</code>
1565 <p>This is done by examining in each prepare call which file descriptors need
1566 to be watched by the other library, registering <code>ev_io</code> watchers for
1567 them and starting an <code>ev_timer</code> watcher for any timeouts (many libraries
1568 provide just this functionality). Then, in the check watcher you check for
1569 any events that occured (by checking the pending status of all watchers
1570 and stopping them) and call back into the library. The I/O and timer
1571 callbacks will never actually be called (but must be valid nevertheless,
1572 because you never know, you know?).</p>
1573 <p>As another example, the Perl Coro module uses these hooks to integrate
1574 coroutines into libev programs, by yielding to other active coroutines
1575 during each prepare and only letting the process block if no coroutines
1576 are ready to run (it's actually more complicated: it only runs coroutines
1577 with priority higher than or equal to the event loop and one coroutine
1578 of lower priority, but only once, using idle watchers to keep the event
1579 loop from blocking if lower-priority coroutines are active, thus mapping
1580 low-priority coroutines to idle/background tasks).</p>
1581 <p>It is recommended to give <code>ev_check</code> watchers highest (<code>EV_MAXPRI</code>)
1582 priority, to ensure that they are being run before any other watchers
1583 after the poll. Also, <code>ev_check</code> watchers (and <code>ev_prepare</code> watchers,
1584 too) should not activate ("feed") events into libev. While libev fully
1585 supports this, they will be called before other <code>ev_check</code> watchers did
1586 their job. As <code>ev_check</code> watchers are often used to embed other event
1587 loops those other event loops might be in an unusable state until their
1588 <code>ev_check</code> watcher ran (always remind yourself to coexist peacefully with
1592 <h3 id="Watcher_Specific_Functions_and_Data_-8">Watcher-Specific Functions and Data Members</h3>
1593 <div id="Watcher_Specific_Functions_and_Data_-2-6">
1595 <dt>ev_prepare_init (ev_prepare *, callback)</dt>
1596 <dt>ev_check_init (ev_check *, callback)</dt>
1598 <p>Initialises and configures the prepare or check watcher - they have no
1599 parameters of any kind. There are <code>ev_prepare_set</code> and <code>ev_check_set</code>
1600 macros, but using them is utterly, utterly and completely pointless.</p>
1603 <p>There are a number of principal ways to embed other event loops or modules
1604 into libev. Here are some ideas on how to include libadns into libev
1605 (there is a Perl module named <code>EV::ADNS</code> that does this, which you could
1606 use for an actually working example. Another Perl module named <code>EV::Glib</code>
1607 embeds a Glib main context into libev, and finally, <code>Glib::EV</code> embeds EV
1608 into the Glib event loop).</p>
1609 <p>Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1610 and in a check watcher, destroy them and call into libadns. What follows
1611 is pseudo-code only of course. This requires you to either use a low
1612 priority for the check watcher or use <code>ev_clear_pending</code> explicitly, as
1613 the callbacks for the IO/timeout watchers might not have been called yet.</p>
1614 <pre> static ev_io iow [nfd];
1618 io_cb (ev_loop *loop, ev_io *w, int revents)
1622 // create io watchers for each fd and a timer before blocking
1624 adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
1626 int timeout = 3600000;
1627 struct pollfd fds [nfd];
1628 // actual code will need to loop here and realloc etc.
1629 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1631 /* the callback is illegal, but won't be called as we stop during check */
1632 ev_timer_init (&tw, 0, timeout * 1e-3);
1633 ev_timer_start (loop, &tw);
1635 // create one ev_io per pollfd
1636 for (int i = 0; i < nfd; ++i)
1638 ev_io_init (iow + i, io_cb, fds [i].fd,
1639 ((fds [i].events & POLLIN ? EV_READ : 0)
1640 | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
1642 fds [i].revents = 0;
1643 ev_io_start (loop, iow + i);
1647 // stop all watchers after blocking
1649 adns_check_cb (ev_loop *loop, ev_check *w, int revents)
1651 ev_timer_stop (loop, &tw);
1653 for (int i = 0; i < nfd; ++i)
1655 // set the relevant poll flags
1656 // could also call adns_processreadable etc. here
1657 struct pollfd *fd = fds + i;
1658 int revents = ev_clear_pending (iow + i);
1659 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1660 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1662 // now stop the watcher
1663 ev_io_stop (loop, iow + i);
1666 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
1670 <p>Method 2: This would be just like method 1, but you run <code>adns_afterpoll</code>
1671 in the prepare watcher and would dispose of the check watcher.</p>
1672 <p>Method 3: If the module to be embedded supports explicit event
1673 notification (adns does), you can also make use of the actual watcher
1674 callbacks, and only destroy/create the watchers in the prepare watcher.</p>
1676 timer_cb (EV_P_ ev_timer *w, int revents)
1678 adns_state ads = (adns_state)w->data;
1681 adns_processtimeouts (ads, &tv_now);
1685 io_cb (EV_P_ ev_io *w, int revents)
1687 adns_state ads = (adns_state)w->data;
1690 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
1691 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
1694 // do not ever call adns_afterpoll
1697 <p>Method 4: Do not use a prepare or check watcher because the module you
1698 want to embed is too inflexible to support it. Instead, youc na override
1699 their poll function. The drawback with this solution is that the main
1700 loop is now no longer controllable by EV. The <code>Glib::EV</code> module does
1703 event_poll_func (GPollFD *fds, guint nfds, gint timeout)
1707 for (n = 0; n < nfds; ++n)
1708 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
1710 if (timeout >= 0)
1711 // create/start timer
1717 if (timeout >= 0)
1718 ev_timer_stop (EV_A_ &to);
1720 // stop io watchers again - their callbacks should have set
1721 for (n = 0; n < nfds; ++n)
1722 ev_io_stop (EV_A_ iow [n]);
1733 <h2 id="code_ev_embed_code_when_one_backend_"><code>ev_embed</code> - when one backend isn't enough...</h2>
1734 <div id="code_ev_embed_code_when_one_backend_-2">
1735 <p>This is a rather advanced watcher type that lets you embed one event loop
1736 into another (currently only <code>ev_io</code> events are supported in the embedded
1737 loop, other types of watchers might be handled in a delayed or incorrect
1738 fashion and must not be used).</p>
1739 <p>There are primarily two reasons you would want that: work around bugs and
1741 <p>As an example for a bug workaround, the kqueue backend might only support
1742 sockets on some platform, so it is unusable as generic backend, but you
1743 still want to make use of it because you have many sockets and it scales
1744 so nicely. In this case, you would create a kqueue-based loop and embed it
1745 into your default loop (which might use e.g. poll). Overall operation will
1746 be a bit slower because first libev has to poll and then call kevent, but
1747 at least you can use both at what they are best.</p>
1748 <p>As for prioritising I/O: rarely you have the case where some fds have
1749 to be watched and handled very quickly (with low latency), and even
1750 priorities and idle watchers might have too much overhead. In this case
1751 you would put all the high priority stuff in one loop and all the rest in
1752 a second one, and embed the second one in the first.</p>
1753 <p>As long as the watcher is active, the callback will be invoked every time
1754 there might be events pending in the embedded loop. The callback must then
1755 call <code>ev_embed_sweep (mainloop, watcher)</code> to make a single sweep and invoke
1756 their callbacks (you could also start an idle watcher to give the embedded
1757 loop strictly lower priority for example). You can also set the callback
1758 to <code>0</code>, in which case the embed watcher will automatically execute the
1759 embedded loop sweep.</p>
1760 <p>As long as the watcher is started it will automatically handle events. The
1761 callback will be invoked whenever some events have been handled. You can
1762 set the callback to <code>0</code> to avoid having to specify one if you are not
1763 interested in that.</p>
1764 <p>Also, there have not currently been made special provisions for forking:
1765 when you fork, you not only have to call <code>ev_loop_fork</code> on both loops,
1766 but you will also have to stop and restart any <code>ev_embed</code> watchers
1768 <p>Unfortunately, not all backends are embeddable, only the ones returned by
1769 <code>ev_embeddable_backends</code> are, which, unfortunately, does not include any
1771 <p>So when you want to use this feature you will always have to be prepared
1772 that you cannot get an embeddable loop. The recommended way to get around
1773 this is to have a separate variables for your embeddable loop, try to
1774 create it, and if that fails, use the normal loop for everything:</p>
1775 <pre> struct ev_loop *loop_hi = ev_default_init (0);
1776 struct ev_loop *loop_lo = 0;
1777 struct ev_embed embed;
1779 // see if there is a chance of getting one that works
1780 // (remember that a flags value of 0 means autodetection)
1781 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
1782 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
1785 // if we got one, then embed it, otherwise default to loop_hi
1788 ev_embed_init (&embed, 0, loop_lo);
1789 ev_embed_start (loop_hi, &embed);
1797 <h3 id="Watcher_Specific_Functions_and_Data_-9">Watcher-Specific Functions and Data Members</h3>
1798 <div id="Watcher_Specific_Functions_and_Data_-2-7">
1800 <dt>ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)</dt>
1801 <dt>ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)</dt>
1803 <p>Configures the watcher to embed the given loop, which must be
1804 embeddable. If the callback is <code>0</code>, then <code>ev_embed_sweep</code> will be
1805 invoked automatically, otherwise it is the responsibility of the callback
1806 to invoke it (it will continue to be called until the sweep has been done,
1807 if you do not want thta, you need to temporarily stop the embed watcher).</p>
1809 <dt>ev_embed_sweep (loop, ev_embed *)</dt>
1811 <p>Make a single, non-blocking sweep over the embedded loop. This works
1812 similarly to <code>ev_loop (embedded_loop, EVLOOP_NONBLOCK)</code>, but in the most
1813 apropriate way for embedded loops.</p>
1815 <dt>struct ev_loop *loop [read-only]</dt>
1817 <p>The embedded event loop.</p>
1826 <h2 id="code_ev_fork_code_the_audacity_to_re"><code>ev_fork</code> - the audacity to resume the event loop after a fork</h2>
1827 <div id="code_ev_fork_code_the_audacity_to_re-2">
1828 <p>Fork watchers are called when a <code>fork ()</code> was detected (usually because
1829 whoever is a good citizen cared to tell libev about it by calling
1830 <code>ev_default_fork</code> or <code>ev_loop_fork</code>). The invocation is done before the
1831 event loop blocks next and before <code>ev_check</code> watchers are being called,
1832 and only in the child after the fork. If whoever good citizen calling
1833 <code>ev_default_fork</code> cheats and calls it in the wrong process, the fork
1834 handlers will be invoked, too, of course.</p>
1836 <dt>ev_fork_init (ev_signal *, callback)</dt>
1838 <p>Initialises and configures the fork watcher - it has no parameters of any
1839 kind. There is a <code>ev_fork_set</code> macro, but using it is utterly pointless,
1849 <h1 id="OTHER_FUNCTIONS">OTHER FUNCTIONS</h1>
1850 <div id="OTHER_FUNCTIONS_CONTENT">
1851 <p>There are some other functions of possible interest. Described. Here. Now.</p>
1853 <dt>ev_once (loop, int fd, int events, ev_tstamp timeout, callback)</dt>
1855 <p>This function combines a simple timer and an I/O watcher, calls your
1856 callback on whichever event happens first and automatically stop both
1857 watchers. This is useful if you want to wait for a single event on an fd
1858 or timeout without having to allocate/configure/start/stop/free one or
1859 more watchers yourself.</p>
1860 <p>If <code>fd</code> is less than 0, then no I/O watcher will be started and events
1861 is being ignored. Otherwise, an <code>ev_io</code> watcher for the given <code>fd</code> and
1862 <code>events</code> set will be craeted and started.</p>
1863 <p>If <code>timeout</code> is less than 0, then no timeout watcher will be
1864 started. Otherwise an <code>ev_timer</code> watcher with after = <code>timeout</code> (and
1865 repeat = 0) will be started. While <code>0</code> is a valid timeout, it is of
1867 <p>The callback has the type <code>void (*cb)(int revents, void *arg)</code> and gets
1868 passed an <code>revents</code> set like normal event callbacks (a combination of
1869 <code>EV_ERROR</code>, <code>EV_READ</code>, <code>EV_WRITE</code> or <code>EV_TIMEOUT</code>) and the <code>arg</code>
1870 value passed to <code>ev_once</code>:</p>
1871 <pre> static void stdin_ready (int revents, void *arg)
1873 if (revents & EV_TIMEOUT)
1874 /* doh, nothing entered */;
1875 else if (revents & EV_READ)
1876 /* stdin might have data for us, joy! */;
1879 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
1883 <dt>ev_feed_event (ev_loop *, watcher *, int revents)</dt>
1885 <p>Feeds the given event set into the event loop, as if the specified event
1886 had happened for the specified watcher (which must be a pointer to an
1887 initialised but not necessarily started event watcher).</p>
1889 <dt>ev_feed_fd_event (ev_loop *, int fd, int revents)</dt>
1891 <p>Feed an event on the given fd, as if a file descriptor backend detected
1892 the given events it.</p>
1894 <dt>ev_feed_signal_event (ev_loop *loop, int signum)</dt>
1896 <p>Feed an event as if the given signal occured (<code>loop</code> must be the default
1906 <h1 id="LIBEVENT_EMULATION">LIBEVENT EMULATION</h1>
1907 <div id="LIBEVENT_EMULATION_CONTENT">
1908 <p>Libev offers a compatibility emulation layer for libevent. It cannot
1909 emulate the internals of libevent, so here are some usage hints:</p>
1911 <dt>* Use it by including <event.h>, as usual.</dt>
1912 <dt>* The following members are fully supported: ev_base, ev_callback,
1913 ev_arg, ev_fd, ev_res, ev_events.</dt>
1914 <dt>* Avoid using ev_flags and the EVLIST_*-macros, while it is
1915 maintained by libev, it does not work exactly the same way as in libevent (consider
1916 it a private API).</dt>
1917 <dt>* Priorities are not currently supported. Initialising priorities
1918 will fail and all watchers will have the same priority, even though there
1919 is an ev_pri field.</dt>
1920 <dt>* Other members are not supported.</dt>
1921 <dt>* The libev emulation is <i>not</i> ABI compatible to libevent, you need
1922 to use the libev header file and library.</dt>
1926 <h1 id="C_SUPPORT">C++ SUPPORT</h1>
1927 <div id="C_SUPPORT_CONTENT">
1928 <p>Libev comes with some simplistic wrapper classes for C++ that mainly allow
1929 you to use some convinience methods to start/stop watchers and also change
1930 the callback model to a model using method callbacks on objects.</p>
1932 <pre> #include <ev++.h>
1935 <p>This automatically includes <cite>ev.h</cite> and puts all of its definitions (many
1936 of them macros) into the global namespace. All C++ specific things are
1937 put into the <code>ev</code> namespace. It should support all the same embedding
1938 options as <cite>ev.h</cite>, most notably <code>EV_MULTIPLICITY</code>.</p>
1939 <p>Care has been taken to keep the overhead low. The only data member the C++
1940 classes add (compared to plain C-style watchers) is the event loop pointer
1941 that the watcher is associated with (or no additional members at all if
1942 you disable <code>EV_MULTIPLICITY</code> when embedding libev).</p>
1943 <p>Currently, functions, and static and non-static member functions can be
1944 used as callbacks. Other types should be easy to add as long as they only
1945 need one additional pointer for context. If you need support for other
1946 types of functors please contact the author (preferably after implementing
1948 <p>Here is a list of things available in the <code>ev</code> namespace:</p>
1950 <dt><code>ev::READ</code>, <code>ev::WRITE</code> etc.</dt>
1952 <p>These are just enum values with the same values as the <code>EV_READ</code> etc.
1953 macros from <cite>ev.h</cite>.</p>
1955 <dt><code>ev::tstamp</code>, <code>ev::now</code></dt>
1957 <p>Aliases to the same types/functions as with the <code>ev_</code> prefix.</p>
1959 <dt><code>ev::io</code>, <code>ev::timer</code>, <code>ev::periodic</code>, <code>ev::idle</code>, <code>ev::sig</code> etc.</dt>
1961 <p>For each <code>ev_TYPE</code> watcher in <cite>ev.h</cite> there is a corresponding class of
1962 the same name in the <code>ev</code> namespace, with the exception of <code>ev_signal</code>
1963 which is called <code>ev::sig</code> to avoid clashes with the <code>signal</code> macro
1964 defines by many implementations.</p>
1965 <p>All of those classes have these methods:</p>
1968 <dt>ev::TYPE::TYPE ()</dt>
1969 <dt>ev::TYPE::TYPE (struct ev_loop *)</dt>
1970 <dt>ev::TYPE::~TYPE</dt>
1972 <p>The constructor (optionally) takes an event loop to associate the watcher
1973 with. If it is omitted, it will use <code>EV_DEFAULT</code>.</p>
1974 <p>The constructor calls <code>ev_init</code> for you, which means you have to call the
1975 <code>set</code> method before starting it.</p>
1976 <p>It will not set a callback, however: You have to call the templated <code>set</code>
1977 method to set a callback before you can start the watcher.</p>
1978 <p>(The reason why you have to use a method is a limitation in C++ which does
1979 not allow explicit template arguments for constructors).</p>
1980 <p>The destructor automatically stops the watcher if it is active.</p>
1982 <dt>w->set<class, &class::method> (object *)</dt>
1984 <p>This method sets the callback method to call. The method has to have a
1985 signature of <code>void (*)(ev_TYPE &, int)</code>, it receives the watcher as
1986 first argument and the <code>revents</code> as second. The object must be given as
1987 parameter and is stored in the <code>data</code> member of the watcher.</p>
1988 <p>This method synthesizes efficient thunking code to call your method from
1989 the C callback that libev requires. If your compiler can inline your
1990 callback (i.e. it is visible to it at the place of the <code>set</code> call and
1991 your compiler is good :), then the method will be fully inlined into the
1992 thunking function, making it as fast as a direct C callback.</p>
1993 <p>Example: simple class declaration and watcher initialisation</p>
1994 <pre> struct myclass
1996 void io_cb (ev::io &w, int revents) { }
2001 iow.set <myclass, &myclass::io_cb> (&obj);
2005 <dt>w->set<function> (void *data = 0)</dt>
2007 <p>Also sets a callback, but uses a static method or plain function as
2008 callback. The optional <code>data</code> argument will be stored in the watcher's
2009 <code>data</code> member and is free for you to use.</p>
2010 <p>The prototype of the <code>function</code> must be <code>void (*)(ev::TYPE &w, int)</code>.</p>
2011 <p>See the method-<code>set</code> above for more details.</p>
2013 <pre> static void io_cb (ev::io &w, int revents) { }
2014 iow.set <io_cb> ();
2018 <dt>w->set (struct ev_loop *)</dt>
2020 <p>Associates a different <code>struct ev_loop</code> with this watcher. You can only
2021 do this when the watcher is inactive (and not pending either).</p>
2023 <dt>w->set ([args])</dt>
2025 <p>Basically the same as <code>ev_TYPE_set</code>, with the same args. Must be
2026 called at least once. Unlike the C counterpart, an active watcher gets
2027 automatically stopped and restarted when reconfiguring it with this
2030 <dt>w->start ()</dt>
2032 <p>Starts the watcher. Note that there is no <code>loop</code> argument, as the
2033 constructor already stores the event loop.</p>
2035 <dt>w->stop ()</dt>
2037 <p>Stops the watcher if it is active. Again, no <code>loop</code> argument.</p>
2039 <dt>w->again () <code>ev::timer</code>, <code>ev::periodic</code> only</dt>
2041 <p>For <code>ev::timer</code> and <code>ev::periodic</code>, this invokes the corresponding
2042 <code>ev_TYPE_again</code> function.</p>
2044 <dt>w->sweep () <code>ev::embed</code> only</dt>
2046 <p>Invokes <code>ev_embed_sweep</code>.</p>
2048 <dt>w->update () <code>ev::stat</code> only</dt>
2050 <p>Invokes <code>ev_stat_stat</code>.</p>
2056 <p>Example: Define a class with an IO and idle watcher, start one of them in
2057 the constructor.</p>
2060 ev_io io; void io_cb (ev::io &w, int revents);
2061 ev_idle idle void idle_cb (ev::idle &w, int revents);
2066 myclass::myclass (int fd)
2068 io .set <myclass, &myclass::io_cb > (this);
2069 idle.set <myclass, &myclass::idle_cb> (this);
2071 io.start (fd, ev::READ);
2080 <h1 id="MACRO_MAGIC">MACRO MAGIC</h1>
2081 <div id="MACRO_MAGIC_CONTENT">
2082 <p>Libev can be compiled with a variety of options, the most fundemantal is
2083 <code>EV_MULTIPLICITY</code>. This option determines whether (most) functions and
2084 callbacks have an initial <code>struct ev_loop *</code> argument.</p>
2085 <p>To make it easier to write programs that cope with either variant, the
2086 following macros are defined:</p>
2088 <dt><code>EV_A</code>, <code>EV_A_</code></dt>
2090 <p>This provides the loop <i>argument</i> for functions, if one is required ("ev
2091 loop argument"). The <code>EV_A</code> form is used when this is the sole argument,
2092 <code>EV_A_</code> is used when other arguments are following. Example:</p>
2093 <pre> ev_unref (EV_A);
2094 ev_timer_add (EV_A_ watcher);
2098 <p>It assumes the variable <code>loop</code> of type <code>struct ev_loop *</code> is in scope,
2099 which is often provided by the following macro.</p>
2101 <dt><code>EV_P</code>, <code>EV_P_</code></dt>
2103 <p>This provides the loop <i>parameter</i> for functions, if one is required ("ev
2104 loop parameter"). The <code>EV_P</code> form is used when this is the sole parameter,
2105 <code>EV_P_</code> is used when other parameters are following. Example:</p>
2106 <pre> // this is how ev_unref is being declared
2107 static void ev_unref (EV_P);
2109 // this is how you can declare your typical callback
2110 static void cb (EV_P_ ev_timer *w, int revents)
2113 <p>It declares a parameter <code>loop</code> of type <code>struct ev_loop *</code>, quite
2114 suitable for use with <code>EV_A</code>.</p>
2116 <dt><code>EV_DEFAULT</code>, <code>EV_DEFAULT_</code></dt>
2118 <p>Similar to the other two macros, this gives you the value of the default
2119 loop, if multiple loops are supported ("ev loop default").</p>
2122 <p>Example: Declare and initialise a check watcher, utilising the above
2123 macros so it will work regardless of whether multiple loops are supported
2126 check_cb (EV_P_ ev_timer *w, int revents)
2128 ev_check_stop (EV_A_ w);
2132 ev_check_init (&check, check_cb);
2133 ev_check_start (EV_DEFAULT_ &check);
2134 ev_loop (EV_DEFAULT_ 0);
2139 <h1 id="EMBEDDING">EMBEDDING</h1>
2140 <div id="EMBEDDING_CONTENT">
2141 <p>Libev can (and often is) directly embedded into host
2142 applications. Examples of applications that embed it include the Deliantra
2143 Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
2144 and rxvt-unicode.</p>
2145 <p>The goal is to enable you to just copy the neecssary files into your
2146 source directory without having to change even a single line in them, so
2147 you can easily upgrade by simply copying (or having a checked-out copy of
2148 libev somewhere in your source tree).</p>
2151 <h2 id="FILESETS">FILESETS</h2>
2152 <div id="FILESETS_CONTENT">
2153 <p>Depending on what features you need you need to include one or more sets of files
2157 <h3 id="CORE_EVENT_LOOP">CORE EVENT LOOP</h3>
2158 <div id="CORE_EVENT_LOOP_CONTENT">
2159 <p>To include only the libev core (all the <code>ev_*</code> functions), with manual
2160 configuration (no autoconf):</p>
2161 <pre> #define EV_STANDALONE 1
2162 #include "ev.c"
2165 <p>This will automatically include <cite>ev.h</cite>, too, and should be done in a
2166 single C source file only to provide the function implementations. To use
2167 it, do the same for <cite>ev.h</cite> in all files wishing to use this API (best
2168 done by writing a wrapper around <cite>ev.h</cite> that you can include instead and
2169 where you can put other configuration options):</p>
2170 <pre> #define EV_STANDALONE 1
2171 #include "ev.h"
2174 <p>Both header files and implementation files can be compiled with a C++
2175 compiler (at least, thats a stated goal, and breakage will be treated
2177 <p>You need the following files in your source tree, or in a directory
2178 in your include path (e.g. in libev/ when using -Ilibev):</p>
2184 ev_win32.c required on win32 platforms only
2186 ev_select.c only when select backend is enabled (which is enabled by default)
2187 ev_poll.c only when poll backend is enabled (disabled by default)
2188 ev_epoll.c only when the epoll backend is enabled (disabled by default)
2189 ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
2190 ev_port.c only when the solaris port backend is enabled (disabled by default)
2193 <p><cite>ev.c</cite> includes the backend files directly when enabled, so you only need
2194 to compile this single file.</p>
2197 <h3 id="LIBEVENT_COMPATIBILITY_API">LIBEVENT COMPATIBILITY API</h3>
2198 <div id="LIBEVENT_COMPATIBILITY_API_CONTENT">
2199 <p>To include the libevent compatibility API, also include:</p>
2200 <pre> #include "event.c"
2203 <p>in the file including <cite>ev.c</cite>, and:</p>
2204 <pre> #include "event.h"
2207 <p>in the files that want to use the libevent API. This also includes <cite>ev.h</cite>.</p>
2208 <p>You need the following additional files for this:</p>
2215 <h3 id="AUTOCONF_SUPPORT">AUTOCONF SUPPORT</h3>
2216 <div id="AUTOCONF_SUPPORT_CONTENT">
2217 <p>Instead of using <code>EV_STANDALONE=1</code> and providing your config in
2218 whatever way you want, you can also <code>m4_include([libev.m4])</code> in your
2219 <cite>configure.ac</cite> and leave <code>EV_STANDALONE</code> undefined. <cite>ev.c</cite> will then
2220 include <cite>config.h</cite> and configure itself accordingly.</p>
2221 <p>For this of course you need the m4 file:</p>
2227 <h2 id="PREPROCESSOR_SYMBOLS_MACROS">PREPROCESSOR SYMBOLS/MACROS</h2>
2228 <div id="PREPROCESSOR_SYMBOLS_MACROS_CONTENT">
2229 <p>Libev can be configured via a variety of preprocessor symbols you have to define
2230 before including any of its files. The default is not to build for multiplicity
2231 and only include the select backend.</p>
2233 <dt>EV_STANDALONE</dt>
2235 <p>Must always be <code>1</code> if you do not use autoconf configuration, which
2236 keeps libev from including <cite>config.h</cite>, and it also defines dummy
2237 implementations for some libevent functions (such as logging, which is not
2238 supported). It will also not define any of the structs usually found in
2239 <cite>event.h</cite> that are not directly supported by the libev core alone.</p>
2241 <dt>EV_USE_MONOTONIC</dt>
2243 <p>If defined to be <code>1</code>, libev will try to detect the availability of the
2244 monotonic clock option at both compiletime and runtime. Otherwise no use
2245 of the monotonic clock option will be attempted. If you enable this, you
2246 usually have to link against librt or something similar. Enabling it when
2247 the functionality isn't available is safe, though, althoguh you have
2248 to make sure you link against any libraries where the <code>clock_gettime</code>
2249 function is hiding in (often <cite>-lrt</cite>).</p>
2251 <dt>EV_USE_REALTIME</dt>
2253 <p>If defined to be <code>1</code>, libev will try to detect the availability of the
2254 realtime clock option at compiletime (and assume its availability at
2255 runtime if successful). Otherwise no use of the realtime clock option will
2256 be attempted. This effectively replaces <code>gettimeofday</code> by <code>clock_get
2257 (CLOCK_REALTIME, ...)</code> and will not normally affect correctness. See tzhe note about libraries
2258 in the description of <code>EV_USE_MONOTONIC</code>, though.</p>
2260 <dt>EV_USE_SELECT</dt>
2262 <p>If undefined or defined to be <code>1</code>, libev will compile in support for the
2263 <code>select</code>(2) backend. No attempt at autodetection will be done: if no
2264 other method takes over, select will be it. Otherwise the select backend
2265 will not be compiled in.</p>
2267 <dt>EV_SELECT_USE_FD_SET</dt>
2269 <p>If defined to <code>1</code>, then the select backend will use the system <code>fd_set</code>
2270 structure. This is useful if libev doesn't compile due to a missing
2271 <code>NFDBITS</code> or <code>fd_mask</code> definition or it misguesses the bitset layout on
2272 exotic systems. This usually limits the range of file descriptors to some
2273 low limit such as 1024 or might have other limitations (winsocket only
2274 allows 64 sockets). The <code>FD_SETSIZE</code> macro, set before compilation, might
2275 influence the size of the <code>fd_set</code> used.</p>
2277 <dt>EV_SELECT_IS_WINSOCKET</dt>
2279 <p>When defined to <code>1</code>, the select backend will assume that
2280 select/socket/connect etc. don't understand file descriptors but
2281 wants osf handles on win32 (this is the case when the select to
2282 be used is the winsock select). This means that it will call
2283 <code>_get_osfhandle</code> on the fd to convert it to an OS handle. Otherwise,
2284 it is assumed that all these functions actually work on fds, even
2285 on win32. Should not be defined on non-win32 platforms.</p>
2287 <dt>EV_USE_POLL</dt>
2289 <p>If defined to be <code>1</code>, libev will compile in support for the <code>poll</code>(2)
2290 backend. Otherwise it will be enabled on non-win32 platforms. It
2291 takes precedence over select.</p>
2293 <dt>EV_USE_EPOLL</dt>
2295 <p>If defined to be <code>1</code>, libev will compile in support for the Linux
2296 <code>epoll</code>(7) backend. Its availability will be detected at runtime,
2297 otherwise another method will be used as fallback. This is the
2298 preferred backend for GNU/Linux systems.</p>
2300 <dt>EV_USE_KQUEUE</dt>
2302 <p>If defined to be <code>1</code>, libev will compile in support for the BSD style
2303 <code>kqueue</code>(2) backend. Its actual availability will be detected at runtime,
2304 otherwise another method will be used as fallback. This is the preferred
2305 backend for BSD and BSD-like systems, although on most BSDs kqueue only
2306 supports some types of fds correctly (the only platform we found that
2307 supports ptys for example was NetBSD), so kqueue might be compiled in, but
2308 not be used unless explicitly requested. The best way to use it is to find
2309 out whether kqueue supports your type of fd properly and use an embedded
2312 <dt>EV_USE_PORT</dt>
2314 <p>If defined to be <code>1</code>, libev will compile in support for the Solaris
2315 10 port style backend. Its availability will be detected at runtime,
2316 otherwise another method will be used as fallback. This is the preferred
2317 backend for Solaris 10 systems.</p>
2319 <dt>EV_USE_DEVPOLL</dt>
2321 <p>reserved for future expansion, works like the USE symbols above.</p>
2323 <dt>EV_USE_INOTIFY</dt>
2325 <p>If defined to be <code>1</code>, libev will compile in support for the Linux inotify
2326 interface to speed up <code>ev_stat</code> watchers. Its actual availability will
2327 be detected at runtime.</p>
2331 <p>The name of the <cite>ev.h</cite> header file used to include it. The default if
2332 undefined is <code><ev.h></code> in <cite>event.h</cite> and <code>"ev.h"</code> in <cite>ev.c</cite>. This
2333 can be used to virtually rename the <cite>ev.h</cite> header file in case of conflicts.</p>
2335 <dt>EV_CONFIG_H</dt>
2337 <p>If <code>EV_STANDALONE</code> isn't <code>1</code>, this variable can be used to override
2338 <cite>ev.c</cite>'s idea of where to find the <cite>config.h</cite> file, similarly to
2339 <code>EV_H</code>, above.</p>
2343 <p>Similarly to <code>EV_H</code>, this macro can be used to override <cite>event.c</cite>'s idea
2344 of how the <cite>event.h</cite> header can be found.</p>
2346 <dt>EV_PROTOTYPES</dt>
2348 <p>If defined to be <code>0</code>, then <cite>ev.h</cite> will not define any function
2349 prototypes, but still define all the structs and other symbols. This is
2350 occasionally useful if you want to provide your own wrapper functions
2351 around libev functions.</p>
2353 <dt>EV_MULTIPLICITY</dt>
2355 <p>If undefined or defined to <code>1</code>, then all event-loop-specific functions
2356 will have the <code>struct ev_loop *</code> as first argument, and you can create
2357 additional independent event loops. Otherwise there will be no support
2358 for multiple event loops and there is no first event loop pointer
2359 argument. Instead, all functions act on the single default loop.</p>
2364 <p>The range of allowed priorities. <code>EV_MINPRI</code> must be smaller or equal to
2365 <code>EV_MAXPRI</code>, but otherwise there are no non-obvious limitations. You can
2366 provide for more priorities by overriding those symbols (usually defined
2367 to be <code>-2</code> and <code>2</code>, respectively).</p>
2368 <p>When doing priority-based operations, libev usually has to linearly search
2369 all the priorities, so having many of them (hundreds) uses a lot of space
2370 and time, so using the defaults of five priorities (-2 .. +2) is usually
2372 <p>If your embedding app does not need any priorities, defining these both to
2373 <code>0</code> will save some memory and cpu.</p>
2375 <dt>EV_PERIODIC_ENABLE</dt>
2377 <p>If undefined or defined to be <code>1</code>, then periodic timers are supported. If
2378 defined to be <code>0</code>, then they are not. Disabling them saves a few kB of
2381 <dt>EV_IDLE_ENABLE</dt>
2383 <p>If undefined or defined to be <code>1</code>, then idle watchers are supported. If
2384 defined to be <code>0</code>, then they are not. Disabling them saves a few kB of
2387 <dt>EV_EMBED_ENABLE</dt>
2389 <p>If undefined or defined to be <code>1</code>, then embed watchers are supported. If
2390 defined to be <code>0</code>, then they are not.</p>
2392 <dt>EV_STAT_ENABLE</dt>
2394 <p>If undefined or defined to be <code>1</code>, then stat watchers are supported. If
2395 defined to be <code>0</code>, then they are not.</p>
2397 <dt>EV_FORK_ENABLE</dt>
2399 <p>If undefined or defined to be <code>1</code>, then fork watchers are supported. If
2400 defined to be <code>0</code>, then they are not.</p>
2404 <p>If you need to shave off some kilobytes of code at the expense of some
2405 speed, define this symbol to <code>1</code>. Currently only used for gcc to override
2406 some inlining decisions, saves roughly 30% codesize of amd64.</p>
2408 <dt>EV_PID_HASHSIZE</dt>
2410 <p><code>ev_child</code> watchers use a small hash table to distribute workload by
2411 pid. The default size is <code>16</code> (or <code>1</code> with <code>EV_MINIMAL</code>), usually more
2412 than enough. If you need to manage thousands of children you might want to
2413 increase this value (<i>must</i> be a power of two).</p>
2415 <dt>EV_INOTIFY_HASHSIZE</dt>
2417 <p><code>ev_staz</code> watchers use a small hash table to distribute workload by
2418 inotify watch id. The default size is <code>16</code> (or <code>1</code> with <code>EV_MINIMAL</code>),
2419 usually more than enough. If you need to manage thousands of <code>ev_stat</code>
2420 watchers you might want to increase this value (<i>must</i> be a power of
2425 <p>By default, all watchers have a <code>void *data</code> member. By redefining
2426 this macro to a something else you can include more and other types of
2427 members. You have to define it each time you include one of the files,
2428 though, and it must be identical each time.</p>
2429 <p>For example, the perl EV module uses something like this:</p>
2430 <pre> #define EV_COMMON \
2431 SV *self; /* contains this struct */ \
2432 SV *cb_sv, *fh /* note no trailing ";" */
2436 <dt>EV_CB_DECLARE (type)</dt>
2437 <dt>EV_CB_INVOKE (watcher, revents)</dt>
2438 <dt>ev_set_cb (ev, cb)</dt>
2440 <p>Can be used to change the callback member declaration in each watcher,
2441 and the way callbacks are invoked and set. Must expand to a struct member
2442 definition and a statement, respectively. See the <cite>ev.v</cite> header file for
2443 their default definitions. One possible use for overriding these is to
2444 avoid the <code>struct ev_loop *</code> as first argument in all cases, or to use
2445 method calls instead of plain function calls in C++.</p>
2448 <h2 id="EXAMPLES">EXAMPLES</h2>
2449 <div id="EXAMPLES_CONTENT">
2450 <p>For a real-world example of a program the includes libev
2451 verbatim, you can have a look at the EV perl module
2452 (<a href="http://software.schmorp.de/pkg/EV.html">http://software.schmorp.de/pkg/EV.html</a>). It has the libev files in
2453 the <cite>libev/</cite> subdirectory and includes them in the <cite>EV/EVAPI.h</cite> (public
2454 interface) and <cite>EV.xs</cite> (implementation) files. Only the <cite>EV.xs</cite> file
2455 will be compiled. It is pretty complex because it provides its own header
2457 <p>The usage in rxvt-unicode is simpler. It has a <cite>ev_cpp.h</cite> header file
2458 that everybody includes and which overrides some configure choices:</p>
2459 <pre> #define EV_MINIMAL 1
2460 #define EV_USE_POLL 0
2461 #define EV_MULTIPLICITY 0
2462 #define EV_PERIODIC_ENABLE 0
2463 #define EV_STAT_ENABLE 0
2464 #define EV_FORK_ENABLE 0
2465 #define EV_CONFIG_H <config.h>
2469 #include "ev++.h"
2472 <p>And a <cite>ev_cpp.C</cite> implementation file that contains libev proper and is compiled:</p>
2473 <pre> #include "ev_cpp.h"
2474 #include "ev.c"
2482 <h1 id="COMPLEXITIES">COMPLEXITIES</h1>
2483 <div id="COMPLEXITIES_CONTENT">
2484 <p>In this section the complexities of (many of) the algorithms used inside
2485 libev will be explained. For complexity discussions about backends see the
2486 documentation for <code>ev_default_init</code>.</p>
2487 <p>All of the following are about amortised time: If an array needs to be
2488 extended, libev needs to realloc and move the whole array, but this
2489 happens asymptotically never with higher number of elements, so O(1) might
2490 mean it might do a lengthy realloc operation in rare cases, but on average
2491 it is much faster and asymptotically approaches constant time.</p>
2494 <dt>Starting and stopping timer/periodic watchers: O(log skipped_other_timers)</dt>
2496 <p>This means that, when you have a watcher that triggers in one hour and
2497 there are 100 watchers that would trigger before that then inserting will
2498 have to skip those 100 watchers.</p>
2500 <dt>Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)</dt>
2502 <p>That means that for changing a timer costs less than removing/adding them
2503 as only the relative motion in the event queue has to be paid for.</p>
2505 <dt>Starting io/check/prepare/idle/signal/child watchers: O(1)</dt>
2507 <p>These just add the watcher into an array or at the head of a list.
2508 =item Stopping check/prepare/idle watchers: O(1)</p>
2510 <dt>Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))</dt>
2512 <p>These watchers are stored in lists then need to be walked to find the
2513 correct watcher to remove. The lists are usually short (you don't usually
2514 have many watchers waiting for the same fd or signal).</p>
2516 <dt>Finding the next timer per loop iteration: O(1)</dt>
2517 <dt>Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)</dt>
2519 <p>A change means an I/O watcher gets started or stopped, which requires
2520 libev to recalculate its status (and possibly tell the kernel).</p>
2522 <dt>Activating one watcher: O(1)</dt>
2523 <dt>Priority handling: O(number_of_priorities)</dt>
2525 <p>Priorities are implemented by allocating some space for each
2526 priority. When doing priority-based operations, libev usually has to
2527 linearly search all the priorities.</p>
2537 <h1 id="AUTHOR">AUTHOR</h1>
2538 <div id="AUTHOR_CONTENT">
2539 <p>Marc Lehmann <libev@schmorp.de>.</p>
projects / software / libev.git / blob