ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev.3
Revision: 1.72
Committed: Tue Oct 21 20:06:52 2008 UTC (15 years, 6 months ago) by root
Branch: MAIN
CVS Tags: rel-3_45
Changes since 1.71: +186 -159 lines
Log Message:
3.45

File Contents

# Content
1 .\" Automatically generated by Pod::Man 2.16 (Pod::Simple 3.05)
2 .\"
3 .\" Standard preamble:
4 .\" ========================================================================
5 .de Sh \" Subsection heading
6 .br
7 .if t .Sp
8 .ne 5
9 .PP
10 \fB\\$1\fR
11 .PP
12 ..
13 .de Sp \" Vertical space (when we can't use .PP)
14 .if t .sp .5v
15 .if n .sp
16 ..
17 .de Vb \" Begin verbatim text
18 .ft CW
19 .nf
20 .ne \\$1
21 ..
22 .de Ve \" End verbatim text
23 .ft R
24 .fi
25 ..
26 .\" Set up some character translations and predefined strings. \*(-- will
27 .\" give an unbreakable dash, \*(PI will give pi, \*(L" will give a left
28 .\" double quote, and \*(R" will give a right double quote. \*(C+ will
29 .\" give a nicer C++. Capital omega is used to do unbreakable dashes and
30 .\" therefore won't be available. \*(C` and \*(C' expand to `' in nroff,
31 .\" nothing in troff, for use with C<>.
32 .tr \(*W-
33 .ds C+ C\v'-.1v'\h'-1p'\s-2+\h'-1p'+\s0\v'.1v'\h'-1p'
34 .ie n \{\
35 . ds -- \(*W-
36 . ds PI pi
37 . if (\n(.H=4u)&(1m=24u) .ds -- \(*W\h'-12u'\(*W\h'-12u'-\" diablo 10 pitch
38 . if (\n(.H=4u)&(1m=20u) .ds -- \(*W\h'-12u'\(*W\h'-8u'-\" diablo 12 pitch
39 . ds L" ""
40 . ds R" ""
41 . ds C` ""
42 . ds C' ""
43 'br\}
44 .el\{\
45 . ds -- \|\(em\|
46 . ds PI \(*p
47 . ds L" ``
48 . ds R" ''
49 'br\}
50 .\"
51 .\" Escape single quotes in literal strings from groff's Unicode transform.
52 .ie \n(.g .ds Aq \(aq
53 .el .ds Aq '
54 .\"
55 .\" If the F register is turned on, we'll generate index entries on stderr for
56 .\" titles (.TH), headers (.SH), subsections (.Sh), items (.Ip), and index
57 .\" entries marked with X<> in POD. Of course, you'll have to process the
58 .\" output yourself in some meaningful fashion.
59 .ie \nF \{\
60 . de IX
61 . tm Index:\\$1\t\\n%\t"\\$2"
62 ..
63 . nr % 0
64 . rr F
65 .\}
66 .el \{\
67 . de IX
68 ..
69 .\}
70 .\"
71 .\" Accent mark definitions (@(#)ms.acc 1.5 88/02/08 SMI; from UCB 4.2).
72 .\" Fear. Run. Save yourself. No user-serviceable parts.
73 . \" fudge factors for nroff and troff
74 .if n \{\
75 . ds #H 0
76 . ds #V .8m
77 . ds #F .3m
78 . ds #[ \f1
79 . ds #] \fP
80 .\}
81 .if t \{\
82 . ds #H ((1u-(\\\\n(.fu%2u))*.13m)
83 . ds #V .6m
84 . ds #F 0
85 . ds #[ \&
86 . ds #] \&
87 .\}
88 . \" simple accents for nroff and troff
89 .if n \{\
90 . ds ' \&
91 . ds ` \&
92 . ds ^ \&
93 . ds , \&
94 . ds ~ ~
95 . ds /
96 .\}
97 .if t \{\
98 . ds ' \\k:\h'-(\\n(.wu*8/10-\*(#H)'\'\h"|\\n:u"
99 . ds ` \\k:\h'-(\\n(.wu*8/10-\*(#H)'\`\h'|\\n:u'
100 . ds ^ \\k:\h'-(\\n(.wu*10/11-\*(#H)'^\h'|\\n:u'
101 . ds , \\k:\h'-(\\n(.wu*8/10)',\h'|\\n:u'
102 . ds ~ \\k:\h'-(\\n(.wu-\*(#H-.1m)'~\h'|\\n:u'
103 . ds / \\k:\h'-(\\n(.wu*8/10-\*(#H)'\z\(sl\h'|\\n:u'
104 .\}
105 . \" troff and (daisy-wheel) nroff accents
106 .ds : \\k:\h'-(\\n(.wu*8/10-\*(#H+.1m+\*(#F)'\v'-\*(#V'\z.\h'.2m+\*(#F'.\h'|\\n:u'\v'\*(#V'
107 .ds 8 \h'\*(#H'\(*b\h'-\*(#H'
108 .ds o \\k:\h'-(\\n(.wu+\w'\(de'u-\*(#H)/2u'\v'-.3n'\*(#[\z\(de\v'.3n'\h'|\\n:u'\*(#]
109 .ds d- \h'\*(#H'\(pd\h'-\w'~'u'\v'-.25m'\f2\(hy\fP\v'.25m'\h'-\*(#H'
110 .ds D- D\\k:\h'-\w'D'u'\v'-.11m'\z\(hy\v'.11m'\h'|\\n:u'
111 .ds th \*(#[\v'.3m'\s+1I\s-1\v'-.3m'\h'-(\w'I'u*2/3)'\s-1o\s+1\*(#]
112 .ds Th \*(#[\s+2I\s-2\h'-\w'I'u*3/5'\v'-.3m'o\v'.3m'\*(#]
113 .ds ae a\h'-(\w'a'u*4/10)'e
114 .ds Ae A\h'-(\w'A'u*4/10)'E
115 . \" corrections for vroff
116 .if v .ds ~ \\k:\h'-(\\n(.wu*9/10-\*(#H)'\s-2\u~\d\s+2\h'|\\n:u'
117 .if v .ds ^ \\k:\h'-(\\n(.wu*10/11-\*(#H)'\v'-.4m'^\v'.4m'\h'|\\n:u'
118 . \" for low resolution devices (crt and lpr)
119 .if \n(.H>23 .if \n(.V>19 \
120 \{\
121 . ds : e
122 . ds 8 ss
123 . ds o a
124 . ds d- d\h'-1'\(ga
125 . ds D- D\h'-1'\(hy
126 . ds th \o'bp'
127 . ds Th \o'LP'
128 . ds ae ae
129 . ds Ae AE
130 .\}
131 .rm #[ #] #H #V #F C
132 .\" ========================================================================
133 .\"
134 .IX Title "LIBEV 3"
135 .TH LIBEV 3 "2008-10-21" "libev-3.45" "libev - high performance full featured event loop"
136 .\" For nroff, turn off justification. Always turn off hyphenation; it makes
137 .\" way too many mistakes in technical documents.
138 .if n .ad l
139 .nh
140 .SH "NAME"
141 libev \- a high performance full\-featured event loop written in C
142 .SH "SYNOPSIS"
143 .IX Header "SYNOPSIS"
144 .Vb 1
145 \& #include <ev.h>
146 .Ve
147 .Sh "\s-1EXAMPLE\s0 \s-1PROGRAM\s0"
148 .IX Subsection "EXAMPLE PROGRAM"
149 .Vb 2
150 \& // a single header file is required
151 \& #include <ev.h>
152 \&
153 \& // every watcher type has its own typedef\*(Aqd struct
154 \& // with the name ev_<type>
155 \& ev_io stdin_watcher;
156 \& ev_timer timeout_watcher;
157 \&
158 \& // all watcher callbacks have a similar signature
159 \& // this callback is called when data is readable on stdin
160 \& static void
161 \& stdin_cb (EV_P_ struct ev_io *w, int revents)
162 \& {
163 \& puts ("stdin ready");
164 \& // for one\-shot events, one must manually stop the watcher
165 \& // with its corresponding stop function.
166 \& ev_io_stop (EV_A_ w);
167 \&
168 \& // this causes all nested ev_loop\*(Aqs to stop iterating
169 \& ev_unloop (EV_A_ EVUNLOOP_ALL);
170 \& }
171 \&
172 \& // another callback, this time for a time\-out
173 \& static void
174 \& timeout_cb (EV_P_ struct ev_timer *w, int revents)
175 \& {
176 \& puts ("timeout");
177 \& // this causes the innermost ev_loop to stop iterating
178 \& ev_unloop (EV_A_ EVUNLOOP_ONE);
179 \& }
180 \&
181 \& int
182 \& main (void)
183 \& {
184 \& // use the default event loop unless you have special needs
185 \& struct ev_loop *loop = ev_default_loop (0);
186 \&
187 \& // initialise an io watcher, then start it
188 \& // this one will watch for stdin to become readable
189 \& ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
190 \& ev_io_start (loop, &stdin_watcher);
191 \&
192 \& // initialise a timer watcher, then start it
193 \& // simple non\-repeating 5.5 second timeout
194 \& ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
195 \& ev_timer_start (loop, &timeout_watcher);
196 \&
197 \& // now wait for events to arrive
198 \& ev_loop (loop, 0);
199 \&
200 \& // unloop was called, so exit
201 \& return 0;
202 \& }
203 .Ve
204 .SH "DESCRIPTION"
205 .IX Header "DESCRIPTION"
206 The newest version of this document is also available as an html-formatted
207 web page you might find easier to navigate when reading it for the first
208 time: <http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
209 .PP
210 Libev is an event loop: you register interest in certain events (such as a
211 file descriptor being readable or a timeout occurring), and it will manage
212 these event sources and provide your program with events.
213 .PP
214 To do this, it must take more or less complete control over your process
215 (or thread) by executing the \fIevent loop\fR handler, and will then
216 communicate events via a callback mechanism.
217 .PP
218 You register interest in certain events by registering so-called \fIevent
219 watchers\fR, which are relatively small C structures you initialise with the
220 details of the event, and then hand it over to libev by \fIstarting\fR the
221 watcher.
222 .Sh "\s-1FEATURES\s0"
223 .IX Subsection "FEATURES"
224 Libev supports \f(CW\*(C`select\*(C'\fR, \f(CW\*(C`poll\*(C'\fR, the Linux-specific \f(CW\*(C`epoll\*(C'\fR, the
225 BSD-specific \f(CW\*(C`kqueue\*(C'\fR and the Solaris-specific event port mechanisms
226 for file descriptor events (\f(CW\*(C`ev_io\*(C'\fR), the Linux \f(CW\*(C`inotify\*(C'\fR interface
227 (for \f(CW\*(C`ev_stat\*(C'\fR), relative timers (\f(CW\*(C`ev_timer\*(C'\fR), absolute timers
228 with customised rescheduling (\f(CW\*(C`ev_periodic\*(C'\fR), synchronous signals
229 (\f(CW\*(C`ev_signal\*(C'\fR), process status change events (\f(CW\*(C`ev_child\*(C'\fR), and event
230 watchers dealing with the event loop mechanism itself (\f(CW\*(C`ev_idle\*(C'\fR,
231 \&\f(CW\*(C`ev_embed\*(C'\fR, \f(CW\*(C`ev_prepare\*(C'\fR and \f(CW\*(C`ev_check\*(C'\fR watchers) as well as
232 file watchers (\f(CW\*(C`ev_stat\*(C'\fR) and even limited support for fork events
233 (\f(CW\*(C`ev_fork\*(C'\fR).
234 .PP
235 It also is quite fast (see this
236 benchmark comparing it to libevent
237 for example).
238 .Sh "\s-1CONVENTIONS\s0"
239 .IX Subsection "CONVENTIONS"
240 Libev is very configurable. In this manual the default (and most common)
241 configuration will be described, which supports multiple event loops. For
242 more info about various configuration options please have a look at
243 \&\fB\s-1EMBED\s0\fR section in this manual. If libev was configured without support
244 for multiple event loops, then all functions taking an initial argument of
245 name \f(CW\*(C`loop\*(C'\fR (which is always of type \f(CW\*(C`struct ev_loop *\*(C'\fR) will not have
246 this argument.
247 .Sh "\s-1TIME\s0 \s-1REPRESENTATION\s0"
248 .IX Subsection "TIME REPRESENTATION"
249 Libev represents time as a single floating point number, representing the
250 (fractional) number of seconds since the (\s-1POSIX\s0) epoch (somewhere near
251 the beginning of 1970, details are complicated, don't ask). This type is
252 called \f(CW\*(C`ev_tstamp\*(C'\fR, which is what you should use too. It usually aliases
253 to the \f(CW\*(C`double\*(C'\fR type in C, and when you need to do any calculations on
254 it, you should treat it as some floating point value. Unlike the name
255 component \f(CW\*(C`stamp\*(C'\fR might indicate, it is also used for time differences
256 throughout libev.
257 .SH "ERROR HANDLING"
258 .IX Header "ERROR HANDLING"
259 Libev knows three classes of errors: operating system errors, usage errors
260 and internal errors (bugs).
261 .PP
262 When libev catches an operating system error it cannot handle (for example
263 a system call indicating a condition libev cannot fix), it calls the callback
264 set via \f(CW\*(C`ev_set_syserr_cb\*(C'\fR, which is supposed to fix the problem or
265 abort. The default is to print a diagnostic message and to call \f(CW\*(C`abort
266 ()\*(C'\fR.
267 .PP
268 When libev detects a usage error such as a negative timer interval, then
269 it will print a diagnostic message and abort (via the \f(CW\*(C`assert\*(C'\fR mechanism,
270 so \f(CW\*(C`NDEBUG\*(C'\fR will disable this checking): these are programming errors in
271 the libev caller and need to be fixed there.
272 .PP
273 Libev also has a few internal error-checking \f(CW\*(C`assert\*(C'\fRions, and also has
274 extensive consistency checking code. These do not trigger under normal
275 circumstances, as they indicate either a bug in libev or worse.
276 .SH "GLOBAL FUNCTIONS"
277 .IX Header "GLOBAL FUNCTIONS"
278 These functions can be called anytime, even before initialising the
279 library in any way.
280 .IP "ev_tstamp ev_time ()" 4
281 .IX Item "ev_tstamp ev_time ()"
282 Returns the current time as libev would use it. Please note that the
283 \&\f(CW\*(C`ev_now\*(C'\fR function is usually faster and also often returns the timestamp
284 you actually want to know.
285 .IP "ev_sleep (ev_tstamp interval)" 4
286 .IX Item "ev_sleep (ev_tstamp interval)"
287 Sleep for the given interval: The current thread will be blocked until
288 either it is interrupted or the given time interval has passed. Basically
289 this is a sub-second-resolution \f(CW\*(C`sleep ()\*(C'\fR.
290 .IP "int ev_version_major ()" 4
291 .IX Item "int ev_version_major ()"
292 .PD 0
293 .IP "int ev_version_minor ()" 4
294 .IX Item "int ev_version_minor ()"
295 .PD
296 You can find out the major and minor \s-1ABI\s0 version numbers of the library
297 you linked against by calling the functions \f(CW\*(C`ev_version_major\*(C'\fR and
298 \&\f(CW\*(C`ev_version_minor\*(C'\fR. If you want, you can compare against the global
299 symbols \f(CW\*(C`EV_VERSION_MAJOR\*(C'\fR and \f(CW\*(C`EV_VERSION_MINOR\*(C'\fR, which specify the
300 version of the library your program was compiled against.
301 .Sp
302 These version numbers refer to the \s-1ABI\s0 version of the library, not the
303 release version.
304 .Sp
305 Usually, it's a good idea to terminate if the major versions mismatch,
306 as this indicates an incompatible change. Minor versions are usually
307 compatible to older versions, so a larger minor version alone is usually
308 not a problem.
309 .Sp
310 Example: Make sure we haven't accidentally been linked against the wrong
311 version.
312 .Sp
313 .Vb 3
314 \& assert (("libev version mismatch",
315 \& ev_version_major () == EV_VERSION_MAJOR
316 \& && ev_version_minor () >= EV_VERSION_MINOR));
317 .Ve
318 .IP "unsigned int ev_supported_backends ()" 4
319 .IX Item "unsigned int ev_supported_backends ()"
320 Return the set of all backends (i.e. their corresponding \f(CW\*(C`EV_BACKEND_*\*(C'\fR
321 value) compiled into this binary of libev (independent of their
322 availability on the system you are running on). See \f(CW\*(C`ev_default_loop\*(C'\fR for
323 a description of the set values.
324 .Sp
325 Example: make sure we have the epoll method, because yeah this is cool and
326 a must have and can we have a torrent of it please!!!11
327 .Sp
328 .Vb 2
329 \& assert (("sorry, no epoll, no sex",
330 \& ev_supported_backends () & EVBACKEND_EPOLL));
331 .Ve
332 .IP "unsigned int ev_recommended_backends ()" 4
333 .IX Item "unsigned int ev_recommended_backends ()"
334 Return the set of all backends compiled into this binary of libev and also
335 recommended for this platform. This set is often smaller than the one
336 returned by \f(CW\*(C`ev_supported_backends\*(C'\fR, as for example kqueue is broken on
337 most BSDs and will not be auto-detected unless you explicitly request it
338 (assuming you know what you are doing). This is the set of backends that
339 libev will probe for if you specify no backends explicitly.
340 .IP "unsigned int ev_embeddable_backends ()" 4
341 .IX Item "unsigned int ev_embeddable_backends ()"
342 Returns the set of backends that are embeddable in other event loops. This
343 is the theoretical, all-platform, value. To find which backends
344 might be supported on the current system, you would need to look at
345 \&\f(CW\*(C`ev_embeddable_backends () & ev_supported_backends ()\*(C'\fR, likewise for
346 recommended ones.
347 .Sp
348 See the description of \f(CW\*(C`ev_embed\*(C'\fR watchers for more info.
349 .IP "ev_set_allocator (void *(*cb)(void *ptr, long size)) [\s-1NOT\s0 \s-1REENTRANT\s0]" 4
350 .IX Item "ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]"
351 Sets the allocation function to use (the prototype is similar \- the
352 semantics are identical to the \f(CW\*(C`realloc\*(C'\fR C89/SuS/POSIX function). It is
353 used to allocate and free memory (no surprises here). If it returns zero
354 when memory needs to be allocated (\f(CW\*(C`size != 0\*(C'\fR), the library might abort
355 or take some potentially destructive action.
356 .Sp
357 Since some systems (at least OpenBSD and Darwin) fail to implement
358 correct \f(CW\*(C`realloc\*(C'\fR semantics, libev will use a wrapper around the system
359 \&\f(CW\*(C`realloc\*(C'\fR and \f(CW\*(C`free\*(C'\fR functions by default.
360 .Sp
361 You could override this function in high-availability programs to, say,
362 free some memory if it cannot allocate memory, to use a special allocator,
363 or even to sleep a while and retry until some memory is available.
364 .Sp
365 Example: Replace the libev allocator with one that waits a bit and then
366 retries (example requires a standards-compliant \f(CW\*(C`realloc\*(C'\fR).
367 .Sp
368 .Vb 6
369 \& static void *
370 \& persistent_realloc (void *ptr, size_t size)
371 \& {
372 \& for (;;)
373 \& {
374 \& void *newptr = realloc (ptr, size);
375 \&
376 \& if (newptr)
377 \& return newptr;
378 \&
379 \& sleep (60);
380 \& }
381 \& }
382 \&
383 \& ...
384 \& ev_set_allocator (persistent_realloc);
385 .Ve
386 .IP "ev_set_syserr_cb (void (*cb)(const char *msg)); [\s-1NOT\s0 \s-1REENTRANT\s0]" 4
387 .IX Item "ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]"
388 Set the callback function to call on a retryable system call error (such
389 as failed select, poll, epoll_wait). The message is a printable string
390 indicating the system call or subsystem causing the problem. If this
391 callback is set, then libev will expect it to remedy the situation, no
392 matter what, when it returns. That is, libev will generally retry the
393 requested operation, or, if the condition doesn't go away, do bad stuff
394 (such as abort).
395 .Sp
396 Example: This is basically the same thing that libev does internally, too.
397 .Sp
398 .Vb 6
399 \& static void
400 \& fatal_error (const char *msg)
401 \& {
402 \& perror (msg);
403 \& abort ();
404 \& }
405 \&
406 \& ...
407 \& ev_set_syserr_cb (fatal_error);
408 .Ve
409 .SH "FUNCTIONS CONTROLLING THE EVENT LOOP"
410 .IX Header "FUNCTIONS CONTROLLING THE EVENT LOOP"
411 An event loop is described by a \f(CW\*(C`struct ev_loop *\*(C'\fR. The library knows two
412 types of such loops, the \fIdefault\fR loop, which supports signals and child
413 events, and dynamically created loops which do not.
414 .IP "struct ev_loop *ev_default_loop (unsigned int flags)" 4
415 .IX Item "struct ev_loop *ev_default_loop (unsigned int flags)"
416 This will initialise the default event loop if it hasn't been initialised
417 yet and return it. If the default loop could not be initialised, returns
418 false. If it already was initialised it simply returns it (and ignores the
419 flags. If that is troubling you, check \f(CW\*(C`ev_backend ()\*(C'\fR afterwards).
420 .Sp
421 If you don't know what event loop to use, use the one returned from this
422 function.
423 .Sp
424 Note that this function is \fInot\fR thread-safe, so if you want to use it
425 from multiple threads, you have to lock (note also that this is unlikely,
426 as loops cannot bes hared easily between threads anyway).
427 .Sp
428 The default loop is the only loop that can handle \f(CW\*(C`ev_signal\*(C'\fR and
429 \&\f(CW\*(C`ev_child\*(C'\fR watchers, and to do this, it always registers a handler
430 for \f(CW\*(C`SIGCHLD\*(C'\fR. If this is a problem for your application you can either
431 create a dynamic loop with \f(CW\*(C`ev_loop_new\*(C'\fR that doesn't do that, or you
432 can simply overwrite the \f(CW\*(C`SIGCHLD\*(C'\fR signal handler \fIafter\fR calling
433 \&\f(CW\*(C`ev_default_init\*(C'\fR.
434 .Sp
435 The flags argument can be used to specify special behaviour or specific
436 backends to use, and is usually specified as \f(CW0\fR (or \f(CW\*(C`EVFLAG_AUTO\*(C'\fR).
437 .Sp
438 The following flags are supported:
439 .RS 4
440 .ie n .IP """EVFLAG_AUTO""" 4
441 .el .IP "\f(CWEVFLAG_AUTO\fR" 4
442 .IX Item "EVFLAG_AUTO"
443 The default flags value. Use this if you have no clue (it's the right
444 thing, believe me).
445 .ie n .IP """EVFLAG_NOENV""" 4
446 .el .IP "\f(CWEVFLAG_NOENV\fR" 4
447 .IX Item "EVFLAG_NOENV"
448 If this flag bit is or'ed into the flag value (or the program runs setuid
449 or setgid) then libev will \fInot\fR look at the environment variable
450 \&\f(CW\*(C`LIBEV_FLAGS\*(C'\fR. Otherwise (the default), this environment variable will
451 override the flags completely if it is found in the environment. This is
452 useful to try out specific backends to test their performance, or to work
453 around bugs.
454 .ie n .IP """EVFLAG_FORKCHECK""" 4
455 .el .IP "\f(CWEVFLAG_FORKCHECK\fR" 4
456 .IX Item "EVFLAG_FORKCHECK"
457 Instead of calling \f(CW\*(C`ev_default_fork\*(C'\fR or \f(CW\*(C`ev_loop_fork\*(C'\fR manually after
458 a fork, you can also make libev check for a fork in each iteration by
459 enabling this flag.
460 .Sp
461 This works by calling \f(CW\*(C`getpid ()\*(C'\fR on every iteration of the loop,
462 and thus this might slow down your event loop if you do a lot of loop
463 iterations and little real work, but is usually not noticeable (on my
464 GNU/Linux system for example, \f(CW\*(C`getpid\*(C'\fR is actually a simple 5\-insn sequence
465 without a system call and thus \fIvery\fR fast, but my GNU/Linux system also has
466 \&\f(CW\*(C`pthread_atfork\*(C'\fR which is even faster).
467 .Sp
468 The big advantage of this flag is that you can forget about fork (and
469 forget about forgetting to tell libev about forking) when you use this
470 flag.
471 .Sp
472 This flag setting cannot be overridden or specified in the \f(CW\*(C`LIBEV_FLAGS\*(C'\fR
473 environment variable.
474 .ie n .IP """EVBACKEND_SELECT"" (value 1, portable select backend)" 4
475 .el .IP "\f(CWEVBACKEND_SELECT\fR (value 1, portable select backend)" 4
476 .IX Item "EVBACKEND_SELECT (value 1, portable select backend)"
477 This is your standard \fIselect\fR\|(2) backend. Not \fIcompletely\fR standard, as
478 libev tries to roll its own fd_set with no limits on the number of fds,
479 but if that fails, expect a fairly low limit on the number of fds when
480 using this backend. It doesn't scale too well (O(highest_fd)), but its
481 usually the fastest backend for a low number of (low-numbered :) fds.
482 .Sp
483 To get good performance out of this backend you need a high amount of
484 parallelism (most of the file descriptors should be busy). If you are
485 writing a server, you should \f(CW\*(C`accept ()\*(C'\fR in a loop to accept as many
486 connections as possible during one iteration. You might also want to have
487 a look at \f(CW\*(C`ev_set_io_collect_interval ()\*(C'\fR to increase the amount of
488 readiness notifications you get per iteration.
489 .Sp
490 This backend maps \f(CW\*(C`EV_READ\*(C'\fR to the \f(CW\*(C`readfds\*(C'\fR set and \f(CW\*(C`EV_WRITE\*(C'\fR to the
491 \&\f(CW\*(C`writefds\*(C'\fR set (and to work around Microsoft Windows bugs, also onto the
492 \&\f(CW\*(C`exceptfds\*(C'\fR set on that platform).
493 .ie n .IP """EVBACKEND_POLL"" (value 2, poll backend, available everywhere except on windows)" 4
494 .el .IP "\f(CWEVBACKEND_POLL\fR (value 2, poll backend, available everywhere except on windows)" 4
495 .IX Item "EVBACKEND_POLL (value 2, poll backend, available everywhere except on windows)"
496 And this is your standard \fIpoll\fR\|(2) backend. It's more complicated
497 than select, but handles sparse fds better and has no artificial
498 limit on the number of fds you can use (except it will slow down
499 considerably with a lot of inactive fds). It scales similarly to select,
500 i.e. O(total_fds). See the entry for \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR, above, for
501 performance tips.
502 .Sp
503 This backend maps \f(CW\*(C`EV_READ\*(C'\fR to \f(CW\*(C`POLLIN | POLLERR | POLLHUP\*(C'\fR, and
504 \&\f(CW\*(C`EV_WRITE\*(C'\fR to \f(CW\*(C`POLLOUT | POLLERR | POLLHUP\*(C'\fR.
505 .ie n .IP """EVBACKEND_EPOLL"" (value 4, Linux)" 4
506 .el .IP "\f(CWEVBACKEND_EPOLL\fR (value 4, Linux)" 4
507 .IX Item "EVBACKEND_EPOLL (value 4, Linux)"
508 For few fds, this backend is a bit little slower than poll and select,
509 but it scales phenomenally better. While poll and select usually scale
510 like O(total_fds) where n is the total number of fds (or the highest fd),
511 epoll scales either O(1) or O(active_fds). The epoll design has a number
512 of shortcomings, such as silently dropping events in some hard-to-detect
513 cases and requiring a system call per fd change, no fork support and bad
514 support for dup.
515 .Sp
516 While stopping, setting and starting an I/O watcher in the same iteration
517 will result in some caching, there is still a system call per such incident
518 (because the fd could point to a different file description now), so its
519 best to avoid that. Also, \f(CW\*(C`dup ()\*(C'\fR'ed file descriptors might not work
520 very well if you register events for both fds.
521 .Sp
522 Please note that epoll sometimes generates spurious notifications, so you
523 need to use non-blocking I/O or other means to avoid blocking when no data
524 (or space) is available.
525 .Sp
526 Best performance from this backend is achieved by not unregistering all
527 watchers for a file descriptor until it has been closed, if possible,
528 i.e. keep at least one watcher active per fd at all times. Stopping and
529 starting a watcher (without re-setting it) also usually doesn't cause
530 extra overhead.
531 .Sp
532 While nominally embeddable in other event loops, this feature is broken in
533 all kernel versions tested so far.
534 .Sp
535 This backend maps \f(CW\*(C`EV_READ\*(C'\fR and \f(CW\*(C`EV_WRITE\*(C'\fR in the same way as
536 \&\f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
537 .ie n .IP """EVBACKEND_KQUEUE"" (value 8, most \s-1BSD\s0 clones)" 4
538 .el .IP "\f(CWEVBACKEND_KQUEUE\fR (value 8, most \s-1BSD\s0 clones)" 4
539 .IX Item "EVBACKEND_KQUEUE (value 8, most BSD clones)"
540 Kqueue deserves special mention, as at the time of this writing, it was
541 broken on all BSDs except NetBSD (usually it doesn't work reliably with
542 anything but sockets and pipes, except on Darwin, where of course it's
543 completely useless). For this reason it's not being \*(L"auto-detected\*(R" unless
544 you explicitly specify it in the flags (i.e. using \f(CW\*(C`EVBACKEND_KQUEUE\*(C'\fR) or
545 libev was compiled on a known-to-be-good (\-enough) system like NetBSD.
546 .Sp
547 You still can embed kqueue into a normal poll or select backend and use it
548 only for sockets (after having made sure that sockets work with kqueue on
549 the target platform). See \f(CW\*(C`ev_embed\*(C'\fR watchers for more info.
550 .Sp
551 It scales in the same way as the epoll backend, but the interface to the
552 kernel is more efficient (which says nothing about its actual speed, of
553 course). While stopping, setting and starting an I/O watcher does never
554 cause an extra system call as with \f(CW\*(C`EVBACKEND_EPOLL\*(C'\fR, it still adds up to
555 two event changes per incident. Support for \f(CW\*(C`fork ()\*(C'\fR is very bad and it
556 drops fds silently in similarly hard-to-detect cases.
557 .Sp
558 This backend usually performs well under most conditions.
559 .Sp
560 While nominally embeddable in other event loops, this doesn't work
561 everywhere, so you might need to test for this. And since it is broken
562 almost everywhere, you should only use it when you have a lot of sockets
563 (for which it usually works), by embedding it into another event loop
564 (e.g. \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or \f(CW\*(C`EVBACKEND_POLL\*(C'\fR) and, did I mention it,
565 using it only for sockets.
566 .Sp
567 This backend maps \f(CW\*(C`EV_READ\*(C'\fR into an \f(CW\*(C`EVFILT_READ\*(C'\fR kevent with
568 \&\f(CW\*(C`NOTE_EOF\*(C'\fR, and \f(CW\*(C`EV_WRITE\*(C'\fR into an \f(CW\*(C`EVFILT_WRITE\*(C'\fR kevent with
569 \&\f(CW\*(C`NOTE_EOF\*(C'\fR.
570 .ie n .IP """EVBACKEND_DEVPOLL"" (value 16, Solaris 8)" 4
571 .el .IP "\f(CWEVBACKEND_DEVPOLL\fR (value 16, Solaris 8)" 4
572 .IX Item "EVBACKEND_DEVPOLL (value 16, Solaris 8)"
573 This is not implemented yet (and might never be, unless you send me an
574 implementation). According to reports, \f(CW\*(C`/dev/poll\*(C'\fR only supports sockets
575 and is not embeddable, which would limit the usefulness of this backend
576 immensely.
577 .ie n .IP """EVBACKEND_PORT"" (value 32, Solaris 10)" 4
578 .el .IP "\f(CWEVBACKEND_PORT\fR (value 32, Solaris 10)" 4
579 .IX Item "EVBACKEND_PORT (value 32, Solaris 10)"
580 This uses the Solaris 10 event port mechanism. As with everything on Solaris,
581 it's really slow, but it still scales very well (O(active_fds)).
582 .Sp
583 Please note that Solaris event ports can deliver a lot of spurious
584 notifications, so you need to use non-blocking I/O or other means to avoid
585 blocking when no data (or space) is available.
586 .Sp
587 While this backend scales well, it requires one system call per active
588 file descriptor per loop iteration. For small and medium numbers of file
589 descriptors a \*(L"slow\*(R" \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or \f(CW\*(C`EVBACKEND_POLL\*(C'\fR backend
590 might perform better.
591 .Sp
592 On the positive side, with the exception of the spurious readiness
593 notifications, this backend actually performed fully to specification
594 in all tests and is fully embeddable, which is a rare feat among the
595 OS-specific backends.
596 .Sp
597 This backend maps \f(CW\*(C`EV_READ\*(C'\fR and \f(CW\*(C`EV_WRITE\*(C'\fR in the same way as
598 \&\f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
599 .ie n .IP """EVBACKEND_ALL""" 4
600 .el .IP "\f(CWEVBACKEND_ALL\fR" 4
601 .IX Item "EVBACKEND_ALL"
602 Try all backends (even potentially broken ones that wouldn't be tried
603 with \f(CW\*(C`EVFLAG_AUTO\*(C'\fR). Since this is a mask, you can do stuff such as
604 \&\f(CW\*(C`EVBACKEND_ALL & ~EVBACKEND_KQUEUE\*(C'\fR.
605 .Sp
606 It is definitely not recommended to use this flag.
607 .RE
608 .RS 4
609 .Sp
610 If one or more of these are or'ed into the flags value, then only these
611 backends will be tried (in the reverse order as listed here). If none are
612 specified, all backends in \f(CW\*(C`ev_recommended_backends ()\*(C'\fR will be tried.
613 .Sp
614 Example: This is the most typical usage.
615 .Sp
616 .Vb 2
617 \& if (!ev_default_loop (0))
618 \& fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
619 .Ve
620 .Sp
621 Example: Restrict libev to the select and poll backends, and do not allow
622 environment settings to be taken into account:
623 .Sp
624 .Vb 1
625 \& ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
626 .Ve
627 .Sp
628 Example: Use whatever libev has to offer, but make sure that kqueue is
629 used if available (warning, breaks stuff, best use only with your own
630 private event loop and only if you know the \s-1OS\s0 supports your types of
631 fds):
632 .Sp
633 .Vb 1
634 \& ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
635 .Ve
636 .RE
637 .IP "struct ev_loop *ev_loop_new (unsigned int flags)" 4
638 .IX Item "struct ev_loop *ev_loop_new (unsigned int flags)"
639 Similar to \f(CW\*(C`ev_default_loop\*(C'\fR, but always creates a new event loop that is
640 always distinct from the default loop. Unlike the default loop, it cannot
641 handle signal and child watchers, and attempts to do so will be greeted by
642 undefined behaviour (or a failed assertion if assertions are enabled).
643 .Sp
644 Note that this function \fIis\fR thread-safe, and the recommended way to use
645 libev with threads is indeed to create one loop per thread, and using the
646 default loop in the \*(L"main\*(R" or \*(L"initial\*(R" thread.
647 .Sp
648 Example: Try to create a event loop that uses epoll and nothing else.
649 .Sp
650 .Vb 3
651 \& struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
652 \& if (!epoller)
653 \& fatal ("no epoll found here, maybe it hides under your chair");
654 .Ve
655 .IP "ev_default_destroy ()" 4
656 .IX Item "ev_default_destroy ()"
657 Destroys the default loop again (frees all memory and kernel state
658 etc.). None of the active event watchers will be stopped in the normal
659 sense, so e.g. \f(CW\*(C`ev_is_active\*(C'\fR might still return true. It is your
660 responsibility to either stop all watchers cleanly yourself \fIbefore\fR
661 calling this function, or cope with the fact afterwards (which is usually
662 the easiest thing, you can just ignore the watchers and/or \f(CW\*(C`free ()\*(C'\fR them
663 for example).
664 .Sp
665 Note that certain global state, such as signal state, will not be freed by
666 this function, and related watchers (such as signal and child watchers)
667 would need to be stopped manually.
668 .Sp
669 In general it is not advisable to call this function except in the
670 rare occasion where you really need to free e.g. the signal handling
671 pipe fds. If you need dynamically allocated loops it is better to use
672 \&\f(CW\*(C`ev_loop_new\*(C'\fR and \f(CW\*(C`ev_loop_destroy\*(C'\fR).
673 .IP "ev_loop_destroy (loop)" 4
674 .IX Item "ev_loop_destroy (loop)"
675 Like \f(CW\*(C`ev_default_destroy\*(C'\fR, but destroys an event loop created by an
676 earlier call to \f(CW\*(C`ev_loop_new\*(C'\fR.
677 .IP "ev_default_fork ()" 4
678 .IX Item "ev_default_fork ()"
679 This function sets a flag that causes subsequent \f(CW\*(C`ev_loop\*(C'\fR iterations
680 to reinitialise the kernel state for backends that have one. Despite the
681 name, you can call it anytime, but it makes most sense after forking, in
682 the child process (or both child and parent, but that again makes little
683 sense). You \fImust\fR call it in the child before using any of the libev
684 functions, and it will only take effect at the next \f(CW\*(C`ev_loop\*(C'\fR iteration.
685 .Sp
686 On the other hand, you only need to call this function in the child
687 process if and only if you want to use the event library in the child. If
688 you just fork+exec, you don't have to call it at all.
689 .Sp
690 The function itself is quite fast and it's usually not a problem to call
691 it just in case after a fork. To make this easy, the function will fit in
692 quite nicely into a call to \f(CW\*(C`pthread_atfork\*(C'\fR:
693 .Sp
694 .Vb 1
695 \& pthread_atfork (0, 0, ev_default_fork);
696 .Ve
697 .IP "ev_loop_fork (loop)" 4
698 .IX Item "ev_loop_fork (loop)"
699 Like \f(CW\*(C`ev_default_fork\*(C'\fR, but acts on an event loop created by
700 \&\f(CW\*(C`ev_loop_new\*(C'\fR. Yes, you have to call this on every allocated event loop
701 after fork that you want to re-use in the child, and how you do this is
702 entirely your own problem.
703 .IP "int ev_is_default_loop (loop)" 4
704 .IX Item "int ev_is_default_loop (loop)"
705 Returns true when the given loop is, in fact, the default loop, and false
706 otherwise.
707 .IP "unsigned int ev_loop_count (loop)" 4
708 .IX Item "unsigned int ev_loop_count (loop)"
709 Returns the count of loop iterations for the loop, which is identical to
710 the number of times libev did poll for new events. It starts at \f(CW0\fR and
711 happily wraps around with enough iterations.
712 .Sp
713 This value can sometimes be useful as a generation counter of sorts (it
714 \&\*(L"ticks\*(R" the number of loop iterations), as it roughly corresponds with
715 \&\f(CW\*(C`ev_prepare\*(C'\fR and \f(CW\*(C`ev_check\*(C'\fR calls.
716 .IP "unsigned int ev_backend (loop)" 4
717 .IX Item "unsigned int ev_backend (loop)"
718 Returns one of the \f(CW\*(C`EVBACKEND_*\*(C'\fR flags indicating the event backend in
719 use.
720 .IP "ev_tstamp ev_now (loop)" 4
721 .IX Item "ev_tstamp ev_now (loop)"
722 Returns the current \*(L"event loop time\*(R", which is the time the event loop
723 received events and started processing them. This timestamp does not
724 change as long as callbacks are being processed, and this is also the base
725 time used for relative timers. You can treat it as the timestamp of the
726 event occurring (or more correctly, libev finding out about it).
727 .IP "ev_now_update (loop)" 4
728 .IX Item "ev_now_update (loop)"
729 Establishes the current time by querying the kernel, updating the time
730 returned by \f(CW\*(C`ev_now ()\*(C'\fR in the progress. This is a costly operation and
731 is usually done automatically within \f(CW\*(C`ev_loop ()\*(C'\fR.
732 .Sp
733 This function is rarely useful, but when some event callback runs for a
734 very long time without entering the event loop, updating libev's idea of
735 the current time is a good idea.
736 .Sp
737 See also \*(L"The special problem of time updates\*(R" in the \f(CW\*(C`ev_timer\*(C'\fR section.
738 .IP "ev_loop (loop, int flags)" 4
739 .IX Item "ev_loop (loop, int flags)"
740 Finally, this is it, the event handler. This function usually is called
741 after you initialised all your watchers and you want to start handling
742 events.
743 .Sp
744 If the flags argument is specified as \f(CW0\fR, it will not return until
745 either no event watchers are active anymore or \f(CW\*(C`ev_unloop\*(C'\fR was called.
746 .Sp
747 Please note that an explicit \f(CW\*(C`ev_unloop\*(C'\fR is usually better than
748 relying on all watchers to be stopped when deciding when a program has
749 finished (especially in interactive programs), but having a program
750 that automatically loops as long as it has to and no longer by virtue
751 of relying on its watchers stopping correctly, that is truly a thing of
752 beauty.
753 .Sp
754 A flags value of \f(CW\*(C`EVLOOP_NONBLOCK\*(C'\fR will look for new events, will handle
755 those events and any already outstanding ones, but will not block your
756 process in case there are no events and will return after one iteration of
757 the loop.
758 .Sp
759 A flags value of \f(CW\*(C`EVLOOP_ONESHOT\*(C'\fR will look for new events (waiting if
760 necessary) and will handle those and any already outstanding ones. It
761 will block your process until at least one new event arrives (which could
762 be an event internal to libev itself, so there is no guarentee that a
763 user-registered callback will be called), and will return after one
764 iteration of the loop.
765 .Sp
766 This is useful if you are waiting for some external event in conjunction
767 with something not expressible using other libev watchers (i.e. "roll your
768 own \f(CW\*(C`ev_loop\*(C'\fR"). However, a pair of \f(CW\*(C`ev_prepare\*(C'\fR/\f(CW\*(C`ev_check\*(C'\fR watchers is
769 usually a better approach for this kind of thing.
770 .Sp
771 Here are the gory details of what \f(CW\*(C`ev_loop\*(C'\fR does:
772 .Sp
773 .Vb 10
774 \& \- Before the first iteration, call any pending watchers.
775 \& * If EVFLAG_FORKCHECK was used, check for a fork.
776 \& \- If a fork was detected (by any means), queue and call all fork watchers.
777 \& \- Queue and call all prepare watchers.
778 \& \- If we have been forked, detach and recreate the kernel state
779 \& as to not disturb the other process.
780 \& \- Update the kernel state with all outstanding changes.
781 \& \- Update the "event loop time" (ev_now ()).
782 \& \- Calculate for how long to sleep or block, if at all
783 \& (active idle watchers, EVLOOP_NONBLOCK or not having
784 \& any active watchers at all will result in not sleeping).
785 \& \- Sleep if the I/O and timer collect interval say so.
786 \& \- Block the process, waiting for any events.
787 \& \- Queue all outstanding I/O (fd) events.
788 \& \- Update the "event loop time" (ev_now ()), and do time jump adjustments.
789 \& \- Queue all expired timers.
790 \& \- Queue all expired periodics.
791 \& \- Unless any events are pending now, queue all idle watchers.
792 \& \- Queue all check watchers.
793 \& \- Call all queued watchers in reverse order (i.e. check watchers first).
794 \& Signals and child watchers are implemented as I/O watchers, and will
795 \& be handled here by queueing them when their watcher gets executed.
796 \& \- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
797 \& were used, or there are no active watchers, return, otherwise
798 \& continue with step *.
799 .Ve
800 .Sp
801 Example: Queue some jobs and then loop until no events are outstanding
802 anymore.
803 .Sp
804 .Vb 4
805 \& ... queue jobs here, make sure they register event watchers as long
806 \& ... as they still have work to do (even an idle watcher will do..)
807 \& ev_loop (my_loop, 0);
808 \& ... jobs done or somebody called unloop. yeah!
809 .Ve
810 .IP "ev_unloop (loop, how)" 4
811 .IX Item "ev_unloop (loop, how)"
812 Can be used to make a call to \f(CW\*(C`ev_loop\*(C'\fR return early (but only after it
813 has processed all outstanding events). The \f(CW\*(C`how\*(C'\fR argument must be either
814 \&\f(CW\*(C`EVUNLOOP_ONE\*(C'\fR, which will make the innermost \f(CW\*(C`ev_loop\*(C'\fR call return, or
815 \&\f(CW\*(C`EVUNLOOP_ALL\*(C'\fR, which will make all nested \f(CW\*(C`ev_loop\*(C'\fR calls return.
816 .Sp
817 This \*(L"unloop state\*(R" will be cleared when entering \f(CW\*(C`ev_loop\*(C'\fR again.
818 .Sp
819 It is safe to call \f(CW\*(C`ev_unloop\*(C'\fR from otuside any \f(CW\*(C`ev_loop\*(C'\fR calls.
820 .IP "ev_ref (loop)" 4
821 .IX Item "ev_ref (loop)"
822 .PD 0
823 .IP "ev_unref (loop)" 4
824 .IX Item "ev_unref (loop)"
825 .PD
826 Ref/unref can be used to add or remove a reference count on the event
827 loop: Every watcher keeps one reference, and as long as the reference
828 count is nonzero, \f(CW\*(C`ev_loop\*(C'\fR will not return on its own.
829 .Sp
830 If you have a watcher you never unregister that should not keep \f(CW\*(C`ev_loop\*(C'\fR
831 from returning, call \fIev_unref()\fR after starting, and \fIev_ref()\fR before
832 stopping it.
833 .Sp
834 As an example, libev itself uses this for its internal signal pipe: It is
835 not visible to the libev user and should not keep \f(CW\*(C`ev_loop\*(C'\fR from exiting
836 if no event watchers registered by it are active. It is also an excellent
837 way to do this for generic recurring timers or from within third-party
838 libraries. Just remember to \fIunref after start\fR and \fIref before stop\fR
839 (but only if the watcher wasn't active before, or was active before,
840 respectively).
841 .Sp
842 Example: Create a signal watcher, but keep it from keeping \f(CW\*(C`ev_loop\*(C'\fR
843 running when nothing else is active.
844 .Sp
845 .Vb 4
846 \& struct ev_signal exitsig;
847 \& ev_signal_init (&exitsig, sig_cb, SIGINT);
848 \& ev_signal_start (loop, &exitsig);
849 \& evf_unref (loop);
850 .Ve
851 .Sp
852 Example: For some weird reason, unregister the above signal handler again.
853 .Sp
854 .Vb 2
855 \& ev_ref (loop);
856 \& ev_signal_stop (loop, &exitsig);
857 .Ve
858 .IP "ev_set_io_collect_interval (loop, ev_tstamp interval)" 4
859 .IX Item "ev_set_io_collect_interval (loop, ev_tstamp interval)"
860 .PD 0
861 .IP "ev_set_timeout_collect_interval (loop, ev_tstamp interval)" 4
862 .IX Item "ev_set_timeout_collect_interval (loop, ev_tstamp interval)"
863 .PD
864 These advanced functions influence the time that libev will spend waiting
865 for events. Both time intervals are by default \f(CW0\fR, meaning that libev
866 will try to invoke timer/periodic callbacks and I/O callbacks with minimum
867 latency.
868 .Sp
869 Setting these to a higher value (the \f(CW\*(C`interval\*(C'\fR \fImust\fR be >= \f(CW0\fR)
870 allows libev to delay invocation of I/O and timer/periodic callbacks
871 to increase efficiency of loop iterations (or to increase power-saving
872 opportunities).
873 .Sp
874 The idea is that sometimes your program runs just fast enough to handle
875 one (or very few) event(s) per loop iteration. While this makes the
876 program responsive, it also wastes a lot of \s-1CPU\s0 time to poll for new
877 events, especially with backends like \f(CW\*(C`select ()\*(C'\fR which have a high
878 overhead for the actual polling but can deliver many events at once.
879 .Sp
880 By setting a higher \fIio collect interval\fR you allow libev to spend more
881 time collecting I/O events, so you can handle more events per iteration,
882 at the cost of increasing latency. Timeouts (both \f(CW\*(C`ev_periodic\*(C'\fR and
883 \&\f(CW\*(C`ev_timer\*(C'\fR) will be not affected. Setting this to a non-null value will
884 introduce an additional \f(CW\*(C`ev_sleep ()\*(C'\fR call into most loop iterations.
885 .Sp
886 Likewise, by setting a higher \fItimeout collect interval\fR you allow libev
887 to spend more time collecting timeouts, at the expense of increased
888 latency/jitter/inexactness (the watcher callback will be called
889 later). \f(CW\*(C`ev_io\*(C'\fR watchers will not be affected. Setting this to a non-null
890 value will not introduce any overhead in libev.
891 .Sp
892 Many (busy) programs can usually benefit by setting the I/O collect
893 interval to a value near \f(CW0.1\fR or so, which is often enough for
894 interactive servers (of course not for games), likewise for timeouts. It
895 usually doesn't make much sense to set it to a lower value than \f(CW0.01\fR,
896 as this approaches the timing granularity of most systems.
897 .Sp
898 Setting the \fItimeout collect interval\fR can improve the opportunity for
899 saving power, as the program will \*(L"bundle\*(R" timer callback invocations that
900 are \*(L"near\*(R" in time together, by delaying some, thus reducing the number of
901 times the process sleeps and wakes up again. Another useful technique to
902 reduce iterations/wake\-ups is to use \f(CW\*(C`ev_periodic\*(C'\fR watchers and make sure
903 they fire on, say, one-second boundaries only.
904 .IP "ev_loop_verify (loop)" 4
905 .IX Item "ev_loop_verify (loop)"
906 This function only does something when \f(CW\*(C`EV_VERIFY\*(C'\fR support has been
907 compiled in. which is the default for non-minimal builds. It tries to go
908 through all internal structures and checks them for validity. If anything
909 is found to be inconsistent, it will print an error message to standard
910 error and call \f(CW\*(C`abort ()\*(C'\fR.
911 .Sp
912 This can be used to catch bugs inside libev itself: under normal
913 circumstances, this function will never abort as of course libev keeps its
914 data structures consistent.
915 .SH "ANATOMY OF A WATCHER"
916 .IX Header "ANATOMY OF A WATCHER"
917 A watcher is a structure that you create and register to record your
918 interest in some event. For instance, if you want to wait for \s-1STDIN\s0 to
919 become readable, you would create an \f(CW\*(C`ev_io\*(C'\fR watcher for that:
920 .PP
921 .Vb 5
922 \& static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)
923 \& {
924 \& ev_io_stop (w);
925 \& ev_unloop (loop, EVUNLOOP_ALL);
926 \& }
927 \&
928 \& struct ev_loop *loop = ev_default_loop (0);
929 \& struct ev_io stdin_watcher;
930 \& ev_init (&stdin_watcher, my_cb);
931 \& ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
932 \& ev_io_start (loop, &stdin_watcher);
933 \& ev_loop (loop, 0);
934 .Ve
935 .PP
936 As you can see, you are responsible for allocating the memory for your
937 watcher structures (and it is usually a bad idea to do this on the stack,
938 although this can sometimes be quite valid).
939 .PP
940 Each watcher structure must be initialised by a call to \f(CW\*(C`ev_init
941 (watcher *, callback)\*(C'\fR, which expects a callback to be provided. This
942 callback gets invoked each time the event occurs (or, in the case of I/O
943 watchers, each time the event loop detects that the file descriptor given
944 is readable and/or writable).
945 .PP
946 Each watcher type has its own \f(CW\*(C`ev_<type>_set (watcher *, ...)\*(C'\fR macro
947 with arguments specific to this watcher type. There is also a macro
948 to combine initialisation and setting in one call: \f(CW\*(C`ev_<type>_init
949 (watcher *, callback, ...)\*(C'\fR.
950 .PP
951 To make the watcher actually watch out for events, you have to start it
952 with a watcher-specific start function (\f(CW\*(C`ev_<type>_start (loop, watcher
953 *)\*(C'\fR), and you can stop watching for events at any time by calling the
954 corresponding stop function (\f(CW\*(C`ev_<type>_stop (loop, watcher *)\*(C'\fR.
955 .PP
956 As long as your watcher is active (has been started but not stopped) you
957 must not touch the values stored in it. Most specifically you must never
958 reinitialise it or call its \f(CW\*(C`set\*(C'\fR macro.
959 .PP
960 Each and every callback receives the event loop pointer as first, the
961 registered watcher structure as second, and a bitset of received events as
962 third argument.
963 .PP
964 The received events usually include a single bit per event type received
965 (you can receive multiple events at the same time). The possible bit masks
966 are:
967 .ie n .IP """EV_READ""" 4
968 .el .IP "\f(CWEV_READ\fR" 4
969 .IX Item "EV_READ"
970 .PD 0
971 .ie n .IP """EV_WRITE""" 4
972 .el .IP "\f(CWEV_WRITE\fR" 4
973 .IX Item "EV_WRITE"
974 .PD
975 The file descriptor in the \f(CW\*(C`ev_io\*(C'\fR watcher has become readable and/or
976 writable.
977 .ie n .IP """EV_TIMEOUT""" 4
978 .el .IP "\f(CWEV_TIMEOUT\fR" 4
979 .IX Item "EV_TIMEOUT"
980 The \f(CW\*(C`ev_timer\*(C'\fR watcher has timed out.
981 .ie n .IP """EV_PERIODIC""" 4
982 .el .IP "\f(CWEV_PERIODIC\fR" 4
983 .IX Item "EV_PERIODIC"
984 The \f(CW\*(C`ev_periodic\*(C'\fR watcher has timed out.
985 .ie n .IP """EV_SIGNAL""" 4
986 .el .IP "\f(CWEV_SIGNAL\fR" 4
987 .IX Item "EV_SIGNAL"
988 The signal specified in the \f(CW\*(C`ev_signal\*(C'\fR watcher has been received by a thread.
989 .ie n .IP """EV_CHILD""" 4
990 .el .IP "\f(CWEV_CHILD\fR" 4
991 .IX Item "EV_CHILD"
992 The pid specified in the \f(CW\*(C`ev_child\*(C'\fR watcher has received a status change.
993 .ie n .IP """EV_STAT""" 4
994 .el .IP "\f(CWEV_STAT\fR" 4
995 .IX Item "EV_STAT"
996 The path specified in the \f(CW\*(C`ev_stat\*(C'\fR watcher changed its attributes somehow.
997 .ie n .IP """EV_IDLE""" 4
998 .el .IP "\f(CWEV_IDLE\fR" 4
999 .IX Item "EV_IDLE"
1000 The \f(CW\*(C`ev_idle\*(C'\fR watcher has determined that you have nothing better to do.
1001 .ie n .IP """EV_PREPARE""" 4
1002 .el .IP "\f(CWEV_PREPARE\fR" 4
1003 .IX Item "EV_PREPARE"
1004 .PD 0
1005 .ie n .IP """EV_CHECK""" 4
1006 .el .IP "\f(CWEV_CHECK\fR" 4
1007 .IX Item "EV_CHECK"
1008 .PD
1009 All \f(CW\*(C`ev_prepare\*(C'\fR watchers are invoked just \fIbefore\fR \f(CW\*(C`ev_loop\*(C'\fR starts
1010 to gather new events, and all \f(CW\*(C`ev_check\*(C'\fR watchers are invoked just after
1011 \&\f(CW\*(C`ev_loop\*(C'\fR has gathered them, but before it invokes any callbacks for any
1012 received events. Callbacks of both watcher types can start and stop as
1013 many watchers as they want, and all of them will be taken into account
1014 (for example, a \f(CW\*(C`ev_prepare\*(C'\fR watcher might start an idle watcher to keep
1015 \&\f(CW\*(C`ev_loop\*(C'\fR from blocking).
1016 .ie n .IP """EV_EMBED""" 4
1017 .el .IP "\f(CWEV_EMBED\fR" 4
1018 .IX Item "EV_EMBED"
1019 The embedded event loop specified in the \f(CW\*(C`ev_embed\*(C'\fR watcher needs attention.
1020 .ie n .IP """EV_FORK""" 4
1021 .el .IP "\f(CWEV_FORK\fR" 4
1022 .IX Item "EV_FORK"
1023 The event loop has been resumed in the child process after fork (see
1024 \&\f(CW\*(C`ev_fork\*(C'\fR).
1025 .ie n .IP """EV_ASYNC""" 4
1026 .el .IP "\f(CWEV_ASYNC\fR" 4
1027 .IX Item "EV_ASYNC"
1028 The given async watcher has been asynchronously notified (see \f(CW\*(C`ev_async\*(C'\fR).
1029 .ie n .IP """EV_ERROR""" 4
1030 .el .IP "\f(CWEV_ERROR\fR" 4
1031 .IX Item "EV_ERROR"
1032 An unspecified error has occurred, the watcher has been stopped. This might
1033 happen because the watcher could not be properly started because libev
1034 ran out of memory, a file descriptor was found to be closed or any other
1035 problem. You best act on it by reporting the problem and somehow coping
1036 with the watcher being stopped.
1037 .Sp
1038 Libev will usually signal a few \*(L"dummy\*(R" events together with an error, for
1039 example it might indicate that a fd is readable or writable, and if your
1040 callbacks is well-written it can just attempt the operation and cope with
1041 the error from \fIread()\fR or \fIwrite()\fR. This will not work in multi-threaded
1042 programs, though, as the fd could already be closed and reused for another
1043 thing, so beware.
1044 .Sh "\s-1GENERIC\s0 \s-1WATCHER\s0 \s-1FUNCTIONS\s0"
1045 .IX Subsection "GENERIC WATCHER FUNCTIONS"
1046 In the following description, \f(CW\*(C`TYPE\*(C'\fR stands for the watcher type,
1047 e.g. \f(CW\*(C`timer\*(C'\fR for \f(CW\*(C`ev_timer\*(C'\fR watchers and \f(CW\*(C`io\*(C'\fR for \f(CW\*(C`ev_io\*(C'\fR watchers.
1048 .ie n .IP """ev_init"" (ev_TYPE *watcher, callback)" 4
1049 .el .IP "\f(CWev_init\fR (ev_TYPE *watcher, callback)" 4
1050 .IX Item "ev_init (ev_TYPE *watcher, callback)"
1051 This macro initialises the generic portion of a watcher. The contents
1052 of the watcher object can be arbitrary (so \f(CW\*(C`malloc\*(C'\fR will do). Only
1053 the generic parts of the watcher are initialised, you \fIneed\fR to call
1054 the type-specific \f(CW\*(C`ev_TYPE_set\*(C'\fR macro afterwards to initialise the
1055 type-specific parts. For each type there is also a \f(CW\*(C`ev_TYPE_init\*(C'\fR macro
1056 which rolls both calls into one.
1057 .Sp
1058 You can reinitialise a watcher at any time as long as it has been stopped
1059 (or never started) and there are no pending events outstanding.
1060 .Sp
1061 The callback is always of type \f(CW\*(C`void (*)(ev_loop *loop, ev_TYPE *watcher,
1062 int revents)\*(C'\fR.
1063 .Sp
1064 Example: Initialise an \f(CW\*(C`ev_io\*(C'\fR watcher in two steps.
1065 .Sp
1066 .Vb 3
1067 \& ev_io w;
1068 \& ev_init (&w, my_cb);
1069 \& ev_io_set (&w, STDIN_FILENO, EV_READ);
1070 .Ve
1071 .ie n .IP """ev_TYPE_set"" (ev_TYPE *, [args])" 4
1072 .el .IP "\f(CWev_TYPE_set\fR (ev_TYPE *, [args])" 4
1073 .IX Item "ev_TYPE_set (ev_TYPE *, [args])"
1074 This macro initialises the type-specific parts of a watcher. You need to
1075 call \f(CW\*(C`ev_init\*(C'\fR at least once before you call this macro, but you can
1076 call \f(CW\*(C`ev_TYPE_set\*(C'\fR any number of times. You must not, however, call this
1077 macro on a watcher that is active (it can be pending, however, which is a
1078 difference to the \f(CW\*(C`ev_init\*(C'\fR macro).
1079 .Sp
1080 Although some watcher types do not have type-specific arguments
1081 (e.g. \f(CW\*(C`ev_prepare\*(C'\fR) you still need to call its \f(CW\*(C`set\*(C'\fR macro.
1082 .Sp
1083 See \f(CW\*(C`ev_init\*(C'\fR, above, for an example.
1084 .ie n .IP """ev_TYPE_init"" (ev_TYPE *watcher, callback, [args])" 4
1085 .el .IP "\f(CWev_TYPE_init\fR (ev_TYPE *watcher, callback, [args])" 4
1086 .IX Item "ev_TYPE_init (ev_TYPE *watcher, callback, [args])"
1087 This convenience macro rolls both \f(CW\*(C`ev_init\*(C'\fR and \f(CW\*(C`ev_TYPE_set\*(C'\fR macro
1088 calls into a single call. This is the most convenient method to initialise
1089 a watcher. The same limitations apply, of course.
1090 .Sp
1091 Example: Initialise and set an \f(CW\*(C`ev_io\*(C'\fR watcher in one step.
1092 .Sp
1093 .Vb 1
1094 \& ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
1095 .Ve
1096 .ie n .IP """ev_TYPE_start"" (loop *, ev_TYPE *watcher)" 4
1097 .el .IP "\f(CWev_TYPE_start\fR (loop *, ev_TYPE *watcher)" 4
1098 .IX Item "ev_TYPE_start (loop *, ev_TYPE *watcher)"
1099 Starts (activates) the given watcher. Only active watchers will receive
1100 events. If the watcher is already active nothing will happen.
1101 .Sp
1102 Example: Start the \f(CW\*(C`ev_io\*(C'\fR watcher that is being abused as example in this
1103 whole section.
1104 .Sp
1105 .Vb 1
1106 \& ev_io_start (EV_DEFAULT_UC, &w);
1107 .Ve
1108 .ie n .IP """ev_TYPE_stop"" (loop *, ev_TYPE *watcher)" 4
1109 .el .IP "\f(CWev_TYPE_stop\fR (loop *, ev_TYPE *watcher)" 4
1110 .IX Item "ev_TYPE_stop (loop *, ev_TYPE *watcher)"
1111 Stops the given watcher if active, and clears the pending status (whether
1112 the watcher was active or not).
1113 .Sp
1114 It is possible that stopped watchers are pending \- for example,
1115 non-repeating timers are being stopped when they become pending \- but
1116 calling \f(CW\*(C`ev_TYPE_stop\*(C'\fR ensures that the watcher is neither active nor
1117 pending. If you want to free or reuse the memory used by the watcher it is
1118 therefore a good idea to always call its \f(CW\*(C`ev_TYPE_stop\*(C'\fR function.
1119 .IP "bool ev_is_active (ev_TYPE *watcher)" 4
1120 .IX Item "bool ev_is_active (ev_TYPE *watcher)"
1121 Returns a true value iff the watcher is active (i.e. it has been started
1122 and not yet been stopped). As long as a watcher is active you must not modify
1123 it.
1124 .IP "bool ev_is_pending (ev_TYPE *watcher)" 4
1125 .IX Item "bool ev_is_pending (ev_TYPE *watcher)"
1126 Returns a true value iff the watcher is pending, (i.e. it has outstanding
1127 events but its callback has not yet been invoked). As long as a watcher
1128 is pending (but not active) you must not call an init function on it (but
1129 \&\f(CW\*(C`ev_TYPE_set\*(C'\fR is safe), you must not change its priority, and you must
1130 make sure the watcher is available to libev (e.g. you cannot \f(CW\*(C`free ()\*(C'\fR
1131 it).
1132 .IP "callback ev_cb (ev_TYPE *watcher)" 4
1133 .IX Item "callback ev_cb (ev_TYPE *watcher)"
1134 Returns the callback currently set on the watcher.
1135 .IP "ev_cb_set (ev_TYPE *watcher, callback)" 4
1136 .IX Item "ev_cb_set (ev_TYPE *watcher, callback)"
1137 Change the callback. You can change the callback at virtually any time
1138 (modulo threads).
1139 .IP "ev_set_priority (ev_TYPE *watcher, priority)" 4
1140 .IX Item "ev_set_priority (ev_TYPE *watcher, priority)"
1141 .PD 0
1142 .IP "int ev_priority (ev_TYPE *watcher)" 4
1143 .IX Item "int ev_priority (ev_TYPE *watcher)"
1144 .PD
1145 Set and query the priority of the watcher. The priority is a small
1146 integer between \f(CW\*(C`EV_MAXPRI\*(C'\fR (default: \f(CW2\fR) and \f(CW\*(C`EV_MINPRI\*(C'\fR
1147 (default: \f(CW\*(C`\-2\*(C'\fR). Pending watchers with higher priority will be invoked
1148 before watchers with lower priority, but priority will not keep watchers
1149 from being executed (except for \f(CW\*(C`ev_idle\*(C'\fR watchers).
1150 .Sp
1151 This means that priorities are \fIonly\fR used for ordering callback
1152 invocation after new events have been received. This is useful, for
1153 example, to reduce latency after idling, or more often, to bind two
1154 watchers on the same event and make sure one is called first.
1155 .Sp
1156 If you need to suppress invocation when higher priority events are pending
1157 you need to look at \f(CW\*(C`ev_idle\*(C'\fR watchers, which provide this functionality.
1158 .Sp
1159 You \fImust not\fR change the priority of a watcher as long as it is active or
1160 pending.
1161 .Sp
1162 The default priority used by watchers when no priority has been set is
1163 always \f(CW0\fR, which is supposed to not be too high and not be too low :).
1164 .Sp
1165 Setting a priority outside the range of \f(CW\*(C`EV_MINPRI\*(C'\fR to \f(CW\*(C`EV_MAXPRI\*(C'\fR is
1166 fine, as long as you do not mind that the priority value you query might
1167 or might not have been adjusted to be within valid range.
1168 .IP "ev_invoke (loop, ev_TYPE *watcher, int revents)" 4
1169 .IX Item "ev_invoke (loop, ev_TYPE *watcher, int revents)"
1170 Invoke the \f(CW\*(C`watcher\*(C'\fR with the given \f(CW\*(C`loop\*(C'\fR and \f(CW\*(C`revents\*(C'\fR. Neither
1171 \&\f(CW\*(C`loop\*(C'\fR nor \f(CW\*(C`revents\*(C'\fR need to be valid as long as the watcher callback
1172 can deal with that fact, as both are simply passed through to the
1173 callback.
1174 .IP "int ev_clear_pending (loop, ev_TYPE *watcher)" 4
1175 .IX Item "int ev_clear_pending (loop, ev_TYPE *watcher)"
1176 If the watcher is pending, this function clears its pending status and
1177 returns its \f(CW\*(C`revents\*(C'\fR bitset (as if its callback was invoked). If the
1178 watcher isn't pending it does nothing and returns \f(CW0\fR.
1179 .Sp
1180 Sometimes it can be useful to \*(L"poll\*(R" a watcher instead of waiting for its
1181 callback to be invoked, which can be accomplished with this function.
1182 .Sh "\s-1ASSOCIATING\s0 \s-1CUSTOM\s0 \s-1DATA\s0 \s-1WITH\s0 A \s-1WATCHER\s0"
1183 .IX Subsection "ASSOCIATING CUSTOM DATA WITH A WATCHER"
1184 Each watcher has, by default, a member \f(CW\*(C`void *data\*(C'\fR that you can change
1185 and read at any time: libev will completely ignore it. This can be used
1186 to associate arbitrary data with your watcher. If you need more data and
1187 don't want to allocate memory and store a pointer to it in that data
1188 member, you can also \*(L"subclass\*(R" the watcher type and provide your own
1189 data:
1190 .PP
1191 .Vb 7
1192 \& struct my_io
1193 \& {
1194 \& struct ev_io io;
1195 \& int otherfd;
1196 \& void *somedata;
1197 \& struct whatever *mostinteresting;
1198 \& };
1199 \&
1200 \& ...
1201 \& struct my_io w;
1202 \& ev_io_init (&w.io, my_cb, fd, EV_READ);
1203 .Ve
1204 .PP
1205 And since your callback will be called with a pointer to the watcher, you
1206 can cast it back to your own type:
1207 .PP
1208 .Vb 5
1209 \& static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
1210 \& {
1211 \& struct my_io *w = (struct my_io *)w_;
1212 \& ...
1213 \& }
1214 .Ve
1215 .PP
1216 More interesting and less C\-conformant ways of casting your callback type
1217 instead have been omitted.
1218 .PP
1219 Another common scenario is to use some data structure with multiple
1220 embedded watchers:
1221 .PP
1222 .Vb 6
1223 \& struct my_biggy
1224 \& {
1225 \& int some_data;
1226 \& ev_timer t1;
1227 \& ev_timer t2;
1228 \& }
1229 .Ve
1230 .PP
1231 In this case getting the pointer to \f(CW\*(C`my_biggy\*(C'\fR is a bit more
1232 complicated: Either you store the address of your \f(CW\*(C`my_biggy\*(C'\fR struct
1233 in the \f(CW\*(C`data\*(C'\fR member of the watcher (for woozies), or you need to use
1234 some pointer arithmetic using \f(CW\*(C`offsetof\*(C'\fR inside your watchers (for real
1235 programmers):
1236 .PP
1237 .Vb 1
1238 \& #include <stddef.h>
1239 \&
1240 \& static void
1241 \& t1_cb (EV_P_ struct ev_timer *w, int revents)
1242 \& {
1243 \& struct my_biggy big = (struct my_biggy *
1244 \& (((char *)w) \- offsetof (struct my_biggy, t1));
1245 \& }
1246 \&
1247 \& static void
1248 \& t2_cb (EV_P_ struct ev_timer *w, int revents)
1249 \& {
1250 \& struct my_biggy big = (struct my_biggy *
1251 \& (((char *)w) \- offsetof (struct my_biggy, t2));
1252 \& }
1253 .Ve
1254 .SH "WATCHER TYPES"
1255 .IX Header "WATCHER TYPES"
1256 This section describes each watcher in detail, but will not repeat
1257 information given in the last section. Any initialisation/set macros,
1258 functions and members specific to the watcher type are explained.
1259 .PP
1260 Members are additionally marked with either \fI[read\-only]\fR, meaning that,
1261 while the watcher is active, you can look at the member and expect some
1262 sensible content, but you must not modify it (you can modify it while the
1263 watcher is stopped to your hearts content), or \fI[read\-write]\fR, which
1264 means you can expect it to have some sensible content while the watcher
1265 is active, but you can also modify it. Modifying it may not do something
1266 sensible or take immediate effect (or do anything at all), but libev will
1267 not crash or malfunction in any way.
1268 .ie n .Sh """ev_io"" \- is this file descriptor readable or writable?"
1269 .el .Sh "\f(CWev_io\fP \- is this file descriptor readable or writable?"
1270 .IX Subsection "ev_io - is this file descriptor readable or writable?"
1271 I/O watchers check whether a file descriptor is readable or writable
1272 in each iteration of the event loop, or, more precisely, when reading
1273 would not block the process and writing would at least be able to write
1274 some data. This behaviour is called level-triggering because you keep
1275 receiving events as long as the condition persists. Remember you can stop
1276 the watcher if you don't want to act on the event and neither want to
1277 receive future events.
1278 .PP
1279 In general you can register as many read and/or write event watchers per
1280 fd as you want (as long as you don't confuse yourself). Setting all file
1281 descriptors to non-blocking mode is also usually a good idea (but not
1282 required if you know what you are doing).
1283 .PP
1284 If you cannot use non-blocking mode, then force the use of a
1285 known-to-be-good backend (at the time of this writing, this includes only
1286 \&\f(CW\*(C`EVBACKEND_SELECT\*(C'\fR and \f(CW\*(C`EVBACKEND_POLL\*(C'\fR).
1287 .PP
1288 Another thing you have to watch out for is that it is quite easy to
1289 receive \*(L"spurious\*(R" readiness notifications, that is your callback might
1290 be called with \f(CW\*(C`EV_READ\*(C'\fR but a subsequent \f(CW\*(C`read\*(C'\fR(2) will actually block
1291 because there is no data. Not only are some backends known to create a
1292 lot of those (for example Solaris ports), it is very easy to get into
1293 this situation even with a relatively standard program structure. Thus
1294 it is best to always use non-blocking I/O: An extra \f(CW\*(C`read\*(C'\fR(2) returning
1295 \&\f(CW\*(C`EAGAIN\*(C'\fR is far preferable to a program hanging until some data arrives.
1296 .PP
1297 If you cannot run the fd in non-blocking mode (for example you should
1298 not play around with an Xlib connection), then you have to separately
1299 re-test whether a file descriptor is really ready with a known-to-be good
1300 interface such as poll (fortunately in our Xlib example, Xlib already
1301 does this on its own, so its quite safe to use). Some people additionally
1302 use \f(CW\*(C`SIGALRM\*(C'\fR and an interval timer, just to be sure you won't block
1303 indefinitely.
1304 .PP
1305 But really, best use non-blocking mode.
1306 .PP
1307 \fIThe special problem of disappearing file descriptors\fR
1308 .IX Subsection "The special problem of disappearing file descriptors"
1309 .PP
1310 Some backends (e.g. kqueue, epoll) need to be told about closing a file
1311 descriptor (either due to calling \f(CW\*(C`close\*(C'\fR explicitly or any other means,
1312 such as \f(CW\*(C`dup2\*(C'\fR). The reason is that you register interest in some file
1313 descriptor, but when it goes away, the operating system will silently drop
1314 this interest. If another file descriptor with the same number then is
1315 registered with libev, there is no efficient way to see that this is, in
1316 fact, a different file descriptor.
1317 .PP
1318 To avoid having to explicitly tell libev about such cases, libev follows
1319 the following policy: Each time \f(CW\*(C`ev_io_set\*(C'\fR is being called, libev
1320 will assume that this is potentially a new file descriptor, otherwise
1321 it is assumed that the file descriptor stays the same. That means that
1322 you \fIhave\fR to call \f(CW\*(C`ev_io_set\*(C'\fR (or \f(CW\*(C`ev_io_init\*(C'\fR) when you change the
1323 descriptor even if the file descriptor number itself did not change.
1324 .PP
1325 This is how one would do it normally anyway, the important point is that
1326 the libev application should not optimise around libev but should leave
1327 optimisations to libev.
1328 .PP
1329 \fIThe special problem of dup'ed file descriptors\fR
1330 .IX Subsection "The special problem of dup'ed file descriptors"
1331 .PP
1332 Some backends (e.g. epoll), cannot register events for file descriptors,
1333 but only events for the underlying file descriptions. That means when you
1334 have \f(CW\*(C`dup ()\*(C'\fR'ed file descriptors or weirder constellations, and register
1335 events for them, only one file descriptor might actually receive events.
1336 .PP
1337 There is no workaround possible except not registering events
1338 for potentially \f(CW\*(C`dup ()\*(C'\fR'ed file descriptors, or to resort to
1339 \&\f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or \f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
1340 .PP
1341 \fIThe special problem of fork\fR
1342 .IX Subsection "The special problem of fork"
1343 .PP
1344 Some backends (epoll, kqueue) do not support \f(CW\*(C`fork ()\*(C'\fR at all or exhibit
1345 useless behaviour. Libev fully supports fork, but needs to be told about
1346 it in the child.
1347 .PP
1348 To support fork in your programs, you either have to call
1349 \&\f(CW\*(C`ev_default_fork ()\*(C'\fR or \f(CW\*(C`ev_loop_fork ()\*(C'\fR after a fork in the child,
1350 enable \f(CW\*(C`EVFLAG_FORKCHECK\*(C'\fR, or resort to \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or
1351 \&\f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
1352 .PP
1353 \fIThe special problem of \s-1SIGPIPE\s0\fR
1354 .IX Subsection "The special problem of SIGPIPE"
1355 .PP
1356 While not really specific to libev, it is easy to forget about \f(CW\*(C`SIGPIPE\*(C'\fR:
1357 when writing to a pipe whose other end has been closed, your program gets
1358 sent a \s-1SIGPIPE\s0, which, by default, aborts your program. For most programs
1359 this is sensible behaviour, for daemons, this is usually undesirable.
1360 .PP
1361 So when you encounter spurious, unexplained daemon exits, make sure you
1362 ignore \s-1SIGPIPE\s0 (and maybe make sure you log the exit status of your daemon
1363 somewhere, as that would have given you a big clue).
1364 .PP
1365 \fIWatcher-Specific Functions\fR
1366 .IX Subsection "Watcher-Specific Functions"
1367 .IP "ev_io_init (ev_io *, callback, int fd, int events)" 4
1368 .IX Item "ev_io_init (ev_io *, callback, int fd, int events)"
1369 .PD 0
1370 .IP "ev_io_set (ev_io *, int fd, int events)" 4
1371 .IX Item "ev_io_set (ev_io *, int fd, int events)"
1372 .PD
1373 Configures an \f(CW\*(C`ev_io\*(C'\fR watcher. The \f(CW\*(C`fd\*(C'\fR is the file descriptor to
1374 receive events for and \f(CW\*(C`events\*(C'\fR is either \f(CW\*(C`EV_READ\*(C'\fR, \f(CW\*(C`EV_WRITE\*(C'\fR or
1375 \&\f(CW\*(C`EV_READ | EV_WRITE\*(C'\fR, to express the desire to receive the given events.
1376 .IP "int fd [read\-only]" 4
1377 .IX Item "int fd [read-only]"
1378 The file descriptor being watched.
1379 .IP "int events [read\-only]" 4
1380 .IX Item "int events [read-only]"
1381 The events being watched.
1382 .PP
1383 \fIExamples\fR
1384 .IX Subsection "Examples"
1385 .PP
1386 Example: Call \f(CW\*(C`stdin_readable_cb\*(C'\fR when \s-1STDIN_FILENO\s0 has become, well
1387 readable, but only once. Since it is likely line-buffered, you could
1388 attempt to read a whole line in the callback.
1389 .PP
1390 .Vb 6
1391 \& static void
1392 \& stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1393 \& {
1394 \& ev_io_stop (loop, w);
1395 \& .. read from stdin here (or from w\->fd) and handle any I/O errors
1396 \& }
1397 \&
1398 \& ...
1399 \& struct ev_loop *loop = ev_default_init (0);
1400 \& struct ev_io stdin_readable;
1401 \& ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1402 \& ev_io_start (loop, &stdin_readable);
1403 \& ev_loop (loop, 0);
1404 .Ve
1405 .ie n .Sh """ev_timer"" \- relative and optionally repeating timeouts"
1406 .el .Sh "\f(CWev_timer\fP \- relative and optionally repeating timeouts"
1407 .IX Subsection "ev_timer - relative and optionally repeating timeouts"
1408 Timer watchers are simple relative timers that generate an event after a
1409 given time, and optionally repeating in regular intervals after that.
1410 .PP
1411 The timers are based on real time, that is, if you register an event that
1412 times out after an hour and you reset your system clock to January last
1413 year, it will still time out after (roughly) one hour. \*(L"Roughly\*(R" because
1414 detecting time jumps is hard, and some inaccuracies are unavoidable (the
1415 monotonic clock option helps a lot here).
1416 .PP
1417 The callback is guaranteed to be invoked only \fIafter\fR its timeout has
1418 passed, but if multiple timers become ready during the same loop iteration
1419 then order of execution is undefined.
1420 .PP
1421 \fIThe special problem of time updates\fR
1422 .IX Subsection "The special problem of time updates"
1423 .PP
1424 Establishing the current time is a costly operation (it usually takes at
1425 least two system calls): \s-1EV\s0 therefore updates its idea of the current
1426 time only before and after \f(CW\*(C`ev_loop\*(C'\fR collects new events, which causes a
1427 growing difference between \f(CW\*(C`ev_now ()\*(C'\fR and \f(CW\*(C`ev_time ()\*(C'\fR when handling
1428 lots of events in one iteration.
1429 .PP
1430 The relative timeouts are calculated relative to the \f(CW\*(C`ev_now ()\*(C'\fR
1431 time. This is usually the right thing as this timestamp refers to the time
1432 of the event triggering whatever timeout you are modifying/starting. If
1433 you suspect event processing to be delayed and you \fIneed\fR to base the
1434 timeout on the current time, use something like this to adjust for this:
1435 .PP
1436 .Vb 1
1437 \& ev_timer_set (&timer, after + ev_now () \- ev_time (), 0.);
1438 .Ve
1439 .PP
1440 If the event loop is suspended for a long time, you can also force an
1441 update of the time returned by \f(CW\*(C`ev_now ()\*(C'\fR by calling \f(CW\*(C`ev_now_update
1442 ()\*(C'\fR.
1443 .PP
1444 \fIWatcher-Specific Functions and Data Members\fR
1445 .IX Subsection "Watcher-Specific Functions and Data Members"
1446 .IP "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)" 4
1447 .IX Item "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)"
1448 .PD 0
1449 .IP "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)" 4
1450 .IX Item "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)"
1451 .PD
1452 Configure the timer to trigger after \f(CW\*(C`after\*(C'\fR seconds. If \f(CW\*(C`repeat\*(C'\fR
1453 is \f(CW0.\fR, then it will automatically be stopped once the timeout is
1454 reached. If it is positive, then the timer will automatically be
1455 configured to trigger again \f(CW\*(C`repeat\*(C'\fR seconds later, again, and again,
1456 until stopped manually.
1457 .Sp
1458 The timer itself will do a best-effort at avoiding drift, that is, if
1459 you configure a timer to trigger every 10 seconds, then it will normally
1460 trigger at exactly 10 second intervals. If, however, your program cannot
1461 keep up with the timer (because it takes longer than those 10 seconds to
1462 do stuff) the timer will not fire more than once per event loop iteration.
1463 .IP "ev_timer_again (loop, ev_timer *)" 4
1464 .IX Item "ev_timer_again (loop, ev_timer *)"
1465 This will act as if the timer timed out and restart it again if it is
1466 repeating. The exact semantics are:
1467 .Sp
1468 If the timer is pending, its pending status is cleared.
1469 .Sp
1470 If the timer is started but non-repeating, stop it (as if it timed out).
1471 .Sp
1472 If the timer is repeating, either start it if necessary (with the
1473 \&\f(CW\*(C`repeat\*(C'\fR value), or reset the running timer to the \f(CW\*(C`repeat\*(C'\fR value.
1474 .Sp
1475 This sounds a bit complicated, but here is a useful and typical
1476 example: Imagine you have a \s-1TCP\s0 connection and you want a so-called idle
1477 timeout, that is, you want to be called when there have been, say, 60
1478 seconds of inactivity on the socket. The easiest way to do this is to
1479 configure an \f(CW\*(C`ev_timer\*(C'\fR with a \f(CW\*(C`repeat\*(C'\fR value of \f(CW60\fR and then call
1480 \&\f(CW\*(C`ev_timer_again\*(C'\fR each time you successfully read or write some data. If
1481 you go into an idle state where you do not expect data to travel on the
1482 socket, you can \f(CW\*(C`ev_timer_stop\*(C'\fR the timer, and \f(CW\*(C`ev_timer_again\*(C'\fR will
1483 automatically restart it if need be.
1484 .Sp
1485 That means you can ignore the \f(CW\*(C`after\*(C'\fR value and \f(CW\*(C`ev_timer_start\*(C'\fR
1486 altogether and only ever use the \f(CW\*(C`repeat\*(C'\fR value and \f(CW\*(C`ev_timer_again\*(C'\fR:
1487 .Sp
1488 .Vb 8
1489 \& ev_timer_init (timer, callback, 0., 5.);
1490 \& ev_timer_again (loop, timer);
1491 \& ...
1492 \& timer\->again = 17.;
1493 \& ev_timer_again (loop, timer);
1494 \& ...
1495 \& timer\->again = 10.;
1496 \& ev_timer_again (loop, timer);
1497 .Ve
1498 .Sp
1499 This is more slightly efficient then stopping/starting the timer each time
1500 you want to modify its timeout value.
1501 .Sp
1502 Note, however, that it is often even more efficient to remember the
1503 time of the last activity and let the timer time-out naturally. In the
1504 callback, you then check whether the time-out is real, or, if there was
1505 some activity, you reschedule the watcher to time-out in \*(L"last_activity +
1506 timeout \- ev_now ()\*(R" seconds.
1507 .IP "ev_tstamp repeat [read\-write]" 4
1508 .IX Item "ev_tstamp repeat [read-write]"
1509 The current \f(CW\*(C`repeat\*(C'\fR value. Will be used each time the watcher times out
1510 or \f(CW\*(C`ev_timer_again\*(C'\fR is called, and determines the next timeout (if any),
1511 which is also when any modifications are taken into account.
1512 .PP
1513 \fIExamples\fR
1514 .IX Subsection "Examples"
1515 .PP
1516 Example: Create a timer that fires after 60 seconds.
1517 .PP
1518 .Vb 5
1519 \& static void
1520 \& one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1521 \& {
1522 \& .. one minute over, w is actually stopped right here
1523 \& }
1524 \&
1525 \& struct ev_timer mytimer;
1526 \& ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1527 \& ev_timer_start (loop, &mytimer);
1528 .Ve
1529 .PP
1530 Example: Create a timeout timer that times out after 10 seconds of
1531 inactivity.
1532 .PP
1533 .Vb 5
1534 \& static void
1535 \& timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1536 \& {
1537 \& .. ten seconds without any activity
1538 \& }
1539 \&
1540 \& struct ev_timer mytimer;
1541 \& ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1542 \& ev_timer_again (&mytimer); /* start timer */
1543 \& ev_loop (loop, 0);
1544 \&
1545 \& // and in some piece of code that gets executed on any "activity":
1546 \& // reset the timeout to start ticking again at 10 seconds
1547 \& ev_timer_again (&mytimer);
1548 .Ve
1549 .ie n .Sh """ev_periodic"" \- to cron or not to cron?"
1550 .el .Sh "\f(CWev_periodic\fP \- to cron or not to cron?"
1551 .IX Subsection "ev_periodic - to cron or not to cron?"
1552 Periodic watchers are also timers of a kind, but they are very versatile
1553 (and unfortunately a bit complex).
1554 .PP
1555 Unlike \f(CW\*(C`ev_timer\*(C'\fR's, they are not based on real time (or relative time)
1556 but on wall clock time (absolute time). You can tell a periodic watcher
1557 to trigger after some specific point in time. For example, if you tell a
1558 periodic watcher to trigger in 10 seconds (by specifying e.g. \f(CW\*(C`ev_now ()
1559 + 10.\*(C'\fR, that is, an absolute time not a delay) and then reset your system
1560 clock to January of the previous year, then it will take more than year
1561 to trigger the event (unlike an \f(CW\*(C`ev_timer\*(C'\fR, which would still trigger
1562 roughly 10 seconds later as it uses a relative timeout).
1563 .PP
1564 \&\f(CW\*(C`ev_periodic\*(C'\fRs can also be used to implement vastly more complex timers,
1565 such as triggering an event on each \*(L"midnight, local time\*(R", or other
1566 complicated rules.
1567 .PP
1568 As with timers, the callback is guaranteed to be invoked only when the
1569 time (\f(CW\*(C`at\*(C'\fR) has passed, but if multiple periodic timers become ready
1570 during the same loop iteration, then order of execution is undefined.
1571 .PP
1572 \fIWatcher-Specific Functions and Data Members\fR
1573 .IX Subsection "Watcher-Specific Functions and Data Members"
1574 .IP "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)" 4
1575 .IX Item "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)"
1576 .PD 0
1577 .IP "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)" 4
1578 .IX Item "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)"
1579 .PD
1580 Lots of arguments, lets sort it out... There are basically three modes of
1581 operation, and we will explain them from simplest to most complex:
1582 .RS 4
1583 .IP "\(bu" 4
1584 absolute timer (at = time, interval = reschedule_cb = 0)
1585 .Sp
1586 In this configuration the watcher triggers an event after the wall clock
1587 time \f(CW\*(C`at\*(C'\fR has passed. It will not repeat and will not adjust when a time
1588 jump occurs, that is, if it is to be run at January 1st 2011 then it will
1589 only run when the system clock reaches or surpasses this time.
1590 .IP "\(bu" 4
1591 repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1592 .Sp
1593 In this mode the watcher will always be scheduled to time out at the next
1594 \&\f(CW\*(C`at + N * interval\*(C'\fR time (for some integer N, which can also be negative)
1595 and then repeat, regardless of any time jumps.
1596 .Sp
1597 This can be used to create timers that do not drift with respect to the
1598 system clock, for example, here is a \f(CW\*(C`ev_periodic\*(C'\fR that triggers each
1599 hour, on the hour:
1600 .Sp
1601 .Vb 1
1602 \& ev_periodic_set (&periodic, 0., 3600., 0);
1603 .Ve
1604 .Sp
1605 This doesn't mean there will always be 3600 seconds in between triggers,
1606 but only that the callback will be called when the system time shows a
1607 full hour (\s-1UTC\s0), or more correctly, when the system time is evenly divisible
1608 by 3600.
1609 .Sp
1610 Another way to think about it (for the mathematically inclined) is that
1611 \&\f(CW\*(C`ev_periodic\*(C'\fR will try to run the callback in this mode at the next possible
1612 time where \f(CW\*(C`time = at (mod interval)\*(C'\fR, regardless of any time jumps.
1613 .Sp
1614 For numerical stability it is preferable that the \f(CW\*(C`at\*(C'\fR value is near
1615 \&\f(CW\*(C`ev_now ()\*(C'\fR (the current time), but there is no range requirement for
1616 this value, and in fact is often specified as zero.
1617 .Sp
1618 Note also that there is an upper limit to how often a timer can fire (\s-1CPU\s0
1619 speed for example), so if \f(CW\*(C`interval\*(C'\fR is very small then timing stability
1620 will of course deteriorate. Libev itself tries to be exact to be about one
1621 millisecond (if the \s-1OS\s0 supports it and the machine is fast enough).
1622 .IP "\(bu" 4
1623 manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1624 .Sp
1625 In this mode the values for \f(CW\*(C`interval\*(C'\fR and \f(CW\*(C`at\*(C'\fR are both being
1626 ignored. Instead, each time the periodic watcher gets scheduled, the
1627 reschedule callback will be called with the watcher as first, and the
1628 current time as second argument.
1629 .Sp
1630 \&\s-1NOTE:\s0 \fIThis callback \s-1MUST\s0 \s-1NOT\s0 stop or destroy any periodic watcher,
1631 ever, or make \s-1ANY\s0 event loop modifications whatsoever\fR.
1632 .Sp
1633 If you need to stop it, return \f(CW\*(C`now + 1e30\*(C'\fR (or so, fudge fudge) and stop
1634 it afterwards (e.g. by starting an \f(CW\*(C`ev_prepare\*(C'\fR watcher, which is the
1635 only event loop modification you are allowed to do).
1636 .Sp
1637 The callback prototype is \f(CW\*(C`ev_tstamp (*reschedule_cb)(struct ev_periodic
1638 *w, ev_tstamp now)\*(C'\fR, e.g.:
1639 .Sp
1640 .Vb 4
1641 \& static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
1642 \& {
1643 \& return now + 60.;
1644 \& }
1645 .Ve
1646 .Sp
1647 It must return the next time to trigger, based on the passed time value
1648 (that is, the lowest time value larger than to the second argument). It
1649 will usually be called just before the callback will be triggered, but
1650 might be called at other times, too.
1651 .Sp
1652 \&\s-1NOTE:\s0 \fIThis callback must always return a time that is higher than or
1653 equal to the passed \f(CI\*(C`now\*(C'\fI value\fR.
1654 .Sp
1655 This can be used to create very complex timers, such as a timer that
1656 triggers on \*(L"next midnight, local time\*(R". To do this, you would calculate the
1657 next midnight after \f(CW\*(C`now\*(C'\fR and return the timestamp value for this. How
1658 you do this is, again, up to you (but it is not trivial, which is the main
1659 reason I omitted it as an example).
1660 .RE
1661 .RS 4
1662 .RE
1663 .IP "ev_periodic_again (loop, ev_periodic *)" 4
1664 .IX Item "ev_periodic_again (loop, ev_periodic *)"
1665 Simply stops and restarts the periodic watcher again. This is only useful
1666 when you changed some parameters or the reschedule callback would return
1667 a different time than the last time it was called (e.g. in a crond like
1668 program when the crontabs have changed).
1669 .IP "ev_tstamp ev_periodic_at (ev_periodic *)" 4
1670 .IX Item "ev_tstamp ev_periodic_at (ev_periodic *)"
1671 When active, returns the absolute time that the watcher is supposed to
1672 trigger next.
1673 .IP "ev_tstamp offset [read\-write]" 4
1674 .IX Item "ev_tstamp offset [read-write]"
1675 When repeating, this contains the offset value, otherwise this is the
1676 absolute point in time (the \f(CW\*(C`at\*(C'\fR value passed to \f(CW\*(C`ev_periodic_set\*(C'\fR).
1677 .Sp
1678 Can be modified any time, but changes only take effect when the periodic
1679 timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being called.
1680 .IP "ev_tstamp interval [read\-write]" 4
1681 .IX Item "ev_tstamp interval [read-write]"
1682 The current interval value. Can be modified any time, but changes only
1683 take effect when the periodic timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being
1684 called.
1685 .IP "ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read\-write]" 4
1686 .IX Item "ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]"
1687 The current reschedule callback, or \f(CW0\fR, if this functionality is
1688 switched off. Can be changed any time, but changes only take effect when
1689 the periodic timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being called.
1690 .PP
1691 \fIExamples\fR
1692 .IX Subsection "Examples"
1693 .PP
1694 Example: Call a callback every hour, or, more precisely, whenever the
1695 system time is divisible by 3600. The callback invocation times have
1696 potentially a lot of jitter, but good long-term stability.
1697 .PP
1698 .Vb 5
1699 \& static void
1700 \& clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1701 \& {
1702 \& ... its now a full hour (UTC, or TAI or whatever your clock follows)
1703 \& }
1704 \&
1705 \& struct ev_periodic hourly_tick;
1706 \& ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1707 \& ev_periodic_start (loop, &hourly_tick);
1708 .Ve
1709 .PP
1710 Example: The same as above, but use a reschedule callback to do it:
1711 .PP
1712 .Vb 1
1713 \& #include <math.h>
1714 \&
1715 \& static ev_tstamp
1716 \& my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)
1717 \& {
1718 \& return now + (3600. \- fmod (now, 3600.));
1719 \& }
1720 \&
1721 \& ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1722 .Ve
1723 .PP
1724 Example: Call a callback every hour, starting now:
1725 .PP
1726 .Vb 4
1727 \& struct ev_periodic hourly_tick;
1728 \& ev_periodic_init (&hourly_tick, clock_cb,
1729 \& fmod (ev_now (loop), 3600.), 3600., 0);
1730 \& ev_periodic_start (loop, &hourly_tick);
1731 .Ve
1732 .ie n .Sh """ev_signal"" \- signal me when a signal gets signalled!"
1733 .el .Sh "\f(CWev_signal\fP \- signal me when a signal gets signalled!"
1734 .IX Subsection "ev_signal - signal me when a signal gets signalled!"
1735 Signal watchers will trigger an event when the process receives a specific
1736 signal one or more times. Even though signals are very asynchronous, libev
1737 will try it's best to deliver signals synchronously, i.e. as part of the
1738 normal event processing, like any other event.
1739 .PP
1740 If you want signals asynchronously, just use \f(CW\*(C`sigaction\*(C'\fR as you would
1741 do without libev and forget about sharing the signal. You can even use
1742 \&\f(CW\*(C`ev_async\*(C'\fR from a signal handler to synchronously wake up an event loop.
1743 .PP
1744 You can configure as many watchers as you like per signal. Only when the
1745 first watcher gets started will libev actually register a signal handler
1746 with the kernel (thus it coexists with your own signal handlers as long as
1747 you don't register any with libev for the same signal). Similarly, when
1748 the last signal watcher for a signal is stopped, libev will reset the
1749 signal handler to \s-1SIG_DFL\s0 (regardless of what it was set to before).
1750 .PP
1751 If possible and supported, libev will install its handlers with
1752 \&\f(CW\*(C`SA_RESTART\*(C'\fR behaviour enabled, so system calls should not be unduly
1753 interrupted. If you have a problem with system calls getting interrupted by
1754 signals you can block all signals in an \f(CW\*(C`ev_check\*(C'\fR watcher and unblock
1755 them in an \f(CW\*(C`ev_prepare\*(C'\fR watcher.
1756 .PP
1757 \fIWatcher-Specific Functions and Data Members\fR
1758 .IX Subsection "Watcher-Specific Functions and Data Members"
1759 .IP "ev_signal_init (ev_signal *, callback, int signum)" 4
1760 .IX Item "ev_signal_init (ev_signal *, callback, int signum)"
1761 .PD 0
1762 .IP "ev_signal_set (ev_signal *, int signum)" 4
1763 .IX Item "ev_signal_set (ev_signal *, int signum)"
1764 .PD
1765 Configures the watcher to trigger on the given signal number (usually one
1766 of the \f(CW\*(C`SIGxxx\*(C'\fR constants).
1767 .IP "int signum [read\-only]" 4
1768 .IX Item "int signum [read-only]"
1769 The signal the watcher watches out for.
1770 .PP
1771 \fIExamples\fR
1772 .IX Subsection "Examples"
1773 .PP
1774 Example: Try to exit cleanly on \s-1SIGINT\s0.
1775 .PP
1776 .Vb 5
1777 \& static void
1778 \& sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
1779 \& {
1780 \& ev_unloop (loop, EVUNLOOP_ALL);
1781 \& }
1782 \&
1783 \& struct ev_signal signal_watcher;
1784 \& ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1785 \& ev_signal_start (loop, &signal_watcher);
1786 .Ve
1787 .ie n .Sh """ev_child"" \- watch out for process status changes"
1788 .el .Sh "\f(CWev_child\fP \- watch out for process status changes"
1789 .IX Subsection "ev_child - watch out for process status changes"
1790 Child watchers trigger when your process receives a \s-1SIGCHLD\s0 in response to
1791 some child status changes (most typically when a child of yours dies or
1792 exits). It is permissible to install a child watcher \fIafter\fR the child
1793 has been forked (which implies it might have already exited), as long
1794 as the event loop isn't entered (or is continued from a watcher), i.e.,
1795 forking and then immediately registering a watcher for the child is fine,
1796 but forking and registering a watcher a few event loop iterations later is
1797 not.
1798 .PP
1799 Only the default event loop is capable of handling signals, and therefore
1800 you can only register child watchers in the default event loop.
1801 .PP
1802 \fIProcess Interaction\fR
1803 .IX Subsection "Process Interaction"
1804 .PP
1805 Libev grabs \f(CW\*(C`SIGCHLD\*(C'\fR as soon as the default event loop is
1806 initialised. This is necessary to guarantee proper behaviour even if
1807 the first child watcher is started after the child exits. The occurrence
1808 of \f(CW\*(C`SIGCHLD\*(C'\fR is recorded asynchronously, but child reaping is done
1809 synchronously as part of the event loop processing. Libev always reaps all
1810 children, even ones not watched.
1811 .PP
1812 \fIOverriding the Built-In Processing\fR
1813 .IX Subsection "Overriding the Built-In Processing"
1814 .PP
1815 Libev offers no special support for overriding the built-in child
1816 processing, but if your application collides with libev's default child
1817 handler, you can override it easily by installing your own handler for
1818 \&\f(CW\*(C`SIGCHLD\*(C'\fR after initialising the default loop, and making sure the
1819 default loop never gets destroyed. You are encouraged, however, to use an
1820 event-based approach to child reaping and thus use libev's support for
1821 that, so other libev users can use \f(CW\*(C`ev_child\*(C'\fR watchers freely.
1822 .PP
1823 \fIStopping the Child Watcher\fR
1824 .IX Subsection "Stopping the Child Watcher"
1825 .PP
1826 Currently, the child watcher never gets stopped, even when the
1827 child terminates, so normally one needs to stop the watcher in the
1828 callback. Future versions of libev might stop the watcher automatically
1829 when a child exit is detected.
1830 .PP
1831 \fIWatcher-Specific Functions and Data Members\fR
1832 .IX Subsection "Watcher-Specific Functions and Data Members"
1833 .IP "ev_child_init (ev_child *, callback, int pid, int trace)" 4
1834 .IX Item "ev_child_init (ev_child *, callback, int pid, int trace)"
1835 .PD 0
1836 .IP "ev_child_set (ev_child *, int pid, int trace)" 4
1837 .IX Item "ev_child_set (ev_child *, int pid, int trace)"
1838 .PD
1839 Configures the watcher to wait for status changes of process \f(CW\*(C`pid\*(C'\fR (or
1840 \&\fIany\fR process if \f(CW\*(C`pid\*(C'\fR is specified as \f(CW0\fR). The callback can look
1841 at the \f(CW\*(C`rstatus\*(C'\fR member of the \f(CW\*(C`ev_child\*(C'\fR watcher structure to see
1842 the status word (use the macros from \f(CW\*(C`sys/wait.h\*(C'\fR and see your systems
1843 \&\f(CW\*(C`waitpid\*(C'\fR documentation). The \f(CW\*(C`rpid\*(C'\fR member contains the pid of the
1844 process causing the status change. \f(CW\*(C`trace\*(C'\fR must be either \f(CW0\fR (only
1845 activate the watcher when the process terminates) or \f(CW1\fR (additionally
1846 activate the watcher when the process is stopped or continued).
1847 .IP "int pid [read\-only]" 4
1848 .IX Item "int pid [read-only]"
1849 The process id this watcher watches out for, or \f(CW0\fR, meaning any process id.
1850 .IP "int rpid [read\-write]" 4
1851 .IX Item "int rpid [read-write]"
1852 The process id that detected a status change.
1853 .IP "int rstatus [read\-write]" 4
1854 .IX Item "int rstatus [read-write]"
1855 The process exit/trace status caused by \f(CW\*(C`rpid\*(C'\fR (see your systems
1856 \&\f(CW\*(C`waitpid\*(C'\fR and \f(CW\*(C`sys/wait.h\*(C'\fR documentation for details).
1857 .PP
1858 \fIExamples\fR
1859 .IX Subsection "Examples"
1860 .PP
1861 Example: \f(CW\*(C`fork()\*(C'\fR a new process and install a child handler to wait for
1862 its completion.
1863 .PP
1864 .Vb 1
1865 \& ev_child cw;
1866 \&
1867 \& static void
1868 \& child_cb (EV_P_ struct ev_child *w, int revents)
1869 \& {
1870 \& ev_child_stop (EV_A_ w);
1871 \& printf ("process %d exited with status %x\en", w\->rpid, w\->rstatus);
1872 \& }
1873 \&
1874 \& pid_t pid = fork ();
1875 \&
1876 \& if (pid < 0)
1877 \& // error
1878 \& else if (pid == 0)
1879 \& {
1880 \& // the forked child executes here
1881 \& exit (1);
1882 \& }
1883 \& else
1884 \& {
1885 \& ev_child_init (&cw, child_cb, pid, 0);
1886 \& ev_child_start (EV_DEFAULT_ &cw);
1887 \& }
1888 .Ve
1889 .ie n .Sh """ev_stat"" \- did the file attributes just change?"
1890 .el .Sh "\f(CWev_stat\fP \- did the file attributes just change?"
1891 .IX Subsection "ev_stat - did the file attributes just change?"
1892 This watches a file system path for attribute changes. That is, it calls
1893 \&\f(CW\*(C`stat\*(C'\fR regularly (or when the \s-1OS\s0 says it changed) and sees if it changed
1894 compared to the last time, invoking the callback if it did.
1895 .PP
1896 The path does not need to exist: changing from \*(L"path exists\*(R" to \*(L"path does
1897 not exist\*(R" is a status change like any other. The condition \*(L"path does
1898 not exist\*(R" is signified by the \f(CW\*(C`st_nlink\*(C'\fR field being zero (which is
1899 otherwise always forced to be at least one) and all the other fields of
1900 the stat buffer having unspecified contents.
1901 .PP
1902 The path \fIshould\fR be absolute and \fImust not\fR end in a slash. If it is
1903 relative and your working directory changes, the behaviour is undefined.
1904 .PP
1905 Since there is no standard kernel interface to do this, the portable
1906 implementation simply calls \f(CW\*(C`stat (2)\*(C'\fR regularly on the path to see if
1907 it changed somehow. You can specify a recommended polling interval for
1908 this case. If you specify a polling interval of \f(CW0\fR (highly recommended!)
1909 then a \fIsuitable, unspecified default\fR value will be used (which
1910 you can expect to be around five seconds, although this might change
1911 dynamically). Libev will also impose a minimum interval which is currently
1912 around \f(CW0.1\fR, but thats usually overkill.
1913 .PP
1914 This watcher type is not meant for massive numbers of stat watchers,
1915 as even with OS-supported change notifications, this can be
1916 resource-intensive.
1917 .PP
1918 At the time of this writing, the only OS-specific interface implemented
1919 is the Linux inotify interface (implementing kqueue support is left as
1920 an exercise for the reader. Note, however, that the author sees no way
1921 of implementing \f(CW\*(C`ev_stat\*(C'\fR semantics with kqueue).
1922 .PP
1923 \fI\s-1ABI\s0 Issues (Largefile Support)\fR
1924 .IX Subsection "ABI Issues (Largefile Support)"
1925 .PP
1926 Libev by default (unless the user overrides this) uses the default
1927 compilation environment, which means that on systems with large file
1928 support disabled by default, you get the 32 bit version of the stat
1929 structure. When using the library from programs that change the \s-1ABI\s0 to
1930 use 64 bit file offsets the programs will fail. In that case you have to
1931 compile libev with the same flags to get binary compatibility. This is
1932 obviously the case with any flags that change the \s-1ABI\s0, but the problem is
1933 most noticeably disabled with ev_stat and large file support.
1934 .PP
1935 The solution for this is to lobby your distribution maker to make large
1936 file interfaces available by default (as e.g. FreeBSD does) and not
1937 optional. Libev cannot simply switch on large file support because it has
1938 to exchange stat structures with application programs compiled using the
1939 default compilation environment.
1940 .PP
1941 \fIInotify and Kqueue\fR
1942 .IX Subsection "Inotify and Kqueue"
1943 .PP
1944 When \f(CW\*(C`inotify (7)\*(C'\fR support has been compiled into libev (generally
1945 only available with Linux 2.6.25 or above due to bugs in earlier
1946 implementations) and present at runtime, it will be used to speed up
1947 change detection where possible. The inotify descriptor will be created
1948 lazily when the first \f(CW\*(C`ev_stat\*(C'\fR watcher is being started.
1949 .PP
1950 Inotify presence does not change the semantics of \f(CW\*(C`ev_stat\*(C'\fR watchers
1951 except that changes might be detected earlier, and in some cases, to avoid
1952 making regular \f(CW\*(C`stat\*(C'\fR calls. Even in the presence of inotify support
1953 there are many cases where libev has to resort to regular \f(CW\*(C`stat\*(C'\fR polling,
1954 but as long as the path exists, libev usually gets away without polling.
1955 .PP
1956 There is no support for kqueue, as apparently it cannot be used to
1957 implement this functionality, due to the requirement of having a file
1958 descriptor open on the object at all times, and detecting renames, unlinks
1959 etc. is difficult.
1960 .PP
1961 \fIThe special problem of stat time resolution\fR
1962 .IX Subsection "The special problem of stat time resolution"
1963 .PP
1964 The \f(CW\*(C`stat ()\*(C'\fR system call only supports full-second resolution portably, and
1965 even on systems where the resolution is higher, most file systems still
1966 only support whole seconds.
1967 .PP
1968 That means that, if the time is the only thing that changes, you can
1969 easily miss updates: on the first update, \f(CW\*(C`ev_stat\*(C'\fR detects a change and
1970 calls your callback, which does something. When there is another update
1971 within the same second, \f(CW\*(C`ev_stat\*(C'\fR will be unable to detect unless the
1972 stat data does change in other ways (e.g. file size).
1973 .PP
1974 The solution to this is to delay acting on a change for slightly more
1975 than a second (or till slightly after the next full second boundary), using
1976 a roughly one-second-delay \f(CW\*(C`ev_timer\*(C'\fR (e.g. \f(CW\*(C`ev_timer_set (w, 0., 1.02);
1977 ev_timer_again (loop, w)\*(C'\fR).
1978 .PP
1979 The \f(CW.02\fR offset is added to work around small timing inconsistencies
1980 of some operating systems (where the second counter of the current time
1981 might be be delayed. One such system is the Linux kernel, where a call to
1982 \&\f(CW\*(C`gettimeofday\*(C'\fR might return a timestamp with a full second later than
1983 a subsequent \f(CW\*(C`time\*(C'\fR call \- if the equivalent of \f(CW\*(C`time ()\*(C'\fR is used to
1984 update file times then there will be a small window where the kernel uses
1985 the previous second to update file times but libev might already execute
1986 the timer callback).
1987 .PP
1988 \fIWatcher-Specific Functions and Data Members\fR
1989 .IX Subsection "Watcher-Specific Functions and Data Members"
1990 .IP "ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)" 4
1991 .IX Item "ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)"
1992 .PD 0
1993 .IP "ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)" 4
1994 .IX Item "ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)"
1995 .PD
1996 Configures the watcher to wait for status changes of the given
1997 \&\f(CW\*(C`path\*(C'\fR. The \f(CW\*(C`interval\*(C'\fR is a hint on how quickly a change is expected to
1998 be detected and should normally be specified as \f(CW0\fR to let libev choose
1999 a suitable value. The memory pointed to by \f(CW\*(C`path\*(C'\fR must point to the same
2000 path for as long as the watcher is active.
2001 .Sp
2002 The callback will receive an \f(CW\*(C`EV_STAT\*(C'\fR event when a change was detected,
2003 relative to the attributes at the time the watcher was started (or the
2004 last change was detected).
2005 .IP "ev_stat_stat (loop, ev_stat *)" 4
2006 .IX Item "ev_stat_stat (loop, ev_stat *)"
2007 Updates the stat buffer immediately with new values. If you change the
2008 watched path in your callback, you could call this function to avoid
2009 detecting this change (while introducing a race condition if you are not
2010 the only one changing the path). Can also be useful simply to find out the
2011 new values.
2012 .IP "ev_statdata attr [read\-only]" 4
2013 .IX Item "ev_statdata attr [read-only]"
2014 The most-recently detected attributes of the file. Although the type is
2015 \&\f(CW\*(C`ev_statdata\*(C'\fR, this is usually the (or one of the) \f(CW\*(C`struct stat\*(C'\fR types
2016 suitable for your system, but you can only rely on the POSIX-standardised
2017 members to be present. If the \f(CW\*(C`st_nlink\*(C'\fR member is \f(CW0\fR, then there was
2018 some error while \f(CW\*(C`stat\*(C'\fRing the file.
2019 .IP "ev_statdata prev [read\-only]" 4
2020 .IX Item "ev_statdata prev [read-only]"
2021 The previous attributes of the file. The callback gets invoked whenever
2022 \&\f(CW\*(C`prev\*(C'\fR != \f(CW\*(C`attr\*(C'\fR, or, more precisely, one or more of these members
2023 differ: \f(CW\*(C`st_dev\*(C'\fR, \f(CW\*(C`st_ino\*(C'\fR, \f(CW\*(C`st_mode\*(C'\fR, \f(CW\*(C`st_nlink\*(C'\fR, \f(CW\*(C`st_uid\*(C'\fR,
2024 \&\f(CW\*(C`st_gid\*(C'\fR, \f(CW\*(C`st_rdev\*(C'\fR, \f(CW\*(C`st_size\*(C'\fR, \f(CW\*(C`st_atime\*(C'\fR, \f(CW\*(C`st_mtime\*(C'\fR, \f(CW\*(C`st_ctime\*(C'\fR.
2025 .IP "ev_tstamp interval [read\-only]" 4
2026 .IX Item "ev_tstamp interval [read-only]"
2027 The specified interval.
2028 .IP "const char *path [read\-only]" 4
2029 .IX Item "const char *path [read-only]"
2030 The file system path that is being watched.
2031 .PP
2032 \fIExamples\fR
2033 .IX Subsection "Examples"
2034 .PP
2035 Example: Watch \f(CW\*(C`/etc/passwd\*(C'\fR for attribute changes.
2036 .PP
2037 .Vb 10
2038 \& static void
2039 \& passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
2040 \& {
2041 \& /* /etc/passwd changed in some way */
2042 \& if (w\->attr.st_nlink)
2043 \& {
2044 \& printf ("passwd current size %ld\en", (long)w\->attr.st_size);
2045 \& printf ("passwd current atime %ld\en", (long)w\->attr.st_mtime);
2046 \& printf ("passwd current mtime %ld\en", (long)w\->attr.st_mtime);
2047 \& }
2048 \& else
2049 \& /* you shalt not abuse printf for puts */
2050 \& puts ("wow, /etc/passwd is not there, expect problems. "
2051 \& "if this is windows, they already arrived\en");
2052 \& }
2053 \&
2054 \& ...
2055 \& ev_stat passwd;
2056 \&
2057 \& ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
2058 \& ev_stat_start (loop, &passwd);
2059 .Ve
2060 .PP
2061 Example: Like above, but additionally use a one-second delay so we do not
2062 miss updates (however, frequent updates will delay processing, too, so
2063 one might do the work both on \f(CW\*(C`ev_stat\*(C'\fR callback invocation \fIand\fR on
2064 \&\f(CW\*(C`ev_timer\*(C'\fR callback invocation).
2065 .PP
2066 .Vb 2
2067 \& static ev_stat passwd;
2068 \& static ev_timer timer;
2069 \&
2070 \& static void
2071 \& timer_cb (EV_P_ ev_timer *w, int revents)
2072 \& {
2073 \& ev_timer_stop (EV_A_ w);
2074 \&
2075 \& /* now it\*(Aqs one second after the most recent passwd change */
2076 \& }
2077 \&
2078 \& static void
2079 \& stat_cb (EV_P_ ev_stat *w, int revents)
2080 \& {
2081 \& /* reset the one\-second timer */
2082 \& ev_timer_again (EV_A_ &timer);
2083 \& }
2084 \&
2085 \& ...
2086 \& ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
2087 \& ev_stat_start (loop, &passwd);
2088 \& ev_timer_init (&timer, timer_cb, 0., 1.02);
2089 .Ve
2090 .ie n .Sh """ev_idle"" \- when you've got nothing better to do..."
2091 .el .Sh "\f(CWev_idle\fP \- when you've got nothing better to do..."
2092 .IX Subsection "ev_idle - when you've got nothing better to do..."
2093 Idle watchers trigger events when no other events of the same or higher
2094 priority are pending (prepare, check and other idle watchers do not count
2095 as receiving \*(L"events\*(R").
2096 .PP
2097 That is, as long as your process is busy handling sockets or timeouts
2098 (or even signals, imagine) of the same or higher priority it will not be
2099 triggered. But when your process is idle (or only lower-priority watchers
2100 are pending), the idle watchers are being called once per event loop
2101 iteration \- until stopped, that is, or your process receives more events
2102 and becomes busy again with higher priority stuff.
2103 .PP
2104 The most noteworthy effect is that as long as any idle watchers are
2105 active, the process will not block when waiting for new events.
2106 .PP
2107 Apart from keeping your process non-blocking (which is a useful
2108 effect on its own sometimes), idle watchers are a good place to do
2109 \&\*(L"pseudo-background processing\*(R", or delay processing stuff to after the
2110 event loop has handled all outstanding events.
2111 .PP
2112 \fIWatcher-Specific Functions and Data Members\fR
2113 .IX Subsection "Watcher-Specific Functions and Data Members"
2114 .IP "ev_idle_init (ev_signal *, callback)" 4
2115 .IX Item "ev_idle_init (ev_signal *, callback)"
2116 Initialises and configures the idle watcher \- it has no parameters of any
2117 kind. There is a \f(CW\*(C`ev_idle_set\*(C'\fR macro, but using it is utterly pointless,
2118 believe me.
2119 .PP
2120 \fIExamples\fR
2121 .IX Subsection "Examples"
2122 .PP
2123 Example: Dynamically allocate an \f(CW\*(C`ev_idle\*(C'\fR watcher, start it, and in the
2124 callback, free it. Also, use no error checking, as usual.
2125 .PP
2126 .Vb 7
2127 \& static void
2128 \& idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)
2129 \& {
2130 \& free (w);
2131 \& // now do something you wanted to do when the program has
2132 \& // no longer anything immediate to do.
2133 \& }
2134 \&
2135 \& struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));
2136 \& ev_idle_init (idle_watcher, idle_cb);
2137 \& ev_idle_start (loop, idle_cb);
2138 .Ve
2139 .ie n .Sh """ev_prepare""\fP and \f(CW""ev_check"" \- customise your event loop!"
2140 .el .Sh "\f(CWev_prepare\fP and \f(CWev_check\fP \- customise your event loop!"
2141 .IX Subsection "ev_prepare and ev_check - customise your event loop!"
2142 Prepare and check watchers are usually (but not always) used in pairs:
2143 prepare watchers get invoked before the process blocks and check watchers
2144 afterwards.
2145 .PP
2146 You \fImust not\fR call \f(CW\*(C`ev_loop\*(C'\fR or similar functions that enter
2147 the current event loop from either \f(CW\*(C`ev_prepare\*(C'\fR or \f(CW\*(C`ev_check\*(C'\fR
2148 watchers. Other loops than the current one are fine, however. The
2149 rationale behind this is that you do not need to check for recursion in
2150 those watchers, i.e. the sequence will always be \f(CW\*(C`ev_prepare\*(C'\fR, blocking,
2151 \&\f(CW\*(C`ev_check\*(C'\fR so if you have one watcher of each kind they will always be
2152 called in pairs bracketing the blocking call.
2153 .PP
2154 Their main purpose is to integrate other event mechanisms into libev and
2155 their use is somewhat advanced. They could be used, for example, to track
2156 variable changes, implement your own watchers, integrate net-snmp or a
2157 coroutine library and lots more. They are also occasionally useful if
2158 you cache some data and want to flush it before blocking (for example,
2159 in X programs you might want to do an \f(CW\*(C`XFlush ()\*(C'\fR in an \f(CW\*(C`ev_prepare\*(C'\fR
2160 watcher).
2161 .PP
2162 This is done by examining in each prepare call which file descriptors
2163 need to be watched by the other library, registering \f(CW\*(C`ev_io\*(C'\fR watchers
2164 for them and starting an \f(CW\*(C`ev_timer\*(C'\fR watcher for any timeouts (many
2165 libraries provide exactly this functionality). Then, in the check watcher,
2166 you check for any events that occurred (by checking the pending status
2167 of all watchers and stopping them) and call back into the library. The
2168 I/O and timer callbacks will never actually be called (but must be valid
2169 nevertheless, because you never know, you know?).
2170 .PP
2171 As another example, the Perl Coro module uses these hooks to integrate
2172 coroutines into libev programs, by yielding to other active coroutines
2173 during each prepare and only letting the process block if no coroutines
2174 are ready to run (it's actually more complicated: it only runs coroutines
2175 with priority higher than or equal to the event loop and one coroutine
2176 of lower priority, but only once, using idle watchers to keep the event
2177 loop from blocking if lower-priority coroutines are active, thus mapping
2178 low-priority coroutines to idle/background tasks).
2179 .PP
2180 It is recommended to give \f(CW\*(C`ev_check\*(C'\fR watchers highest (\f(CW\*(C`EV_MAXPRI\*(C'\fR)
2181 priority, to ensure that they are being run before any other watchers
2182 after the poll (this doesn't matter for \f(CW\*(C`ev_prepare\*(C'\fR watchers).
2183 .PP
2184 Also, \f(CW\*(C`ev_check\*(C'\fR watchers (and \f(CW\*(C`ev_prepare\*(C'\fR watchers, too) should not
2185 activate (\*(L"feed\*(R") events into libev. While libev fully supports this, they
2186 might get executed before other \f(CW\*(C`ev_check\*(C'\fR watchers did their job. As
2187 \&\f(CW\*(C`ev_check\*(C'\fR watchers are often used to embed other (non-libev) event
2188 loops those other event loops might be in an unusable state until their
2189 \&\f(CW\*(C`ev_check\*(C'\fR watcher ran (always remind yourself to coexist peacefully with
2190 others).
2191 .PP
2192 \fIWatcher-Specific Functions and Data Members\fR
2193 .IX Subsection "Watcher-Specific Functions and Data Members"
2194 .IP "ev_prepare_init (ev_prepare *, callback)" 4
2195 .IX Item "ev_prepare_init (ev_prepare *, callback)"
2196 .PD 0
2197 .IP "ev_check_init (ev_check *, callback)" 4
2198 .IX Item "ev_check_init (ev_check *, callback)"
2199 .PD
2200 Initialises and configures the prepare or check watcher \- they have no
2201 parameters of any kind. There are \f(CW\*(C`ev_prepare_set\*(C'\fR and \f(CW\*(C`ev_check_set\*(C'\fR
2202 macros, but using them is utterly, utterly, utterly and completely
2203 pointless.
2204 .PP
2205 \fIExamples\fR
2206 .IX Subsection "Examples"
2207 .PP
2208 There are a number of principal ways to embed other event loops or modules
2209 into libev. Here are some ideas on how to include libadns into libev
2210 (there is a Perl module named \f(CW\*(C`EV::ADNS\*(C'\fR that does this, which you could
2211 use as a working example. Another Perl module named \f(CW\*(C`EV::Glib\*(C'\fR embeds a
2212 Glib main context into libev, and finally, \f(CW\*(C`Glib::EV\*(C'\fR embeds \s-1EV\s0 into the
2213 Glib event loop).
2214 .PP
2215 Method 1: Add \s-1IO\s0 watchers and a timeout watcher in a prepare handler,
2216 and in a check watcher, destroy them and call into libadns. What follows
2217 is pseudo-code only of course. This requires you to either use a low
2218 priority for the check watcher or use \f(CW\*(C`ev_clear_pending\*(C'\fR explicitly, as
2219 the callbacks for the IO/timeout watchers might not have been called yet.
2220 .PP
2221 .Vb 2
2222 \& static ev_io iow [nfd];
2223 \& static ev_timer tw;
2224 \&
2225 \& static void
2226 \& io_cb (ev_loop *loop, ev_io *w, int revents)
2227 \& {
2228 \& }
2229 \&
2230 \& // create io watchers for each fd and a timer before blocking
2231 \& static void
2232 \& adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
2233 \& {
2234 \& int timeout = 3600000;
2235 \& struct pollfd fds [nfd];
2236 \& // actual code will need to loop here and realloc etc.
2237 \& adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
2238 \&
2239 \& /* the callback is illegal, but won\*(Aqt be called as we stop during check */
2240 \& ev_timer_init (&tw, 0, timeout * 1e\-3);
2241 \& ev_timer_start (loop, &tw);
2242 \&
2243 \& // create one ev_io per pollfd
2244 \& for (int i = 0; i < nfd; ++i)
2245 \& {
2246 \& ev_io_init (iow + i, io_cb, fds [i].fd,
2247 \& ((fds [i].events & POLLIN ? EV_READ : 0)
2248 \& | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
2249 \&
2250 \& fds [i].revents = 0;
2251 \& ev_io_start (loop, iow + i);
2252 \& }
2253 \& }
2254 \&
2255 \& // stop all watchers after blocking
2256 \& static void
2257 \& adns_check_cb (ev_loop *loop, ev_check *w, int revents)
2258 \& {
2259 \& ev_timer_stop (loop, &tw);
2260 \&
2261 \& for (int i = 0; i < nfd; ++i)
2262 \& {
2263 \& // set the relevant poll flags
2264 \& // could also call adns_processreadable etc. here
2265 \& struct pollfd *fd = fds + i;
2266 \& int revents = ev_clear_pending (iow + i);
2267 \& if (revents & EV_READ ) fd\->revents |= fd\->events & POLLIN;
2268 \& if (revents & EV_WRITE) fd\->revents |= fd\->events & POLLOUT;
2269 \&
2270 \& // now stop the watcher
2271 \& ev_io_stop (loop, iow + i);
2272 \& }
2273 \&
2274 \& adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
2275 \& }
2276 .Ve
2277 .PP
2278 Method 2: This would be just like method 1, but you run \f(CW\*(C`adns_afterpoll\*(C'\fR
2279 in the prepare watcher and would dispose of the check watcher.
2280 .PP
2281 Method 3: If the module to be embedded supports explicit event
2282 notification (libadns does), you can also make use of the actual watcher
2283 callbacks, and only destroy/create the watchers in the prepare watcher.
2284 .PP
2285 .Vb 5
2286 \& static void
2287 \& timer_cb (EV_P_ ev_timer *w, int revents)
2288 \& {
2289 \& adns_state ads = (adns_state)w\->data;
2290 \& update_now (EV_A);
2291 \&
2292 \& adns_processtimeouts (ads, &tv_now);
2293 \& }
2294 \&
2295 \& static void
2296 \& io_cb (EV_P_ ev_io *w, int revents)
2297 \& {
2298 \& adns_state ads = (adns_state)w\->data;
2299 \& update_now (EV_A);
2300 \&
2301 \& if (revents & EV_READ ) adns_processreadable (ads, w\->fd, &tv_now);
2302 \& if (revents & EV_WRITE) adns_processwriteable (ads, w\->fd, &tv_now);
2303 \& }
2304 \&
2305 \& // do not ever call adns_afterpoll
2306 .Ve
2307 .PP
2308 Method 4: Do not use a prepare or check watcher because the module you
2309 want to embed is not flexible enough to support it. Instead, you can
2310 override their poll function. The drawback with this solution is that the
2311 main loop is now no longer controllable by \s-1EV\s0. The \f(CW\*(C`Glib::EV\*(C'\fR module uses
2312 this approach, effectively embedding \s-1EV\s0 as a client into the horrible
2313 libglib event loop.
2314 .PP
2315 .Vb 4
2316 \& static gint
2317 \& event_poll_func (GPollFD *fds, guint nfds, gint timeout)
2318 \& {
2319 \& int got_events = 0;
2320 \&
2321 \& for (n = 0; n < nfds; ++n)
2322 \& // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
2323 \&
2324 \& if (timeout >= 0)
2325 \& // create/start timer
2326 \&
2327 \& // poll
2328 \& ev_loop (EV_A_ 0);
2329 \&
2330 \& // stop timer again
2331 \& if (timeout >= 0)
2332 \& ev_timer_stop (EV_A_ &to);
2333 \&
2334 \& // stop io watchers again \- their callbacks should have set
2335 \& for (n = 0; n < nfds; ++n)
2336 \& ev_io_stop (EV_A_ iow [n]);
2337 \&
2338 \& return got_events;
2339 \& }
2340 .Ve
2341 .ie n .Sh """ev_embed"" \- when one backend isn't enough..."
2342 .el .Sh "\f(CWev_embed\fP \- when one backend isn't enough..."
2343 .IX Subsection "ev_embed - when one backend isn't enough..."
2344 This is a rather advanced watcher type that lets you embed one event loop
2345 into another (currently only \f(CW\*(C`ev_io\*(C'\fR events are supported in the embedded
2346 loop, other types of watchers might be handled in a delayed or incorrect
2347 fashion and must not be used).
2348 .PP
2349 There are primarily two reasons you would want that: work around bugs and
2350 prioritise I/O.
2351 .PP
2352 As an example for a bug workaround, the kqueue backend might only support
2353 sockets on some platform, so it is unusable as generic backend, but you
2354 still want to make use of it because you have many sockets and it scales
2355 so nicely. In this case, you would create a kqueue-based loop and embed
2356 it into your default loop (which might use e.g. poll). Overall operation
2357 will be a bit slower because first libev has to call \f(CW\*(C`poll\*(C'\fR and then
2358 \&\f(CW\*(C`kevent\*(C'\fR, but at least you can use both mechanisms for what they are
2359 best: \f(CW\*(C`kqueue\*(C'\fR for scalable sockets and \f(CW\*(C`poll\*(C'\fR if you want it to work :)
2360 .PP
2361 As for prioritising I/O: under rare circumstances you have the case where
2362 some fds have to be watched and handled very quickly (with low latency),
2363 and even priorities and idle watchers might have too much overhead. In
2364 this case you would put all the high priority stuff in one loop and all
2365 the rest in a second one, and embed the second one in the first.
2366 .PP
2367 As long as the watcher is active, the callback will be invoked every time
2368 there might be events pending in the embedded loop. The callback must then
2369 call \f(CW\*(C`ev_embed_sweep (mainloop, watcher)\*(C'\fR to make a single sweep and invoke
2370 their callbacks (you could also start an idle watcher to give the embedded
2371 loop strictly lower priority for example). You can also set the callback
2372 to \f(CW0\fR, in which case the embed watcher will automatically execute the
2373 embedded loop sweep.
2374 .PP
2375 As long as the watcher is started it will automatically handle events. The
2376 callback will be invoked whenever some events have been handled. You can
2377 set the callback to \f(CW0\fR to avoid having to specify one if you are not
2378 interested in that.
2379 .PP
2380 Also, there have not currently been made special provisions for forking:
2381 when you fork, you not only have to call \f(CW\*(C`ev_loop_fork\*(C'\fR on both loops,
2382 but you will also have to stop and restart any \f(CW\*(C`ev_embed\*(C'\fR watchers
2383 yourself \- but you can use a fork watcher to handle this automatically,
2384 and future versions of libev might do just that.
2385 .PP
2386 Unfortunately, not all backends are embeddable: only the ones returned by
2387 \&\f(CW\*(C`ev_embeddable_backends\*(C'\fR are, which, unfortunately, does not include any
2388 portable one.
2389 .PP
2390 So when you want to use this feature you will always have to be prepared
2391 that you cannot get an embeddable loop. The recommended way to get around
2392 this is to have a separate variables for your embeddable loop, try to
2393 create it, and if that fails, use the normal loop for everything.
2394 .PP
2395 \fI\f(CI\*(C`ev_embed\*(C'\fI and fork\fR
2396 .IX Subsection "ev_embed and fork"
2397 .PP
2398 While the \f(CW\*(C`ev_embed\*(C'\fR watcher is running, forks in the embedding loop will
2399 automatically be applied to the embedded loop as well, so no special
2400 fork handling is required in that case. When the watcher is not running,
2401 however, it is still the task of the libev user to call \f(CW\*(C`ev_loop_fork ()\*(C'\fR
2402 as applicable.
2403 .PP
2404 \fIWatcher-Specific Functions and Data Members\fR
2405 .IX Subsection "Watcher-Specific Functions and Data Members"
2406 .IP "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)" 4
2407 .IX Item "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)"
2408 .PD 0
2409 .IP "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)" 4
2410 .IX Item "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)"
2411 .PD
2412 Configures the watcher to embed the given loop, which must be
2413 embeddable. If the callback is \f(CW0\fR, then \f(CW\*(C`ev_embed_sweep\*(C'\fR will be
2414 invoked automatically, otherwise it is the responsibility of the callback
2415 to invoke it (it will continue to be called until the sweep has been done,
2416 if you do not want that, you need to temporarily stop the embed watcher).
2417 .IP "ev_embed_sweep (loop, ev_embed *)" 4
2418 .IX Item "ev_embed_sweep (loop, ev_embed *)"
2419 Make a single, non-blocking sweep over the embedded loop. This works
2420 similarly to \f(CW\*(C`ev_loop (embedded_loop, EVLOOP_NONBLOCK)\*(C'\fR, but in the most
2421 appropriate way for embedded loops.
2422 .IP "struct ev_loop *other [read\-only]" 4
2423 .IX Item "struct ev_loop *other [read-only]"
2424 The embedded event loop.
2425 .PP
2426 \fIExamples\fR
2427 .IX Subsection "Examples"
2428 .PP
2429 Example: Try to get an embeddable event loop and embed it into the default
2430 event loop. If that is not possible, use the default loop. The default
2431 loop is stored in \f(CW\*(C`loop_hi\*(C'\fR, while the embeddable loop is stored in
2432 \&\f(CW\*(C`loop_lo\*(C'\fR (which is \f(CW\*(C`loop_hi\*(C'\fR in the case no embeddable loop can be
2433 used).
2434 .PP
2435 .Vb 3
2436 \& struct ev_loop *loop_hi = ev_default_init (0);
2437 \& struct ev_loop *loop_lo = 0;
2438 \& struct ev_embed embed;
2439 \&
2440 \& // see if there is a chance of getting one that works
2441 \& // (remember that a flags value of 0 means autodetection)
2442 \& loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
2443 \& ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
2444 \& : 0;
2445 \&
2446 \& // if we got one, then embed it, otherwise default to loop_hi
2447 \& if (loop_lo)
2448 \& {
2449 \& ev_embed_init (&embed, 0, loop_lo);
2450 \& ev_embed_start (loop_hi, &embed);
2451 \& }
2452 \& else
2453 \& loop_lo = loop_hi;
2454 .Ve
2455 .PP
2456 Example: Check if kqueue is available but not recommended and create
2457 a kqueue backend for use with sockets (which usually work with any
2458 kqueue implementation). Store the kqueue/socket\-only event loop in
2459 \&\f(CW\*(C`loop_socket\*(C'\fR. (One might optionally use \f(CW\*(C`EVFLAG_NOENV\*(C'\fR, too).
2460 .PP
2461 .Vb 3
2462 \& struct ev_loop *loop = ev_default_init (0);
2463 \& struct ev_loop *loop_socket = 0;
2464 \& struct ev_embed embed;
2465 \&
2466 \& if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2467 \& if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2468 \& {
2469 \& ev_embed_init (&embed, 0, loop_socket);
2470 \& ev_embed_start (loop, &embed);
2471 \& }
2472 \&
2473 \& if (!loop_socket)
2474 \& loop_socket = loop;
2475 \&
2476 \& // now use loop_socket for all sockets, and loop for everything else
2477 .Ve
2478 .ie n .Sh """ev_fork"" \- the audacity to resume the event loop after a fork"
2479 .el .Sh "\f(CWev_fork\fP \- the audacity to resume the event loop after a fork"
2480 .IX Subsection "ev_fork - the audacity to resume the event loop after a fork"
2481 Fork watchers are called when a \f(CW\*(C`fork ()\*(C'\fR was detected (usually because
2482 whoever is a good citizen cared to tell libev about it by calling
2483 \&\f(CW\*(C`ev_default_fork\*(C'\fR or \f(CW\*(C`ev_loop_fork\*(C'\fR). The invocation is done before the
2484 event loop blocks next and before \f(CW\*(C`ev_check\*(C'\fR watchers are being called,
2485 and only in the child after the fork. If whoever good citizen calling
2486 \&\f(CW\*(C`ev_default_fork\*(C'\fR cheats and calls it in the wrong process, the fork
2487 handlers will be invoked, too, of course.
2488 .PP
2489 \fIWatcher-Specific Functions and Data Members\fR
2490 .IX Subsection "Watcher-Specific Functions and Data Members"
2491 .IP "ev_fork_init (ev_signal *, callback)" 4
2492 .IX Item "ev_fork_init (ev_signal *, callback)"
2493 Initialises and configures the fork watcher \- it has no parameters of any
2494 kind. There is a \f(CW\*(C`ev_fork_set\*(C'\fR macro, but using it is utterly pointless,
2495 believe me.
2496 .ie n .Sh """ev_async"" \- how to wake up another event loop"
2497 .el .Sh "\f(CWev_async\fP \- how to wake up another event loop"
2498 .IX Subsection "ev_async - how to wake up another event loop"
2499 In general, you cannot use an \f(CW\*(C`ev_loop\*(C'\fR from multiple threads or other
2500 asynchronous sources such as signal handlers (as opposed to multiple event
2501 loops \- those are of course safe to use in different threads).
2502 .PP
2503 Sometimes, however, you need to wake up another event loop you do not
2504 control, for example because it belongs to another thread. This is what
2505 \&\f(CW\*(C`ev_async\*(C'\fR watchers do: as long as the \f(CW\*(C`ev_async\*(C'\fR watcher is active, you
2506 can signal it by calling \f(CW\*(C`ev_async_send\*(C'\fR, which is thread\- and signal
2507 safe.
2508 .PP
2509 This functionality is very similar to \f(CW\*(C`ev_signal\*(C'\fR watchers, as signals,
2510 too, are asynchronous in nature, and signals, too, will be compressed
2511 (i.e. the number of callback invocations may be less than the number of
2512 \&\f(CW\*(C`ev_async_sent\*(C'\fR calls).
2513 .PP
2514 Unlike \f(CW\*(C`ev_signal\*(C'\fR watchers, \f(CW\*(C`ev_async\*(C'\fR works with any event loop, not
2515 just the default loop.
2516 .PP
2517 \fIQueueing\fR
2518 .IX Subsection "Queueing"
2519 .PP
2520 \&\f(CW\*(C`ev_async\*(C'\fR does not support queueing of data in any way. The reason
2521 is that the author does not know of a simple (or any) algorithm for a
2522 multiple-writer-single-reader queue that works in all cases and doesn't
2523 need elaborate support such as pthreads.
2524 .PP
2525 That means that if you want to queue data, you have to provide your own
2526 queue. But at least I can tell you how to implement locking around your
2527 queue:
2528 .IP "queueing from a signal handler context" 4
2529 .IX Item "queueing from a signal handler context"
2530 To implement race-free queueing, you simply add to the queue in the signal
2531 handler but you block the signal handler in the watcher callback. Here is
2532 an example that does that for some fictitious \s-1SIGUSR1\s0 handler:
2533 .Sp
2534 .Vb 1
2535 \& static ev_async mysig;
2536 \&
2537 \& static void
2538 \& sigusr1_handler (void)
2539 \& {
2540 \& sometype data;
2541 \&
2542 \& // no locking etc.
2543 \& queue_put (data);
2544 \& ev_async_send (EV_DEFAULT_ &mysig);
2545 \& }
2546 \&
2547 \& static void
2548 \& mysig_cb (EV_P_ ev_async *w, int revents)
2549 \& {
2550 \& sometype data;
2551 \& sigset_t block, prev;
2552 \&
2553 \& sigemptyset (&block);
2554 \& sigaddset (&block, SIGUSR1);
2555 \& sigprocmask (SIG_BLOCK, &block, &prev);
2556 \&
2557 \& while (queue_get (&data))
2558 \& process (data);
2559 \&
2560 \& if (sigismember (&prev, SIGUSR1)
2561 \& sigprocmask (SIG_UNBLOCK, &block, 0);
2562 \& }
2563 .Ve
2564 .Sp
2565 (Note: pthreads in theory requires you to use \f(CW\*(C`pthread_setmask\*(C'\fR
2566 instead of \f(CW\*(C`sigprocmask\*(C'\fR when you use threads, but libev doesn't do it
2567 either...).
2568 .IP "queueing from a thread context" 4
2569 .IX Item "queueing from a thread context"
2570 The strategy for threads is different, as you cannot (easily) block
2571 threads but you can easily preempt them, so to queue safely you need to
2572 employ a traditional mutex lock, such as in this pthread example:
2573 .Sp
2574 .Vb 2
2575 \& static ev_async mysig;
2576 \& static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
2577 \&
2578 \& static void
2579 \& otherthread (void)
2580 \& {
2581 \& // only need to lock the actual queueing operation
2582 \& pthread_mutex_lock (&mymutex);
2583 \& queue_put (data);
2584 \& pthread_mutex_unlock (&mymutex);
2585 \&
2586 \& ev_async_send (EV_DEFAULT_ &mysig);
2587 \& }
2588 \&
2589 \& static void
2590 \& mysig_cb (EV_P_ ev_async *w, int revents)
2591 \& {
2592 \& pthread_mutex_lock (&mymutex);
2593 \&
2594 \& while (queue_get (&data))
2595 \& process (data);
2596 \&
2597 \& pthread_mutex_unlock (&mymutex);
2598 \& }
2599 .Ve
2600 .PP
2601 \fIWatcher-Specific Functions and Data Members\fR
2602 .IX Subsection "Watcher-Specific Functions and Data Members"
2603 .IP "ev_async_init (ev_async *, callback)" 4
2604 .IX Item "ev_async_init (ev_async *, callback)"
2605 Initialises and configures the async watcher \- it has no parameters of any
2606 kind. There is a \f(CW\*(C`ev_asynd_set\*(C'\fR macro, but using it is utterly pointless,
2607 trust me.
2608 .IP "ev_async_send (loop, ev_async *)" 4
2609 .IX Item "ev_async_send (loop, ev_async *)"
2610 Sends/signals/activates the given \f(CW\*(C`ev_async\*(C'\fR watcher, that is, feeds
2611 an \f(CW\*(C`EV_ASYNC\*(C'\fR event on the watcher into the event loop. Unlike
2612 \&\f(CW\*(C`ev_feed_event\*(C'\fR, this call is safe to do from other threads, signal or
2613 similar contexts (see the discussion of \f(CW\*(C`EV_ATOMIC_T\*(C'\fR in the embedding
2614 section below on what exactly this means).
2615 .Sp
2616 This call incurs the overhead of a system call only once per loop iteration,
2617 so while the overhead might be noticeable, it doesn't apply to repeated
2618 calls to \f(CW\*(C`ev_async_send\*(C'\fR.
2619 .IP "bool = ev_async_pending (ev_async *)" 4
2620 .IX Item "bool = ev_async_pending (ev_async *)"
2621 Returns a non-zero value when \f(CW\*(C`ev_async_send\*(C'\fR has been called on the
2622 watcher but the event has not yet been processed (or even noted) by the
2623 event loop.
2624 .Sp
2625 \&\f(CW\*(C`ev_async_send\*(C'\fR sets a flag in the watcher and wakes up the loop. When
2626 the loop iterates next and checks for the watcher to have become active,
2627 it will reset the flag again. \f(CW\*(C`ev_async_pending\*(C'\fR can be used to very
2628 quickly check whether invoking the loop might be a good idea.
2629 .Sp
2630 Not that this does \fInot\fR check whether the watcher itself is pending, only
2631 whether it has been requested to make this watcher pending.
2632 .SH "OTHER FUNCTIONS"
2633 .IX Header "OTHER FUNCTIONS"
2634 There are some other functions of possible interest. Described. Here. Now.
2635 .IP "ev_once (loop, int fd, int events, ev_tstamp timeout, callback)" 4
2636 .IX Item "ev_once (loop, int fd, int events, ev_tstamp timeout, callback)"
2637 This function combines a simple timer and an I/O watcher, calls your
2638 callback on whichever event happens first and automatically stops both
2639 watchers. This is useful if you want to wait for a single event on an fd
2640 or timeout without having to allocate/configure/start/stop/free one or
2641 more watchers yourself.
2642 .Sp
2643 If \f(CW\*(C`fd\*(C'\fR is less than 0, then no I/O watcher will be started and the
2644 \&\f(CW\*(C`events\*(C'\fR argument is being ignored. Otherwise, an \f(CW\*(C`ev_io\*(C'\fR watcher for
2645 the given \f(CW\*(C`fd\*(C'\fR and \f(CW\*(C`events\*(C'\fR set will be created and started.
2646 .Sp
2647 If \f(CW\*(C`timeout\*(C'\fR is less than 0, then no timeout watcher will be
2648 started. Otherwise an \f(CW\*(C`ev_timer\*(C'\fR watcher with after = \f(CW\*(C`timeout\*(C'\fR (and
2649 repeat = 0) will be started. \f(CW0\fR is a valid timeout.
2650 .Sp
2651 The callback has the type \f(CW\*(C`void (*cb)(int revents, void *arg)\*(C'\fR and gets
2652 passed an \f(CW\*(C`revents\*(C'\fR set like normal event callbacks (a combination of
2653 \&\f(CW\*(C`EV_ERROR\*(C'\fR, \f(CW\*(C`EV_READ\*(C'\fR, \f(CW\*(C`EV_WRITE\*(C'\fR or \f(CW\*(C`EV_TIMEOUT\*(C'\fR) and the \f(CW\*(C`arg\*(C'\fR
2654 value passed to \f(CW\*(C`ev_once\*(C'\fR. Note that it is possible to receive \fIboth\fR
2655 a timeout and an io event at the same time \- you probably should give io
2656 events precedence.
2657 .Sp
2658 Example: wait up to ten seconds for data to appear on \s-1STDIN_FILENO\s0.
2659 .Sp
2660 .Vb 7
2661 \& static void stdin_ready (int revents, void *arg)
2662 \& {
2663 \& if (revents & EV_READ)
2664 \& /* stdin might have data for us, joy! */;
2665 \& else if (revents & EV_TIMEOUT)
2666 \& /* doh, nothing entered */;
2667 \& }
2668 \&
2669 \& ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
2670 .Ve
2671 .IP "ev_feed_event (ev_loop *, watcher *, int revents)" 4
2672 .IX Item "ev_feed_event (ev_loop *, watcher *, int revents)"
2673 Feeds the given event set into the event loop, as if the specified event
2674 had happened for the specified watcher (which must be a pointer to an
2675 initialised but not necessarily started event watcher).
2676 .IP "ev_feed_fd_event (ev_loop *, int fd, int revents)" 4
2677 .IX Item "ev_feed_fd_event (ev_loop *, int fd, int revents)"
2678 Feed an event on the given fd, as if a file descriptor backend detected
2679 the given events it.
2680 .IP "ev_feed_signal_event (ev_loop *loop, int signum)" 4
2681 .IX Item "ev_feed_signal_event (ev_loop *loop, int signum)"
2682 Feed an event as if the given signal occurred (\f(CW\*(C`loop\*(C'\fR must be the default
2683 loop!).
2684 .SH "LIBEVENT EMULATION"
2685 .IX Header "LIBEVENT EMULATION"
2686 Libev offers a compatibility emulation layer for libevent. It cannot
2687 emulate the internals of libevent, so here are some usage hints:
2688 .IP "\(bu" 4
2689 Use it by including <event.h>, as usual.
2690 .IP "\(bu" 4
2691 The following members are fully supported: ev_base, ev_callback,
2692 ev_arg, ev_fd, ev_res, ev_events.
2693 .IP "\(bu" 4
2694 Avoid using ev_flags and the EVLIST_*\-macros, while it is
2695 maintained by libev, it does not work exactly the same way as in libevent (consider
2696 it a private \s-1API\s0).
2697 .IP "\(bu" 4
2698 Priorities are not currently supported. Initialising priorities
2699 will fail and all watchers will have the same priority, even though there
2700 is an ev_pri field.
2701 .IP "\(bu" 4
2702 In libevent, the last base created gets the signals, in libev, the
2703 first base created (== the default loop) gets the signals.
2704 .IP "\(bu" 4
2705 Other members are not supported.
2706 .IP "\(bu" 4
2707 The libev emulation is \fInot\fR \s-1ABI\s0 compatible to libevent, you need
2708 to use the libev header file and library.
2709 .SH "\*(C+ SUPPORT"
2710 .IX Header " SUPPORT"
2711 Libev comes with some simplistic wrapper classes for \*(C+ that mainly allow
2712 you to use some convenience methods to start/stop watchers and also change
2713 the callback model to a model using method callbacks on objects.
2714 .PP
2715 To use it,
2716 .PP
2717 .Vb 1
2718 \& #include <ev++.h>
2719 .Ve
2720 .PP
2721 This automatically includes \fIev.h\fR and puts all of its definitions (many
2722 of them macros) into the global namespace. All \*(C+ specific things are
2723 put into the \f(CW\*(C`ev\*(C'\fR namespace. It should support all the same embedding
2724 options as \fIev.h\fR, most notably \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR.
2725 .PP
2726 Care has been taken to keep the overhead low. The only data member the \*(C+
2727 classes add (compared to plain C\-style watchers) is the event loop pointer
2728 that the watcher is associated with (or no additional members at all if
2729 you disable \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR when embedding libev).
2730 .PP
2731 Currently, functions, and static and non-static member functions can be
2732 used as callbacks. Other types should be easy to add as long as they only
2733 need one additional pointer for context. If you need support for other
2734 types of functors please contact the author (preferably after implementing
2735 it).
2736 .PP
2737 Here is a list of things available in the \f(CW\*(C`ev\*(C'\fR namespace:
2738 .ie n .IP """ev::READ""\fR, \f(CW""ev::WRITE"" etc." 4
2739 .el .IP "\f(CWev::READ\fR, \f(CWev::WRITE\fR etc." 4
2740 .IX Item "ev::READ, ev::WRITE etc."
2741 These are just enum values with the same values as the \f(CW\*(C`EV_READ\*(C'\fR etc.
2742 macros from \fIev.h\fR.
2743 .ie n .IP """ev::tstamp""\fR, \f(CW""ev::now""" 4
2744 .el .IP "\f(CWev::tstamp\fR, \f(CWev::now\fR" 4
2745 .IX Item "ev::tstamp, ev::now"
2746 Aliases to the same types/functions as with the \f(CW\*(C`ev_\*(C'\fR prefix.
2747 .ie n .IP """ev::io""\fR, \f(CW""ev::timer""\fR, \f(CW""ev::periodic""\fR, \f(CW""ev::idle""\fR, \f(CW""ev::sig"" etc." 4
2748 .el .IP "\f(CWev::io\fR, \f(CWev::timer\fR, \f(CWev::periodic\fR, \f(CWev::idle\fR, \f(CWev::sig\fR etc." 4
2749 .IX Item "ev::io, ev::timer, ev::periodic, ev::idle, ev::sig etc."
2750 For each \f(CW\*(C`ev_TYPE\*(C'\fR watcher in \fIev.h\fR there is a corresponding class of
2751 the same name in the \f(CW\*(C`ev\*(C'\fR namespace, with the exception of \f(CW\*(C`ev_signal\*(C'\fR
2752 which is called \f(CW\*(C`ev::sig\*(C'\fR to avoid clashes with the \f(CW\*(C`signal\*(C'\fR macro
2753 defines by many implementations.
2754 .Sp
2755 All of those classes have these methods:
2756 .RS 4
2757 .IP "ev::TYPE::TYPE ()" 4
2758 .IX Item "ev::TYPE::TYPE ()"
2759 .PD 0
2760 .IP "ev::TYPE::TYPE (struct ev_loop *)" 4
2761 .IX Item "ev::TYPE::TYPE (struct ev_loop *)"
2762 .IP "ev::TYPE::~TYPE" 4
2763 .IX Item "ev::TYPE::~TYPE"
2764 .PD
2765 The constructor (optionally) takes an event loop to associate the watcher
2766 with. If it is omitted, it will use \f(CW\*(C`EV_DEFAULT\*(C'\fR.
2767 .Sp
2768 The constructor calls \f(CW\*(C`ev_init\*(C'\fR for you, which means you have to call the
2769 \&\f(CW\*(C`set\*(C'\fR method before starting it.
2770 .Sp
2771 It will not set a callback, however: You have to call the templated \f(CW\*(C`set\*(C'\fR
2772 method to set a callback before you can start the watcher.
2773 .Sp
2774 (The reason why you have to use a method is a limitation in \*(C+ which does
2775 not allow explicit template arguments for constructors).
2776 .Sp
2777 The destructor automatically stops the watcher if it is active.
2778 .IP "w\->set<class, &class::method> (object *)" 4
2779 .IX Item "w->set<class, &class::method> (object *)"
2780 This method sets the callback method to call. The method has to have a
2781 signature of \f(CW\*(C`void (*)(ev_TYPE &, int)\*(C'\fR, it receives the watcher as
2782 first argument and the \f(CW\*(C`revents\*(C'\fR as second. The object must be given as
2783 parameter and is stored in the \f(CW\*(C`data\*(C'\fR member of the watcher.
2784 .Sp
2785 This method synthesizes efficient thunking code to call your method from
2786 the C callback that libev requires. If your compiler can inline your
2787 callback (i.e. it is visible to it at the place of the \f(CW\*(C`set\*(C'\fR call and
2788 your compiler is good :), then the method will be fully inlined into the
2789 thunking function, making it as fast as a direct C callback.
2790 .Sp
2791 Example: simple class declaration and watcher initialisation
2792 .Sp
2793 .Vb 4
2794 \& struct myclass
2795 \& {
2796 \& void io_cb (ev::io &w, int revents) { }
2797 \& }
2798 \&
2799 \& myclass obj;
2800 \& ev::io iow;
2801 \& iow.set <myclass, &myclass::io_cb> (&obj);
2802 .Ve
2803 .IP "w\->set<function> (void *data = 0)" 4
2804 .IX Item "w->set<function> (void *data = 0)"
2805 Also sets a callback, but uses a static method or plain function as
2806 callback. The optional \f(CW\*(C`data\*(C'\fR argument will be stored in the watcher's
2807 \&\f(CW\*(C`data\*(C'\fR member and is free for you to use.
2808 .Sp
2809 The prototype of the \f(CW\*(C`function\*(C'\fR must be \f(CW\*(C`void (*)(ev::TYPE &w, int)\*(C'\fR.
2810 .Sp
2811 See the method\-\f(CW\*(C`set\*(C'\fR above for more details.
2812 .Sp
2813 Example: Use a plain function as callback.
2814 .Sp
2815 .Vb 2
2816 \& static void io_cb (ev::io &w, int revents) { }
2817 \& iow.set <io_cb> ();
2818 .Ve
2819 .IP "w\->set (struct ev_loop *)" 4
2820 .IX Item "w->set (struct ev_loop *)"
2821 Associates a different \f(CW\*(C`struct ev_loop\*(C'\fR with this watcher. You can only
2822 do this when the watcher is inactive (and not pending either).
2823 .IP "w\->set ([arguments])" 4
2824 .IX Item "w->set ([arguments])"
2825 Basically the same as \f(CW\*(C`ev_TYPE_set\*(C'\fR, with the same arguments. Must be
2826 called at least once. Unlike the C counterpart, an active watcher gets
2827 automatically stopped and restarted when reconfiguring it with this
2828 method.
2829 .IP "w\->start ()" 4
2830 .IX Item "w->start ()"
2831 Starts the watcher. Note that there is no \f(CW\*(C`loop\*(C'\fR argument, as the
2832 constructor already stores the event loop.
2833 .IP "w\->stop ()" 4
2834 .IX Item "w->stop ()"
2835 Stops the watcher if it is active. Again, no \f(CW\*(C`loop\*(C'\fR argument.
2836 .ie n .IP "w\->again () (""ev::timer""\fR, \f(CW""ev::periodic"" only)" 4
2837 .el .IP "w\->again () (\f(CWev::timer\fR, \f(CWev::periodic\fR only)" 4
2838 .IX Item "w->again () (ev::timer, ev::periodic only)"
2839 For \f(CW\*(C`ev::timer\*(C'\fR and \f(CW\*(C`ev::periodic\*(C'\fR, this invokes the corresponding
2840 \&\f(CW\*(C`ev_TYPE_again\*(C'\fR function.
2841 .ie n .IP "w\->sweep () (""ev::embed"" only)" 4
2842 .el .IP "w\->sweep () (\f(CWev::embed\fR only)" 4
2843 .IX Item "w->sweep () (ev::embed only)"
2844 Invokes \f(CW\*(C`ev_embed_sweep\*(C'\fR.
2845 .ie n .IP "w\->update () (""ev::stat"" only)" 4
2846 .el .IP "w\->update () (\f(CWev::stat\fR only)" 4
2847 .IX Item "w->update () (ev::stat only)"
2848 Invokes \f(CW\*(C`ev_stat_stat\*(C'\fR.
2849 .RE
2850 .RS 4
2851 .RE
2852 .PP
2853 Example: Define a class with an \s-1IO\s0 and idle watcher, start one of them in
2854 the constructor.
2855 .PP
2856 .Vb 4
2857 \& class myclass
2858 \& {
2859 \& ev::io io ; void io_cb (ev::io &w, int revents);
2860 \& ev::idle idle; void idle_cb (ev::idle &w, int revents);
2861 \&
2862 \& myclass (int fd)
2863 \& {
2864 \& io .set <myclass, &myclass::io_cb > (this);
2865 \& idle.set <myclass, &myclass::idle_cb> (this);
2866 \&
2867 \& io.start (fd, ev::READ);
2868 \& }
2869 \& };
2870 .Ve
2871 .SH "OTHER LANGUAGE BINDINGS"
2872 .IX Header "OTHER LANGUAGE BINDINGS"
2873 Libev does not offer other language bindings itself, but bindings for a
2874 number of languages exist in the form of third-party packages. If you know
2875 any interesting language binding in addition to the ones listed here, drop
2876 me a note.
2877 .IP "Perl" 4
2878 .IX Item "Perl"
2879 The \s-1EV\s0 module implements the full libev \s-1API\s0 and is actually used to test
2880 libev. \s-1EV\s0 is developed together with libev. Apart from the \s-1EV\s0 core module,
2881 there are additional modules that implement libev-compatible interfaces
2882 to \f(CW\*(C`libadns\*(C'\fR (\f(CW\*(C`EV::ADNS\*(C'\fR, but \f(CW\*(C`AnyEvent::DNS\*(C'\fR is preferred nowadays),
2883 \&\f(CW\*(C`Net::SNMP\*(C'\fR (\f(CW\*(C`Net::SNMP::EV\*(C'\fR) and the \f(CW\*(C`libglib\*(C'\fR event core (\f(CW\*(C`Glib::EV\*(C'\fR
2884 and \f(CW\*(C`EV::Glib\*(C'\fR).
2885 .Sp
2886 It can be found and installed via \s-1CPAN\s0, its homepage is at
2887 <http://software.schmorp.de/pkg/EV>.
2888 .IP "Python" 4
2889 .IX Item "Python"
2890 Python bindings can be found at <http://code.google.com/p/pyev/>. It
2891 seems to be quite complete and well-documented. Note, however, that the
2892 patch they require for libev is outright dangerous as it breaks the \s-1ABI\s0
2893 for everybody else, and therefore, should never be applied in an installed
2894 libev (if python requires an incompatible \s-1ABI\s0 then it needs to embed
2895 libev).
2896 .IP "Ruby" 4
2897 .IX Item "Ruby"
2898 Tony Arcieri has written a ruby extension that offers access to a subset
2899 of the libev \s-1API\s0 and adds file handle abstractions, asynchronous \s-1DNS\s0 and
2900 more on top of it. It can be found via gem servers. Its homepage is at
2901 <http://rev.rubyforge.org/>.
2902 .IP "D" 4
2903 .IX Item "D"
2904 Leandro Lucarella has written a D language binding (\fIev.d\fR) for libev, to
2905 be found at <http://proj.llucax.com.ar/wiki/evd>.
2906 .SH "MACRO MAGIC"
2907 .IX Header "MACRO MAGIC"
2908 Libev can be compiled with a variety of options, the most fundamental
2909 of which is \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR. This option determines whether (most)
2910 functions and callbacks have an initial \f(CW\*(C`struct ev_loop *\*(C'\fR argument.
2911 .PP
2912 To make it easier to write programs that cope with either variant, the
2913 following macros are defined:
2914 .ie n .IP """EV_A""\fR, \f(CW""EV_A_""" 4
2915 .el .IP "\f(CWEV_A\fR, \f(CWEV_A_\fR" 4
2916 .IX Item "EV_A, EV_A_"
2917 This provides the loop \fIargument\fR for functions, if one is required (\*(L"ev
2918 loop argument\*(R"). The \f(CW\*(C`EV_A\*(C'\fR form is used when this is the sole argument,
2919 \&\f(CW\*(C`EV_A_\*(C'\fR is used when other arguments are following. Example:
2920 .Sp
2921 .Vb 3
2922 \& ev_unref (EV_A);
2923 \& ev_timer_add (EV_A_ watcher);
2924 \& ev_loop (EV_A_ 0);
2925 .Ve
2926 .Sp
2927 It assumes the variable \f(CW\*(C`loop\*(C'\fR of type \f(CW\*(C`struct ev_loop *\*(C'\fR is in scope,
2928 which is often provided by the following macro.
2929 .ie n .IP """EV_P""\fR, \f(CW""EV_P_""" 4
2930 .el .IP "\f(CWEV_P\fR, \f(CWEV_P_\fR" 4
2931 .IX Item "EV_P, EV_P_"
2932 This provides the loop \fIparameter\fR for functions, if one is required (\*(L"ev
2933 loop parameter\*(R"). The \f(CW\*(C`EV_P\*(C'\fR form is used when this is the sole parameter,
2934 \&\f(CW\*(C`EV_P_\*(C'\fR is used when other parameters are following. Example:
2935 .Sp
2936 .Vb 2
2937 \& // this is how ev_unref is being declared
2938 \& static void ev_unref (EV_P);
2939 \&
2940 \& // this is how you can declare your typical callback
2941 \& static void cb (EV_P_ ev_timer *w, int revents)
2942 .Ve
2943 .Sp
2944 It declares a parameter \f(CW\*(C`loop\*(C'\fR of type \f(CW\*(C`struct ev_loop *\*(C'\fR, quite
2945 suitable for use with \f(CW\*(C`EV_A\*(C'\fR.
2946 .ie n .IP """EV_DEFAULT""\fR, \f(CW""EV_DEFAULT_""" 4
2947 .el .IP "\f(CWEV_DEFAULT\fR, \f(CWEV_DEFAULT_\fR" 4
2948 .IX Item "EV_DEFAULT, EV_DEFAULT_"
2949 Similar to the other two macros, this gives you the value of the default
2950 loop, if multiple loops are supported (\*(L"ev loop default\*(R").
2951 .ie n .IP """EV_DEFAULT_UC""\fR, \f(CW""EV_DEFAULT_UC_""" 4
2952 .el .IP "\f(CWEV_DEFAULT_UC\fR, \f(CWEV_DEFAULT_UC_\fR" 4
2953 .IX Item "EV_DEFAULT_UC, EV_DEFAULT_UC_"
2954 Usage identical to \f(CW\*(C`EV_DEFAULT\*(C'\fR and \f(CW\*(C`EV_DEFAULT_\*(C'\fR, but requires that the
2955 default loop has been initialised (\f(CW\*(C`UC\*(C'\fR == unchecked). Their behaviour
2956 is undefined when the default loop has not been initialised by a previous
2957 execution of \f(CW\*(C`EV_DEFAULT\*(C'\fR, \f(CW\*(C`EV_DEFAULT_\*(C'\fR or \f(CW\*(C`ev_default_init (...)\*(C'\fR.
2958 .Sp
2959 It is often prudent to use \f(CW\*(C`EV_DEFAULT\*(C'\fR when initialising the first
2960 watcher in a function but use \f(CW\*(C`EV_DEFAULT_UC\*(C'\fR afterwards.
2961 .PP
2962 Example: Declare and initialise a check watcher, utilising the above
2963 macros so it will work regardless of whether multiple loops are supported
2964 or not.
2965 .PP
2966 .Vb 5
2967 \& static void
2968 \& check_cb (EV_P_ ev_timer *w, int revents)
2969 \& {
2970 \& ev_check_stop (EV_A_ w);
2971 \& }
2972 \&
2973 \& ev_check check;
2974 \& ev_check_init (&check, check_cb);
2975 \& ev_check_start (EV_DEFAULT_ &check);
2976 \& ev_loop (EV_DEFAULT_ 0);
2977 .Ve
2978 .SH "EMBEDDING"
2979 .IX Header "EMBEDDING"
2980 Libev can (and often is) directly embedded into host
2981 applications. Examples of applications that embed it include the Deliantra
2982 Game Server, the \s-1EV\s0 perl module, the \s-1GNU\s0 Virtual Private Ethernet (gvpe)
2983 and rxvt-unicode.
2984 .PP
2985 The goal is to enable you to just copy the necessary files into your
2986 source directory without having to change even a single line in them, so
2987 you can easily upgrade by simply copying (or having a checked-out copy of
2988 libev somewhere in your source tree).
2989 .Sh "\s-1FILESETS\s0"
2990 .IX Subsection "FILESETS"
2991 Depending on what features you need you need to include one or more sets of files
2992 in your application.
2993 .PP
2994 \fI\s-1CORE\s0 \s-1EVENT\s0 \s-1LOOP\s0\fR
2995 .IX Subsection "CORE EVENT LOOP"
2996 .PP
2997 To include only the libev core (all the \f(CW\*(C`ev_*\*(C'\fR functions), with manual
2998 configuration (no autoconf):
2999 .PP
3000 .Vb 2
3001 \& #define EV_STANDALONE 1
3002 \& #include "ev.c"
3003 .Ve
3004 .PP
3005 This will automatically include \fIev.h\fR, too, and should be done in a
3006 single C source file only to provide the function implementations. To use
3007 it, do the same for \fIev.h\fR in all files wishing to use this \s-1API\s0 (best
3008 done by writing a wrapper around \fIev.h\fR that you can include instead and
3009 where you can put other configuration options):
3010 .PP
3011 .Vb 2
3012 \& #define EV_STANDALONE 1
3013 \& #include "ev.h"
3014 .Ve
3015 .PP
3016 Both header files and implementation files can be compiled with a \*(C+
3017 compiler (at least, thats a stated goal, and breakage will be treated
3018 as a bug).
3019 .PP
3020 You need the following files in your source tree, or in a directory
3021 in your include path (e.g. in libev/ when using \-Ilibev):
3022 .PP
3023 .Vb 4
3024 \& ev.h
3025 \& ev.c
3026 \& ev_vars.h
3027 \& ev_wrap.h
3028 \&
3029 \& ev_win32.c required on win32 platforms only
3030 \&
3031 \& ev_select.c only when select backend is enabled (which is enabled by default)
3032 \& ev_poll.c only when poll backend is enabled (disabled by default)
3033 \& ev_epoll.c only when the epoll backend is enabled (disabled by default)
3034 \& ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
3035 \& ev_port.c only when the solaris port backend is enabled (disabled by default)
3036 .Ve
3037 .PP
3038 \&\fIev.c\fR includes the backend files directly when enabled, so you only need
3039 to compile this single file.
3040 .PP
3041 \fI\s-1LIBEVENT\s0 \s-1COMPATIBILITY\s0 \s-1API\s0\fR
3042 .IX Subsection "LIBEVENT COMPATIBILITY API"
3043 .PP
3044 To include the libevent compatibility \s-1API\s0, also include:
3045 .PP
3046 .Vb 1
3047 \& #include "event.c"
3048 .Ve
3049 .PP
3050 in the file including \fIev.c\fR, and:
3051 .PP
3052 .Vb 1
3053 \& #include "event.h"
3054 .Ve
3055 .PP
3056 in the files that want to use the libevent \s-1API\s0. This also includes \fIev.h\fR.
3057 .PP
3058 You need the following additional files for this:
3059 .PP
3060 .Vb 2
3061 \& event.h
3062 \& event.c
3063 .Ve
3064 .PP
3065 \fI\s-1AUTOCONF\s0 \s-1SUPPORT\s0\fR
3066 .IX Subsection "AUTOCONF SUPPORT"
3067 .PP
3068 Instead of using \f(CW\*(C`EV_STANDALONE=1\*(C'\fR and providing your configuration in
3069 whatever way you want, you can also \f(CW\*(C`m4_include([libev.m4])\*(C'\fR in your
3070 \&\fIconfigure.ac\fR and leave \f(CW\*(C`EV_STANDALONE\*(C'\fR undefined. \fIev.c\fR will then
3071 include \fIconfig.h\fR and configure itself accordingly.
3072 .PP
3073 For this of course you need the m4 file:
3074 .PP
3075 .Vb 1
3076 \& libev.m4
3077 .Ve
3078 .Sh "\s-1PREPROCESSOR\s0 \s-1SYMBOLS/MACROS\s0"
3079 .IX Subsection "PREPROCESSOR SYMBOLS/MACROS"
3080 Libev can be configured via a variety of preprocessor symbols you have to
3081 define before including any of its files. The default in the absence of
3082 autoconf is documented for every option.
3083 .IP "\s-1EV_STANDALONE\s0" 4
3084 .IX Item "EV_STANDALONE"
3085 Must always be \f(CW1\fR if you do not use autoconf configuration, which
3086 keeps libev from including \fIconfig.h\fR, and it also defines dummy
3087 implementations for some libevent functions (such as logging, which is not
3088 supported). It will also not define any of the structs usually found in
3089 \&\fIevent.h\fR that are not directly supported by the libev core alone.
3090 .IP "\s-1EV_USE_MONOTONIC\s0" 4
3091 .IX Item "EV_USE_MONOTONIC"
3092 If defined to be \f(CW1\fR, libev will try to detect the availability of the
3093 monotonic clock option at both compile time and runtime. Otherwise no use
3094 of the monotonic clock option will be attempted. If you enable this, you
3095 usually have to link against librt or something similar. Enabling it when
3096 the functionality isn't available is safe, though, although you have
3097 to make sure you link against any libraries where the \f(CW\*(C`clock_gettime\*(C'\fR
3098 function is hiding in (often \fI\-lrt\fR).
3099 .IP "\s-1EV_USE_REALTIME\s0" 4
3100 .IX Item "EV_USE_REALTIME"
3101 If defined to be \f(CW1\fR, libev will try to detect the availability of the
3102 real-time clock option at compile time (and assume its availability at
3103 runtime if successful). Otherwise no use of the real-time clock option will
3104 be attempted. This effectively replaces \f(CW\*(C`gettimeofday\*(C'\fR by \f(CW\*(C`clock_get
3105 (CLOCK_REALTIME, ...)\*(C'\fR and will not normally affect correctness. See the
3106 note about libraries in the description of \f(CW\*(C`EV_USE_MONOTONIC\*(C'\fR, though.
3107 .IP "\s-1EV_USE_NANOSLEEP\s0" 4
3108 .IX Item "EV_USE_NANOSLEEP"
3109 If defined to be \f(CW1\fR, libev will assume that \f(CW\*(C`nanosleep ()\*(C'\fR is available
3110 and will use it for delays. Otherwise it will use \f(CW\*(C`select ()\*(C'\fR.
3111 .IP "\s-1EV_USE_EVENTFD\s0" 4
3112 .IX Item "EV_USE_EVENTFD"
3113 If defined to be \f(CW1\fR, then libev will assume that \f(CW\*(C`eventfd ()\*(C'\fR is
3114 available and will probe for kernel support at runtime. This will improve
3115 \&\f(CW\*(C`ev_signal\*(C'\fR and \f(CW\*(C`ev_async\*(C'\fR performance and reduce resource consumption.
3116 If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
3117 2.7 or newer, otherwise disabled.
3118 .IP "\s-1EV_USE_SELECT\s0" 4
3119 .IX Item "EV_USE_SELECT"
3120 If undefined or defined to be \f(CW1\fR, libev will compile in support for the
3121 \&\f(CW\*(C`select\*(C'\fR(2) backend. No attempt at auto-detection will be done: if no
3122 other method takes over, select will be it. Otherwise the select backend
3123 will not be compiled in.
3124 .IP "\s-1EV_SELECT_USE_FD_SET\s0" 4
3125 .IX Item "EV_SELECT_USE_FD_SET"
3126 If defined to \f(CW1\fR, then the select backend will use the system \f(CW\*(C`fd_set\*(C'\fR
3127 structure. This is useful if libev doesn't compile due to a missing
3128 \&\f(CW\*(C`NFDBITS\*(C'\fR or \f(CW\*(C`fd_mask\*(C'\fR definition or it mis-guesses the bitset layout on
3129 exotic systems. This usually limits the range of file descriptors to some
3130 low limit such as 1024 or might have other limitations (winsocket only
3131 allows 64 sockets). The \f(CW\*(C`FD_SETSIZE\*(C'\fR macro, set before compilation, might
3132 influence the size of the \f(CW\*(C`fd_set\*(C'\fR used.
3133 .IP "\s-1EV_SELECT_IS_WINSOCKET\s0" 4
3134 .IX Item "EV_SELECT_IS_WINSOCKET"
3135 When defined to \f(CW1\fR, the select backend will assume that
3136 select/socket/connect etc. don't understand file descriptors but
3137 wants osf handles on win32 (this is the case when the select to
3138 be used is the winsock select). This means that it will call
3139 \&\f(CW\*(C`_get_osfhandle\*(C'\fR on the fd to convert it to an \s-1OS\s0 handle. Otherwise,
3140 it is assumed that all these functions actually work on fds, even
3141 on win32. Should not be defined on non\-win32 platforms.
3142 .IP "\s-1EV_FD_TO_WIN32_HANDLE\s0" 4
3143 .IX Item "EV_FD_TO_WIN32_HANDLE"
3144 If \f(CW\*(C`EV_SELECT_IS_WINSOCKET\*(C'\fR is enabled, then libev needs a way to map
3145 file descriptors to socket handles. When not defining this symbol (the
3146 default), then libev will call \f(CW\*(C`_get_osfhandle\*(C'\fR, which is usually
3147 correct. In some cases, programs use their own file descriptor management,
3148 in which case they can provide this function to map fds to socket handles.
3149 .IP "\s-1EV_USE_POLL\s0" 4
3150 .IX Item "EV_USE_POLL"
3151 If defined to be \f(CW1\fR, libev will compile in support for the \f(CW\*(C`poll\*(C'\fR(2)
3152 backend. Otherwise it will be enabled on non\-win32 platforms. It
3153 takes precedence over select.
3154 .IP "\s-1EV_USE_EPOLL\s0" 4
3155 .IX Item "EV_USE_EPOLL"
3156 If defined to be \f(CW1\fR, libev will compile in support for the Linux
3157 \&\f(CW\*(C`epoll\*(C'\fR(7) backend. Its availability will be detected at runtime,
3158 otherwise another method will be used as fallback. This is the preferred
3159 backend for GNU/Linux systems. If undefined, it will be enabled if the
3160 headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
3161 .IP "\s-1EV_USE_KQUEUE\s0" 4
3162 .IX Item "EV_USE_KQUEUE"
3163 If defined to be \f(CW1\fR, libev will compile in support for the \s-1BSD\s0 style
3164 \&\f(CW\*(C`kqueue\*(C'\fR(2) backend. Its actual availability will be detected at runtime,
3165 otherwise another method will be used as fallback. This is the preferred
3166 backend for \s-1BSD\s0 and BSD-like systems, although on most BSDs kqueue only
3167 supports some types of fds correctly (the only platform we found that
3168 supports ptys for example was NetBSD), so kqueue might be compiled in, but
3169 not be used unless explicitly requested. The best way to use it is to find
3170 out whether kqueue supports your type of fd properly and use an embedded
3171 kqueue loop.
3172 .IP "\s-1EV_USE_PORT\s0" 4
3173 .IX Item "EV_USE_PORT"
3174 If defined to be \f(CW1\fR, libev will compile in support for the Solaris
3175 10 port style backend. Its availability will be detected at runtime,
3176 otherwise another method will be used as fallback. This is the preferred
3177 backend for Solaris 10 systems.
3178 .IP "\s-1EV_USE_DEVPOLL\s0" 4
3179 .IX Item "EV_USE_DEVPOLL"
3180 Reserved for future expansion, works like the \s-1USE\s0 symbols above.
3181 .IP "\s-1EV_USE_INOTIFY\s0" 4
3182 .IX Item "EV_USE_INOTIFY"
3183 If defined to be \f(CW1\fR, libev will compile in support for the Linux inotify
3184 interface to speed up \f(CW\*(C`ev_stat\*(C'\fR watchers. Its actual availability will
3185 be detected at runtime. If undefined, it will be enabled if the headers
3186 indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
3187 .IP "\s-1EV_ATOMIC_T\s0" 4
3188 .IX Item "EV_ATOMIC_T"
3189 Libev requires an integer type (suitable for storing \f(CW0\fR or \f(CW1\fR) whose
3190 access is atomic with respect to other threads or signal contexts. No such
3191 type is easily found in the C language, so you can provide your own type
3192 that you know is safe for your purposes. It is used both for signal handler \*(L"locking\*(R"
3193 as well as for signal and thread safety in \f(CW\*(C`ev_async\*(C'\fR watchers.
3194 .Sp
3195 In the absence of this define, libev will use \f(CW\*(C`sig_atomic_t volatile\*(C'\fR
3196 (from \fIsignal.h\fR), which is usually good enough on most platforms.
3197 .IP "\s-1EV_H\s0" 4
3198 .IX Item "EV_H"
3199 The name of the \fIev.h\fR header file used to include it. The default if
3200 undefined is \f(CW"ev.h"\fR in \fIevent.h\fR, \fIev.c\fR and \fIev++.h\fR. This can be
3201 used to virtually rename the \fIev.h\fR header file in case of conflicts.
3202 .IP "\s-1EV_CONFIG_H\s0" 4
3203 .IX Item "EV_CONFIG_H"
3204 If \f(CW\*(C`EV_STANDALONE\*(C'\fR isn't \f(CW1\fR, this variable can be used to override
3205 \&\fIev.c\fR's idea of where to find the \fIconfig.h\fR file, similarly to
3206 \&\f(CW\*(C`EV_H\*(C'\fR, above.
3207 .IP "\s-1EV_EVENT_H\s0" 4
3208 .IX Item "EV_EVENT_H"
3209 Similarly to \f(CW\*(C`EV_H\*(C'\fR, this macro can be used to override \fIevent.c\fR's idea
3210 of how the \fIevent.h\fR header can be found, the default is \f(CW"event.h"\fR.
3211 .IP "\s-1EV_PROTOTYPES\s0" 4
3212 .IX Item "EV_PROTOTYPES"
3213 If defined to be \f(CW0\fR, then \fIev.h\fR will not define any function
3214 prototypes, but still define all the structs and other symbols. This is
3215 occasionally useful if you want to provide your own wrapper functions
3216 around libev functions.
3217 .IP "\s-1EV_MULTIPLICITY\s0" 4
3218 .IX Item "EV_MULTIPLICITY"
3219 If undefined or defined to \f(CW1\fR, then all event-loop-specific functions
3220 will have the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument, and you can create
3221 additional independent event loops. Otherwise there will be no support
3222 for multiple event loops and there is no first event loop pointer
3223 argument. Instead, all functions act on the single default loop.
3224 .IP "\s-1EV_MINPRI\s0" 4
3225 .IX Item "EV_MINPRI"
3226 .PD 0
3227 .IP "\s-1EV_MAXPRI\s0" 4
3228 .IX Item "EV_MAXPRI"
3229 .PD
3230 The range of allowed priorities. \f(CW\*(C`EV_MINPRI\*(C'\fR must be smaller or equal to
3231 \&\f(CW\*(C`EV_MAXPRI\*(C'\fR, but otherwise there are no non-obvious limitations. You can
3232 provide for more priorities by overriding those symbols (usually defined
3233 to be \f(CW\*(C`\-2\*(C'\fR and \f(CW2\fR, respectively).
3234 .Sp
3235 When doing priority-based operations, libev usually has to linearly search
3236 all the priorities, so having many of them (hundreds) uses a lot of space
3237 and time, so using the defaults of five priorities (\-2 .. +2) is usually
3238 fine.
3239 .Sp
3240 If your embedding application does not need any priorities, defining these
3241 both to \f(CW0\fR will save some memory and \s-1CPU\s0.
3242 .IP "\s-1EV_PERIODIC_ENABLE\s0" 4
3243 .IX Item "EV_PERIODIC_ENABLE"
3244 If undefined or defined to be \f(CW1\fR, then periodic timers are supported. If
3245 defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of
3246 code.
3247 .IP "\s-1EV_IDLE_ENABLE\s0" 4
3248 .IX Item "EV_IDLE_ENABLE"
3249 If undefined or defined to be \f(CW1\fR, then idle watchers are supported. If
3250 defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of
3251 code.
3252 .IP "\s-1EV_EMBED_ENABLE\s0" 4
3253 .IX Item "EV_EMBED_ENABLE"
3254 If undefined or defined to be \f(CW1\fR, then embed watchers are supported. If
3255 defined to be \f(CW0\fR, then they are not. Embed watchers rely on most other
3256 watcher types, which therefore must not be disabled.
3257 .IP "\s-1EV_STAT_ENABLE\s0" 4
3258 .IX Item "EV_STAT_ENABLE"
3259 If undefined or defined to be \f(CW1\fR, then stat watchers are supported. If
3260 defined to be \f(CW0\fR, then they are not.
3261 .IP "\s-1EV_FORK_ENABLE\s0" 4
3262 .IX Item "EV_FORK_ENABLE"
3263 If undefined or defined to be \f(CW1\fR, then fork watchers are supported. If
3264 defined to be \f(CW0\fR, then they are not.
3265 .IP "\s-1EV_ASYNC_ENABLE\s0" 4
3266 .IX Item "EV_ASYNC_ENABLE"
3267 If undefined or defined to be \f(CW1\fR, then async watchers are supported. If
3268 defined to be \f(CW0\fR, then they are not.
3269 .IP "\s-1EV_MINIMAL\s0" 4
3270 .IX Item "EV_MINIMAL"
3271 If you need to shave off some kilobytes of code at the expense of some
3272 speed, define this symbol to \f(CW1\fR. Currently this is used to override some
3273 inlining decisions, saves roughly 30% code size on amd64. It also selects a
3274 much smaller 2\-heap for timer management over the default 4\-heap.
3275 .IP "\s-1EV_PID_HASHSIZE\s0" 4
3276 .IX Item "EV_PID_HASHSIZE"
3277 \&\f(CW\*(C`ev_child\*(C'\fR watchers use a small hash table to distribute workload by
3278 pid. The default size is \f(CW16\fR (or \f(CW1\fR with \f(CW\*(C`EV_MINIMAL\*(C'\fR), usually more
3279 than enough. If you need to manage thousands of children you might want to
3280 increase this value (\fImust\fR be a power of two).
3281 .IP "\s-1EV_INOTIFY_HASHSIZE\s0" 4
3282 .IX Item "EV_INOTIFY_HASHSIZE"
3283 \&\f(CW\*(C`ev_stat\*(C'\fR watchers use a small hash table to distribute workload by
3284 inotify watch id. The default size is \f(CW16\fR (or \f(CW1\fR with \f(CW\*(C`EV_MINIMAL\*(C'\fR),
3285 usually more than enough. If you need to manage thousands of \f(CW\*(C`ev_stat\*(C'\fR
3286 watchers you might want to increase this value (\fImust\fR be a power of
3287 two).
3288 .IP "\s-1EV_USE_4HEAP\s0" 4
3289 .IX Item "EV_USE_4HEAP"
3290 Heaps are not very cache-efficient. To improve the cache-efficiency of the
3291 timer and periodics heaps, libev uses a 4\-heap when this symbol is defined
3292 to \f(CW1\fR. The 4\-heap uses more complicated (longer) code but has noticeably
3293 faster performance with many (thousands) of watchers.
3294 .Sp
3295 The default is \f(CW1\fR unless \f(CW\*(C`EV_MINIMAL\*(C'\fR is set in which case it is \f(CW0\fR
3296 (disabled).
3297 .IP "\s-1EV_HEAP_CACHE_AT\s0" 4
3298 .IX Item "EV_HEAP_CACHE_AT"
3299 Heaps are not very cache-efficient. To improve the cache-efficiency of the
3300 timer and periodics heaps, libev can cache the timestamp (\fIat\fR) within
3301 the heap structure (selected by defining \f(CW\*(C`EV_HEAP_CACHE_AT\*(C'\fR to \f(CW1\fR),
3302 which uses 8\-12 bytes more per watcher and a few hundred bytes more code,
3303 but avoids random read accesses on heap changes. This improves performance
3304 noticeably with many (hundreds) of watchers.
3305 .Sp
3306 The default is \f(CW1\fR unless \f(CW\*(C`EV_MINIMAL\*(C'\fR is set in which case it is \f(CW0\fR
3307 (disabled).
3308 .IP "\s-1EV_VERIFY\s0" 4
3309 .IX Item "EV_VERIFY"
3310 Controls how much internal verification (see \f(CW\*(C`ev_loop_verify ()\*(C'\fR) will
3311 be done: If set to \f(CW0\fR, no internal verification code will be compiled
3312 in. If set to \f(CW1\fR, then verification code will be compiled in, but not
3313 called. If set to \f(CW2\fR, then the internal verification code will be
3314 called once per loop, which can slow down libev. If set to \f(CW3\fR, then the
3315 verification code will be called very frequently, which will slow down
3316 libev considerably.
3317 .Sp
3318 The default is \f(CW1\fR, unless \f(CW\*(C`EV_MINIMAL\*(C'\fR is set, in which case it will be
3319 \&\f(CW0\fR.
3320 .IP "\s-1EV_COMMON\s0" 4
3321 .IX Item "EV_COMMON"
3322 By default, all watchers have a \f(CW\*(C`void *data\*(C'\fR member. By redefining
3323 this macro to a something else you can include more and other types of
3324 members. You have to define it each time you include one of the files,
3325 though, and it must be identical each time.
3326 .Sp
3327 For example, the perl \s-1EV\s0 module uses something like this:
3328 .Sp
3329 .Vb 3
3330 \& #define EV_COMMON \e
3331 \& SV *self; /* contains this struct */ \e
3332 \& SV *cb_sv, *fh /* note no trailing ";" */
3333 .Ve
3334 .IP "\s-1EV_CB_DECLARE\s0 (type)" 4
3335 .IX Item "EV_CB_DECLARE (type)"
3336 .PD 0
3337 .IP "\s-1EV_CB_INVOKE\s0 (watcher, revents)" 4
3338 .IX Item "EV_CB_INVOKE (watcher, revents)"
3339 .IP "ev_set_cb (ev, cb)" 4
3340 .IX Item "ev_set_cb (ev, cb)"
3341 .PD
3342 Can be used to change the callback member declaration in each watcher,
3343 and the way callbacks are invoked and set. Must expand to a struct member
3344 definition and a statement, respectively. See the \fIev.h\fR header file for
3345 their default definitions. One possible use for overriding these is to
3346 avoid the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument in all cases, or to use
3347 method calls instead of plain function calls in \*(C+.
3348 .Sh "\s-1EXPORTED\s0 \s-1API\s0 \s-1SYMBOLS\s0"
3349 .IX Subsection "EXPORTED API SYMBOLS"
3350 If you need to re-export the \s-1API\s0 (e.g. via a \s-1DLL\s0) and you need a list of
3351 exported symbols, you can use the provided \fISymbol.*\fR files which list
3352 all public symbols, one per line:
3353 .PP
3354 .Vb 2
3355 \& Symbols.ev for libev proper
3356 \& Symbols.event for the libevent emulation
3357 .Ve
3358 .PP
3359 This can also be used to rename all public symbols to avoid clashes with
3360 multiple versions of libev linked together (which is obviously bad in
3361 itself, but sometimes it is inconvenient to avoid this).
3362 .PP
3363 A sed command like this will create wrapper \f(CW\*(C`#define\*(C'\fR's that you need to
3364 include before including \fIev.h\fR:
3365 .PP
3366 .Vb 1
3367 \& <Symbols.ev sed \-e "s/.*/#define & myprefix_&/" >wrap.h
3368 .Ve
3369 .PP
3370 This would create a file \fIwrap.h\fR which essentially looks like this:
3371 .PP
3372 .Vb 4
3373 \& #define ev_backend myprefix_ev_backend
3374 \& #define ev_check_start myprefix_ev_check_start
3375 \& #define ev_check_stop myprefix_ev_check_stop
3376 \& ...
3377 .Ve
3378 .Sh "\s-1EXAMPLES\s0"
3379 .IX Subsection "EXAMPLES"
3380 For a real-world example of a program the includes libev
3381 verbatim, you can have a look at the \s-1EV\s0 perl module
3382 (<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
3383 the \fIlibev/\fR subdirectory and includes them in the \fI\s-1EV/EVAPI\s0.h\fR (public
3384 interface) and \fI\s-1EV\s0.xs\fR (implementation) files. Only the \fI\s-1EV\s0.xs\fR file
3385 will be compiled. It is pretty complex because it provides its own header
3386 file.
3387 .PP
3388 The usage in rxvt-unicode is simpler. It has a \fIev_cpp.h\fR header file
3389 that everybody includes and which overrides some configure choices:
3390 .PP
3391 .Vb 9
3392 \& #define EV_MINIMAL 1
3393 \& #define EV_USE_POLL 0
3394 \& #define EV_MULTIPLICITY 0
3395 \& #define EV_PERIODIC_ENABLE 0
3396 \& #define EV_STAT_ENABLE 0
3397 \& #define EV_FORK_ENABLE 0
3398 \& #define EV_CONFIG_H <config.h>
3399 \& #define EV_MINPRI 0
3400 \& #define EV_MAXPRI 0
3401 \&
3402 \& #include "ev++.h"
3403 .Ve
3404 .PP
3405 And a \fIev_cpp.C\fR implementation file that contains libev proper and is compiled:
3406 .PP
3407 .Vb 2
3408 \& #include "ev_cpp.h"
3409 \& #include "ev.c"
3410 .Ve
3411 .SH "INTERACTION WITH OTHER PROGRAMS OR LIBRARIES"
3412 .IX Header "INTERACTION WITH OTHER PROGRAMS OR LIBRARIES"
3413 .Sh "\s-1THREADS\s0 \s-1AND\s0 \s-1COROUTINES\s0"
3414 .IX Subsection "THREADS AND COROUTINES"
3415 \fI\s-1THREADS\s0\fR
3416 .IX Subsection "THREADS"
3417 .PP
3418 All libev functions are reentrant and thread-safe unless explicitly
3419 documented otherwise, but libev implements no locking itself. This means
3420 that you can use as many loops as you want in parallel, as long as there
3421 are no concurrent calls into any libev function with the same loop
3422 parameter (\f(CW\*(C`ev_default_*\*(C'\fR calls have an implicit default loop parameter,
3423 of course): libev guarantees that different event loops share no data
3424 structures that need any locking.
3425 .PP
3426 Or to put it differently: calls with different loop parameters can be done
3427 concurrently from multiple threads, calls with the same loop parameter
3428 must be done serially (but can be done from different threads, as long as
3429 only one thread ever is inside a call at any point in time, e.g. by using
3430 a mutex per loop).
3431 .PP
3432 Specifically to support threads (and signal handlers), libev implements
3433 so-called \f(CW\*(C`ev_async\*(C'\fR watchers, which allow some limited form of
3434 concurrency on the same event loop, namely waking it up \*(L"from the
3435 outside\*(R".
3436 .PP
3437 If you want to know which design (one loop, locking, or multiple loops
3438 without or something else still) is best for your problem, then I cannot
3439 help you, but here is some generic advice:
3440 .IP "\(bu" 4
3441 most applications have a main thread: use the default libev loop
3442 in that thread, or create a separate thread running only the default loop.
3443 .Sp
3444 This helps integrating other libraries or software modules that use libev
3445 themselves and don't care/know about threading.
3446 .IP "\(bu" 4
3447 one loop per thread is usually a good model.
3448 .Sp
3449 Doing this is almost never wrong, sometimes a better-performance model
3450 exists, but it is always a good start.
3451 .IP "\(bu" 4
3452 other models exist, such as the leader/follower pattern, where one
3453 loop is handed through multiple threads in a kind of round-robin fashion.
3454 .Sp
3455 Choosing a model is hard \- look around, learn, know that usually you can do
3456 better than you currently do :\-)
3457 .IP "\(bu" 4
3458 often you need to talk to some other thread which blocks in the
3459 event loop.
3460 .Sp
3461 \&\f(CW\*(C`ev_async\*(C'\fR watchers can be used to wake them up from other threads safely
3462 (or from signal contexts...).
3463 .Sp
3464 An example use would be to communicate signals or other events that only
3465 work in the default loop by registering the signal watcher with the
3466 default loop and triggering an \f(CW\*(C`ev_async\*(C'\fR watcher from the default loop
3467 watcher callback into the event loop interested in the signal.
3468 .PP
3469 \fI\s-1COROUTINES\s0\fR
3470 .IX Subsection "COROUTINES"
3471 .PP
3472 Libev is very accommodating to coroutines (\*(L"cooperative threads\*(R"):
3473 libev fully supports nesting calls to its functions from different
3474 coroutines (e.g. you can call \f(CW\*(C`ev_loop\*(C'\fR on the same loop from two
3475 different coroutines, and switch freely between both coroutines running the
3476 loop, as long as you don't confuse yourself). The only exception is that
3477 you must not do this from \f(CW\*(C`ev_periodic\*(C'\fR reschedule callbacks.
3478 .PP
3479 Care has been taken to ensure that libev does not keep local state inside
3480 \&\f(CW\*(C`ev_loop\*(C'\fR, and other calls do not usually allow for coroutine switches as
3481 they do not clal any callbacks.
3482 .Sh "\s-1COMPILER\s0 \s-1WARNINGS\s0"
3483 .IX Subsection "COMPILER WARNINGS"
3484 Depending on your compiler and compiler settings, you might get no or a
3485 lot of warnings when compiling libev code. Some people are apparently
3486 scared by this.
3487 .PP
3488 However, these are unavoidable for many reasons. For one, each compiler
3489 has different warnings, and each user has different tastes regarding
3490 warning options. \*(L"Warn-free\*(R" code therefore cannot be a goal except when
3491 targeting a specific compiler and compiler-version.
3492 .PP
3493 Another reason is that some compiler warnings require elaborate
3494 workarounds, or other changes to the code that make it less clear and less
3495 maintainable.
3496 .PP
3497 And of course, some compiler warnings are just plain stupid, or simply
3498 wrong (because they don't actually warn about the condition their message
3499 seems to warn about). For example, certain older gcc versions had some
3500 warnings that resulted an extreme number of false positives. These have
3501 been fixed, but some people still insist on making code warn-free with
3502 such buggy versions.
3503 .PP
3504 While libev is written to generate as few warnings as possible,
3505 \&\*(L"warn-free\*(R" code is not a goal, and it is recommended not to build libev
3506 with any compiler warnings enabled unless you are prepared to cope with
3507 them (e.g. by ignoring them). Remember that warnings are just that:
3508 warnings, not errors, or proof of bugs.
3509 .Sh "\s-1VALGRIND\s0"
3510 .IX Subsection "VALGRIND"
3511 Valgrind has a special section here because it is a popular tool that is
3512 highly useful. Unfortunately, valgrind reports are very hard to interpret.
3513 .PP
3514 If you think you found a bug (memory leak, uninitialised data access etc.)
3515 in libev, then check twice: If valgrind reports something like:
3516 .PP
3517 .Vb 3
3518 \& ==2274== definitely lost: 0 bytes in 0 blocks.
3519 \& ==2274== possibly lost: 0 bytes in 0 blocks.
3520 \& ==2274== still reachable: 256 bytes in 1 blocks.
3521 .Ve
3522 .PP
3523 Then there is no memory leak, just as memory accounted to global variables
3524 is not a memleak \- the memory is still being refernced, and didn't leak.
3525 .PP
3526 Similarly, under some circumstances, valgrind might report kernel bugs
3527 as if it were a bug in libev (e.g. in realloc or in the poll backend,
3528 although an acceptable workaround has been found here), or it might be
3529 confused.
3530 .PP
3531 Keep in mind that valgrind is a very good tool, but only a tool. Don't
3532 make it into some kind of religion.
3533 .PP
3534 If you are unsure about something, feel free to contact the mailing list
3535 with the full valgrind report and an explanation on why you think this
3536 is a bug in libev (best check the archives, too :). However, don't be
3537 annoyed when you get a brisk \*(L"this is no bug\*(R" answer and take the chance
3538 of learning how to interpret valgrind properly.
3539 .PP
3540 If you need, for some reason, empty reports from valgrind for your project
3541 I suggest using suppression lists.
3542 .SH "PORTABILITY NOTES"
3543 .IX Header "PORTABILITY NOTES"
3544 .Sh "\s-1WIN32\s0 \s-1PLATFORM\s0 \s-1LIMITATIONS\s0 \s-1AND\s0 \s-1WORKAROUNDS\s0"
3545 .IX Subsection "WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS"
3546 Win32 doesn't support any of the standards (e.g. \s-1POSIX\s0) that libev
3547 requires, and its I/O model is fundamentally incompatible with the \s-1POSIX\s0
3548 model. Libev still offers limited functionality on this platform in
3549 the form of the \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR backend, and only supports socket
3550 descriptors. This only applies when using Win32 natively, not when using
3551 e.g. cygwin.
3552 .PP
3553 Lifting these limitations would basically require the full
3554 re-implementation of the I/O system. If you are into these kinds of
3555 things, then note that glib does exactly that for you in a very portable
3556 way (note also that glib is the slowest event library known to man).
3557 .PP
3558 There is no supported compilation method available on windows except
3559 embedding it into other applications.
3560 .PP
3561 Not a libev limitation but worth mentioning: windows apparently doesn't
3562 accept large writes: instead of resulting in a partial write, windows will
3563 either accept everything or return \f(CW\*(C`ENOBUFS\*(C'\fR if the buffer is too large,
3564 so make sure you only write small amounts into your sockets (less than a
3565 megabyte seems safe, but this apparently depends on the amount of memory
3566 available).
3567 .PP
3568 Due to the many, low, and arbitrary limits on the win32 platform and
3569 the abysmal performance of winsockets, using a large number of sockets
3570 is not recommended (and not reasonable). If your program needs to use
3571 more than a hundred or so sockets, then likely it needs to use a totally
3572 different implementation for windows, as libev offers the \s-1POSIX\s0 readiness
3573 notification model, which cannot be implemented efficiently on windows
3574 (Microsoft monopoly games).
3575 .PP
3576 A typical way to use libev under windows is to embed it (see the embedding
3577 section for details) and use the following \fIevwrap.h\fR header file instead
3578 of \fIev.h\fR:
3579 .PP
3580 .Vb 2
3581 \& #define EV_STANDALONE /* keeps ev from requiring config.h */
3582 \& #define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
3583 \&
3584 \& #include "ev.h"
3585 .Ve
3586 .PP
3587 And compile the following \fIevwrap.c\fR file into your project (make sure
3588 you do \fInot\fR compile the \fIev.c\fR or any other embedded source files!):
3589 .PP
3590 .Vb 2
3591 \& #include "evwrap.h"
3592 \& #include "ev.c"
3593 .Ve
3594 .IP "The winsocket select function" 4
3595 .IX Item "The winsocket select function"
3596 The winsocket \f(CW\*(C`select\*(C'\fR function doesn't follow \s-1POSIX\s0 in that it
3597 requires socket \fIhandles\fR and not socket \fIfile descriptors\fR (it is
3598 also extremely buggy). This makes select very inefficient, and also
3599 requires a mapping from file descriptors to socket handles (the Microsoft
3600 C runtime provides the function \f(CW\*(C`_open_osfhandle\*(C'\fR for this). See the
3601 discussion of the \f(CW\*(C`EV_SELECT_USE_FD_SET\*(C'\fR, \f(CW\*(C`EV_SELECT_IS_WINSOCKET\*(C'\fR and
3602 \&\f(CW\*(C`EV_FD_TO_WIN32_HANDLE\*(C'\fR preprocessor symbols for more info.
3603 .Sp
3604 The configuration for a \*(L"naked\*(R" win32 using the Microsoft runtime
3605 libraries and raw winsocket select is:
3606 .Sp
3607 .Vb 2
3608 \& #define EV_USE_SELECT 1
3609 \& #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
3610 .Ve
3611 .Sp
3612 Note that winsockets handling of fd sets is O(n), so you can easily get a
3613 complexity in the O(nA\*^X) range when using win32.
3614 .IP "Limited number of file descriptors" 4
3615 .IX Item "Limited number of file descriptors"
3616 Windows has numerous arbitrary (and low) limits on things.
3617 .Sp
3618 Early versions of winsocket's select only supported waiting for a maximum
3619 of \f(CW64\fR handles (probably owning to the fact that all windows kernels
3620 can only wait for \f(CW64\fR things at the same time internally; Microsoft
3621 recommends spawning a chain of threads and wait for 63 handles and the
3622 previous thread in each. Great).
3623 .Sp
3624 Newer versions support more handles, but you need to define \f(CW\*(C`FD_SETSIZE\*(C'\fR
3625 to some high number (e.g. \f(CW2048\fR) before compiling the winsocket select
3626 call (which might be in libev or elsewhere, for example, perl does its own
3627 select emulation on windows).
3628 .Sp
3629 Another limit is the number of file descriptors in the Microsoft runtime
3630 libraries, which by default is \f(CW64\fR (there must be a hidden \fI64\fR fetish
3631 or something like this inside Microsoft). You can increase this by calling
3632 \&\f(CW\*(C`_setmaxstdio\*(C'\fR, which can increase this limit to \f(CW2048\fR (another
3633 arbitrary limit), but is broken in many versions of the Microsoft runtime
3634 libraries.
3635 .Sp
3636 This might get you to about \f(CW512\fR or \f(CW2048\fR sockets (depending on
3637 windows version and/or the phase of the moon). To get more, you need to
3638 wrap all I/O functions and provide your own fd management, but the cost of
3639 calling select (O(nA\*^X)) will likely make this unworkable.
3640 .Sh "\s-1PORTABILITY\s0 \s-1REQUIREMENTS\s0"
3641 .IX Subsection "PORTABILITY REQUIREMENTS"
3642 In addition to a working ISO-C implementation and of course the
3643 backend-specific APIs, libev relies on a few additional extensions:
3644 .ie n .IP """void (*)(ev_watcher_type *, int revents)""\fR must have compatible calling conventions regardless of \f(CW""ev_watcher_type *""." 4
3645 .el .IP "\f(CWvoid (*)(ev_watcher_type *, int revents)\fR must have compatible calling conventions regardless of \f(CWev_watcher_type *\fR." 4
3646 .IX Item "void (*)(ev_watcher_type *, int revents) must have compatible calling conventions regardless of ev_watcher_type *."
3647 Libev assumes not only that all watcher pointers have the same internal
3648 structure (guaranteed by \s-1POSIX\s0 but not by \s-1ISO\s0 C for example), but it also
3649 assumes that the same (machine) code can be used to call any watcher
3650 callback: The watcher callbacks have different type signatures, but libev
3651 calls them using an \f(CW\*(C`ev_watcher *\*(C'\fR internally.
3652 .ie n .IP """sig_atomic_t volatile"" must be thread-atomic as well" 4
3653 .el .IP "\f(CWsig_atomic_t volatile\fR must be thread-atomic as well" 4
3654 .IX Item "sig_atomic_t volatile must be thread-atomic as well"
3655 The type \f(CW\*(C`sig_atomic_t volatile\*(C'\fR (or whatever is defined as
3656 \&\f(CW\*(C`EV_ATOMIC_T\*(C'\fR) must be atomic with respect to accesses from different
3657 threads. This is not part of the specification for \f(CW\*(C`sig_atomic_t\*(C'\fR, but is
3658 believed to be sufficiently portable.
3659 .ie n .IP """sigprocmask"" must work in a threaded environment" 4
3660 .el .IP "\f(CWsigprocmask\fR must work in a threaded environment" 4
3661 .IX Item "sigprocmask must work in a threaded environment"
3662 Libev uses \f(CW\*(C`sigprocmask\*(C'\fR to temporarily block signals. This is not
3663 allowed in a threaded program (\f(CW\*(C`pthread_sigmask\*(C'\fR has to be used). Typical
3664 pthread implementations will either allow \f(CW\*(C`sigprocmask\*(C'\fR in the \*(L"main
3665 thread\*(R" or will block signals process-wide, both behaviours would
3666 be compatible with libev. Interaction between \f(CW\*(C`sigprocmask\*(C'\fR and
3667 \&\f(CW\*(C`pthread_sigmask\*(C'\fR could complicate things, however.
3668 .Sp
3669 The most portable way to handle signals is to block signals in all threads
3670 except the initial one, and run the default loop in the initial thread as
3671 well.
3672 .ie n .IP """long"" must be large enough for common memory allocation sizes" 4
3673 .el .IP "\f(CWlong\fR must be large enough for common memory allocation sizes" 4
3674 .IX Item "long must be large enough for common memory allocation sizes"
3675 To improve portability and simplify its \s-1API\s0, libev uses \f(CW\*(C`long\*(C'\fR internally
3676 instead of \f(CW\*(C`size_t\*(C'\fR when allocating its data structures. On non-POSIX
3677 systems (Microsoft...) this might be unexpectedly low, but is still at
3678 least 31 bits everywhere, which is enough for hundreds of millions of
3679 watchers.
3680 .ie n .IP """double"" must hold a time value in seconds with enough accuracy" 4
3681 .el .IP "\f(CWdouble\fR must hold a time value in seconds with enough accuracy" 4
3682 .IX Item "double must hold a time value in seconds with enough accuracy"
3683 The type \f(CW\*(C`double\*(C'\fR is used to represent timestamps. It is required to
3684 have at least 51 bits of mantissa (and 9 bits of exponent), which is good
3685 enough for at least into the year 4000. This requirement is fulfilled by
3686 implementations implementing \s-1IEEE\s0 754 (basically all existing ones).
3687 .PP
3688 If you know of other additional requirements drop me a note.
3689 .SH "ALGORITHMIC COMPLEXITIES"
3690 .IX Header "ALGORITHMIC COMPLEXITIES"
3691 In this section the complexities of (many of) the algorithms used inside
3692 libev will be documented. For complexity discussions about backends see
3693 the documentation for \f(CW\*(C`ev_default_init\*(C'\fR.
3694 .PP
3695 All of the following are about amortised time: If an array needs to be
3696 extended, libev needs to realloc and move the whole array, but this
3697 happens asymptotically rarer with higher number of elements, so O(1) might
3698 mean that libev does a lengthy realloc operation in rare cases, but on
3699 average it is much faster and asymptotically approaches constant time.
3700 .IP "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)" 4
3701 .IX Item "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)"
3702 This means that, when you have a watcher that triggers in one hour and
3703 there are 100 watchers that would trigger before that, then inserting will
3704 have to skip roughly seven (\f(CW\*(C`ld 100\*(C'\fR) of these watchers.
3705 .IP "Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)" 4
3706 .IX Item "Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)"
3707 That means that changing a timer costs less than removing/adding them,
3708 as only the relative motion in the event queue has to be paid for.
3709 .IP "Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)" 4
3710 .IX Item "Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)"
3711 These just add the watcher into an array or at the head of a list.
3712 .IP "Stopping check/prepare/idle/fork/async watchers: O(1)" 4
3713 .IX Item "Stopping check/prepare/idle/fork/async watchers: O(1)"
3714 .PD 0
3715 .IP "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % \s-1EV_PID_HASHSIZE\s0))" 4
3716 .IX Item "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))"
3717 .PD
3718 These watchers are stored in lists, so they need to be walked to find the
3719 correct watcher to remove. The lists are usually short (you don't usually
3720 have many watchers waiting for the same fd or signal: one is typical, two
3721 is rare).
3722 .IP "Finding the next timer in each loop iteration: O(1)" 4
3723 .IX Item "Finding the next timer in each loop iteration: O(1)"
3724 By virtue of using a binary or 4\-heap, the next timer is always found at a
3725 fixed position in the storage array.
3726 .IP "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)" 4
3727 .IX Item "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)"
3728 A change means an I/O watcher gets started or stopped, which requires
3729 libev to recalculate its status (and possibly tell the kernel, depending
3730 on backend and whether \f(CW\*(C`ev_io_set\*(C'\fR was used).
3731 .IP "Activating one watcher (putting it into the pending state): O(1)" 4
3732 .IX Item "Activating one watcher (putting it into the pending state): O(1)"
3733 .PD 0
3734 .IP "Priority handling: O(number_of_priorities)" 4
3735 .IX Item "Priority handling: O(number_of_priorities)"
3736 .PD
3737 Priorities are implemented by allocating some space for each
3738 priority. When doing priority-based operations, libev usually has to
3739 linearly search all the priorities, but starting/stopping and activating
3740 watchers becomes O(1) with respect to priority handling.
3741 .IP "Sending an ev_async: O(1)" 4
3742 .IX Item "Sending an ev_async: O(1)"
3743 .PD 0
3744 .IP "Processing ev_async_send: O(number_of_async_watchers)" 4
3745 .IX Item "Processing ev_async_send: O(number_of_async_watchers)"
3746 .IP "Processing signals: O(max_signal_number)" 4
3747 .IX Item "Processing signals: O(max_signal_number)"
3748 .PD
3749 Sending involves a system call \fIiff\fR there were no other \f(CW\*(C`ev_async_send\*(C'\fR
3750 calls in the current loop iteration. Checking for async and signal events
3751 involves iterating over all running async watchers or all signal numbers.
3752 .SH "AUTHOR"
3753 .IX Header "AUTHOR"
3754 Marc Lehmann <libev@schmorp.de>.