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

Comparing libev/ev.pod (file contents):
Revision 1.91 by root, Thu Dec 20 07:12:57 2007 UTC vs.
Revision 1.204 by root, Mon Oct 27 11:08:29 2008 UTC

2 2
3libev - a high performance full-featured event loop written in C 3libev - a high performance full-featured event loop written in C
4 4
5=head1 SYNOPSIS 5=head1 SYNOPSIS
6 6
7 #include <ev.h> 7 #include <ev.h>
8 8
9=head1 EXAMPLE PROGRAM 9=head2 EXAMPLE PROGRAM
10 10
11 // a single header file is required
11 #include <ev.h> 12 #include <ev.h>
12 13
14 // every watcher type has its own typedef'd struct
15 // with the name ev_TYPE
13 ev_io stdin_watcher; 16 ev_io stdin_watcher;
14 ev_timer timeout_watcher; 17 ev_timer timeout_watcher;
15 18
16 /* called when data readable on stdin */ 19 // all watcher callbacks have a similar signature
20 // this callback is called when data is readable on stdin
17 static void 21 static void
18 stdin_cb (EV_P_ struct ev_io *w, int revents) 22 stdin_cb (EV_P_ ev_io *w, int revents)
19 { 23 {
20 /* puts ("stdin ready"); */ 24 puts ("stdin ready");
21 ev_io_stop (EV_A_ w); /* just a syntax example */ 25 // for one-shot events, one must manually stop the watcher
22 ev_unloop (EV_A_ EVUNLOOP_ALL); /* leave all loop calls */ 26 // with its corresponding stop function.
27 ev_io_stop (EV_A_ w);
28
29 // this causes all nested ev_loop's to stop iterating
30 ev_unloop (EV_A_ EVUNLOOP_ALL);
23 } 31 }
24 32
33 // another callback, this time for a time-out
25 static void 34 static void
26 timeout_cb (EV_P_ struct ev_timer *w, int revents) 35 timeout_cb (EV_P_ ev_timer *w, int revents)
27 { 36 {
28 /* puts ("timeout"); */ 37 puts ("timeout");
29 ev_unloop (EV_A_ EVUNLOOP_ONE); /* leave one loop call */ 38 // this causes the innermost ev_loop to stop iterating
39 ev_unloop (EV_A_ EVUNLOOP_ONE);
30 } 40 }
31 41
32 int 42 int
33 main (void) 43 main (void)
34 { 44 {
45 // use the default event loop unless you have special needs
35 struct ev_loop *loop = ev_default_loop (0); 46 ev_loop *loop = ev_default_loop (0);
36 47
37 /* initialise an io watcher, then start it */ 48 // initialise an io watcher, then start it
49 // this one will watch for stdin to become readable
38 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ); 50 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
39 ev_io_start (loop, &stdin_watcher); 51 ev_io_start (loop, &stdin_watcher);
40 52
53 // initialise a timer watcher, then start it
41 /* simple non-repeating 5.5 second timeout */ 54 // simple non-repeating 5.5 second timeout
42 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.); 55 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
43 ev_timer_start (loop, &timeout_watcher); 56 ev_timer_start (loop, &timeout_watcher);
44 57
45 /* loop till timeout or data ready */ 58 // now wait for events to arrive
46 ev_loop (loop, 0); 59 ev_loop (loop, 0);
47 60
61 // unloop was called, so exit
48 return 0; 62 return 0;
49 } 63 }
50 64
51=head1 DESCRIPTION 65=head1 DESCRIPTION
52 66
53The newest version of this document is also available as a html-formatted 67The newest version of this document is also available as an html-formatted
54web page you might find easier to navigate when reading it for the first 68web page you might find easier to navigate when reading it for the first
55time: L<http://cvs.schmorp.de/libev/ev.html>. 69time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
56 70
57Libev is an event loop: you register interest in certain events (such as a 71Libev is an event loop: you register interest in certain events (such as a
58file descriptor being readable or a timeout occuring), and it will manage 72file descriptor being readable or a timeout occurring), and it will manage
59these event sources and provide your program with events. 73these event sources and provide your program with events.
60 74
61To do this, it must take more or less complete control over your process 75To do this, it must take more or less complete control over your process
62(or thread) by executing the I<event loop> handler, and will then 76(or thread) by executing the I<event loop> handler, and will then
63communicate events via a callback mechanism. 77communicate events via a callback mechanism.
65You register interest in certain events by registering so-called I<event 79You register interest in certain events by registering so-called I<event
66watchers>, which are relatively small C structures you initialise with the 80watchers>, which are relatively small C structures you initialise with the
67details of the event, and then hand it over to libev by I<starting> the 81details of the event, and then hand it over to libev by I<starting> the
68watcher. 82watcher.
69 83
70=head1 FEATURES 84=head2 FEATURES
71 85
72Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the 86Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
73BSD-specific C<kqueue> and the Solaris-specific event port mechanisms 87BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
74for file descriptor events (C<ev_io>), the Linux C<inotify> interface 88for file descriptor events (C<ev_io>), the Linux C<inotify> interface
75(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers 89(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
82 96
83It also is quite fast (see this 97It also is quite fast (see this
84L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent 98L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
85for example). 99for example).
86 100
87=head1 CONVENTIONS 101=head2 CONVENTIONS
88 102
89Libev is very configurable. In this manual the default configuration will 103Libev is very configurable. In this manual the default (and most common)
90be described, which supports multiple event loops. For more info about 104configuration will be described, which supports multiple event loops. For
91various configuration options please have a look at B<EMBED> section in 105more info about various configuration options please have a look at
92this manual. If libev was configured without support for multiple event 106B<EMBED> section in this manual. If libev was configured without support
93loops, then all functions taking an initial argument of name C<loop> 107for multiple event loops, then all functions taking an initial argument of
94(which is always of type C<struct ev_loop *>) will not have this argument. 108name C<loop> (which is always of type C<ev_loop *>) will not have
109this argument.
95 110
96=head1 TIME REPRESENTATION 111=head2 TIME REPRESENTATION
97 112
98Libev represents time as a single floating point number, representing the 113Libev represents time as a single floating point number, representing the
99(fractional) number of seconds since the (POSIX) epoch (somewhere near 114(fractional) number of seconds since the (POSIX) epoch (somewhere near
100the beginning of 1970, details are complicated, don't ask). This type is 115the beginning of 1970, details are complicated, don't ask). This type is
101called C<ev_tstamp>, which is what you should use too. It usually aliases 116called C<ev_tstamp>, which is what you should use too. It usually aliases
102to the C<double> type in C, and when you need to do any calculations on 117to the C<double> type in C, and when you need to do any calculations on
103it, you should treat it as some floatingpoint value. Unlike the name 118it, you should treat it as some floating point value. Unlike the name
104component C<stamp> might indicate, it is also used for time differences 119component C<stamp> might indicate, it is also used for time differences
105throughout libev. 120throughout libev.
121
122=head1 ERROR HANDLING
123
124Libev knows three classes of errors: operating system errors, usage errors
125and internal errors (bugs).
126
127When libev catches an operating system error it cannot handle (for example
128a system call indicating a condition libev cannot fix), it calls the callback
129set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
130abort. The default is to print a diagnostic message and to call C<abort
131()>.
132
133When libev detects a usage error such as a negative timer interval, then
134it will print a diagnostic message and abort (via the C<assert> mechanism,
135so C<NDEBUG> will disable this checking): these are programming errors in
136the libev caller and need to be fixed there.
137
138Libev also has a few internal error-checking C<assert>ions, and also has
139extensive consistency checking code. These do not trigger under normal
140circumstances, as they indicate either a bug in libev or worse.
141
106 142
107=head1 GLOBAL FUNCTIONS 143=head1 GLOBAL FUNCTIONS
108 144
109These functions can be called anytime, even before initialising the 145These functions can be called anytime, even before initialising the
110library in any way. 146library in any way.
114=item ev_tstamp ev_time () 150=item ev_tstamp ev_time ()
115 151
116Returns the current time as libev would use it. Please note that the 152Returns the current time as libev would use it. Please note that the
117C<ev_now> function is usually faster and also often returns the timestamp 153C<ev_now> function is usually faster and also often returns the timestamp
118you actually want to know. 154you actually want to know.
155
156=item ev_sleep (ev_tstamp interval)
157
158Sleep for the given interval: The current thread will be blocked until
159either it is interrupted or the given time interval has passed. Basically
160this is a sub-second-resolution C<sleep ()>.
119 161
120=item int ev_version_major () 162=item int ev_version_major ()
121 163
122=item int ev_version_minor () 164=item int ev_version_minor ()
123 165
136not a problem. 178not a problem.
137 179
138Example: Make sure we haven't accidentally been linked against the wrong 180Example: Make sure we haven't accidentally been linked against the wrong
139version. 181version.
140 182
141 assert (("libev version mismatch", 183 assert (("libev version mismatch",
142 ev_version_major () == EV_VERSION_MAJOR 184 ev_version_major () == EV_VERSION_MAJOR
143 && ev_version_minor () >= EV_VERSION_MINOR)); 185 && ev_version_minor () >= EV_VERSION_MINOR));
144 186
145=item unsigned int ev_supported_backends () 187=item unsigned int ev_supported_backends ()
146 188
147Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*> 189Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
148value) compiled into this binary of libev (independent of their 190value) compiled into this binary of libev (independent of their
150a description of the set values. 192a description of the set values.
151 193
152Example: make sure we have the epoll method, because yeah this is cool and 194Example: make sure we have the epoll method, because yeah this is cool and
153a must have and can we have a torrent of it please!!!11 195a must have and can we have a torrent of it please!!!11
154 196
155 assert (("sorry, no epoll, no sex", 197 assert (("sorry, no epoll, no sex",
156 ev_supported_backends () & EVBACKEND_EPOLL)); 198 ev_supported_backends () & EVBACKEND_EPOLL));
157 199
158=item unsigned int ev_recommended_backends () 200=item unsigned int ev_recommended_backends ()
159 201
160Return the set of all backends compiled into this binary of libev and also 202Return the set of all backends compiled into this binary of libev and also
161recommended for this platform. This set is often smaller than the one 203recommended for this platform. This set is often smaller than the one
162returned by C<ev_supported_backends>, as for example kqueue is broken on 204returned by C<ev_supported_backends>, as for example kqueue is broken on
163most BSDs and will not be autodetected unless you explicitly request it 205most BSDs and will not be auto-detected unless you explicitly request it
164(assuming you know what you are doing). This is the set of backends that 206(assuming you know what you are doing). This is the set of backends that
165libev will probe for if you specify no backends explicitly. 207libev will probe for if you specify no backends explicitly.
166 208
167=item unsigned int ev_embeddable_backends () 209=item unsigned int ev_embeddable_backends ()
168 210
172C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for 214C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
173recommended ones. 215recommended ones.
174 216
175See the description of C<ev_embed> watchers for more info. 217See the description of C<ev_embed> watchers for more info.
176 218
177=item ev_set_allocator (void *(*cb)(void *ptr, long size)) 219=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
178 220
179Sets the allocation function to use (the prototype is similar - the 221Sets the allocation function to use (the prototype is similar - the
180semantics is identical - to the realloc C function). It is used to 222semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
181allocate and free memory (no surprises here). If it returns zero when 223used to allocate and free memory (no surprises here). If it returns zero
182memory needs to be allocated, the library might abort or take some 224when memory needs to be allocated (C<size != 0>), the library might abort
183potentially destructive action. The default is your system realloc 225or take some potentially destructive action.
184function. 226
227Since some systems (at least OpenBSD and Darwin) fail to implement
228correct C<realloc> semantics, libev will use a wrapper around the system
229C<realloc> and C<free> functions by default.
185 230
186You could override this function in high-availability programs to, say, 231You could override this function in high-availability programs to, say,
187free some memory if it cannot allocate memory, to use a special allocator, 232free some memory if it cannot allocate memory, to use a special allocator,
188or even to sleep a while and retry until some memory is available. 233or even to sleep a while and retry until some memory is available.
189 234
190Example: Replace the libev allocator with one that waits a bit and then 235Example: Replace the libev allocator with one that waits a bit and then
191retries). 236retries (example requires a standards-compliant C<realloc>).
192 237
193 static void * 238 static void *
194 persistent_realloc (void *ptr, size_t size) 239 persistent_realloc (void *ptr, size_t size)
195 { 240 {
196 for (;;) 241 for (;;)
205 } 250 }
206 251
207 ... 252 ...
208 ev_set_allocator (persistent_realloc); 253 ev_set_allocator (persistent_realloc);
209 254
210=item ev_set_syserr_cb (void (*cb)(const char *msg)); 255=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
211 256
212Set the callback function to call on a retryable syscall error (such 257Set the callback function to call on a retryable system call error (such
213as failed select, poll, epoll_wait). The message is a printable string 258as failed select, poll, epoll_wait). The message is a printable string
214indicating the system call or subsystem causing the problem. If this 259indicating the system call or subsystem causing the problem. If this
215callback is set, then libev will expect it to remedy the sitution, no 260callback is set, then libev will expect it to remedy the situation, no
216matter what, when it returns. That is, libev will generally retry the 261matter what, when it returns. That is, libev will generally retry the
217requested operation, or, if the condition doesn't go away, do bad stuff 262requested operation, or, if the condition doesn't go away, do bad stuff
218(such as abort). 263(such as abort).
219 264
220Example: This is basically the same thing that libev does internally, too. 265Example: This is basically the same thing that libev does internally, too.
231 276
232=back 277=back
233 278
234=head1 FUNCTIONS CONTROLLING THE EVENT LOOP 279=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
235 280
236An event loop is described by a C<struct ev_loop *>. The library knows two 281An event loop is described by a C<struct ev_loop *> (the C<struct>
237types of such loops, the I<default> loop, which supports signals and child 282is I<not> optional in this case, as there is also an C<ev_loop>
238events, and dynamically created loops which do not. 283I<function>).
239 284
240If you use threads, a common model is to run the default event loop 285The library knows two types of such loops, the I<default> loop, which
241in your main thread (or in a separate thread) and for each thread you 286supports signals and child events, and dynamically created loops which do
242create, you also create another event loop. Libev itself does no locking 287not.
243whatsoever, so if you mix calls to the same event loop in different
244threads, make sure you lock (this is usually a bad idea, though, even if
245done correctly, because it's hideous and inefficient).
246 288
247=over 4 289=over 4
248 290
249=item struct ev_loop *ev_default_loop (unsigned int flags) 291=item struct ev_loop *ev_default_loop (unsigned int flags)
250 292
254flags. If that is troubling you, check C<ev_backend ()> afterwards). 296flags. If that is troubling you, check C<ev_backend ()> afterwards).
255 297
256If you don't know what event loop to use, use the one returned from this 298If you don't know what event loop to use, use the one returned from this
257function. 299function.
258 300
301Note that this function is I<not> thread-safe, so if you want to use it
302from multiple threads, you have to lock (note also that this is unlikely,
303as loops cannot bes hared easily between threads anyway).
304
305The default loop is the only loop that can handle C<ev_signal> and
306C<ev_child> watchers, and to do this, it always registers a handler
307for C<SIGCHLD>. If this is a problem for your application you can either
308create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
309can simply overwrite the C<SIGCHLD> signal handler I<after> calling
310C<ev_default_init>.
311
259The flags argument can be used to specify special behaviour or specific 312The flags argument can be used to specify special behaviour or specific
260backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>). 313backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
261 314
262The following flags are supported: 315The following flags are supported:
263 316
268The default flags value. Use this if you have no clue (it's the right 321The default flags value. Use this if you have no clue (it's the right
269thing, believe me). 322thing, believe me).
270 323
271=item C<EVFLAG_NOENV> 324=item C<EVFLAG_NOENV>
272 325
273If this flag bit is ored into the flag value (or the program runs setuid 326If this flag bit is or'ed into the flag value (or the program runs setuid
274or setgid) then libev will I<not> look at the environment variable 327or setgid) then libev will I<not> look at the environment variable
275C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will 328C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
276override the flags completely if it is found in the environment. This is 329override the flags completely if it is found in the environment. This is
277useful to try out specific backends to test their performance, or to work 330useful to try out specific backends to test their performance, or to work
278around bugs. 331around bugs.
284enabling this flag. 337enabling this flag.
285 338
286This works by calling C<getpid ()> on every iteration of the loop, 339This works by calling C<getpid ()> on every iteration of the loop,
287and thus this might slow down your event loop if you do a lot of loop 340and thus this might slow down your event loop if you do a lot of loop
288iterations and little real work, but is usually not noticeable (on my 341iterations and little real work, but is usually not noticeable (on my
289Linux system for example, C<getpid> is actually a simple 5-insn sequence 342GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
290without a syscall and thus I<very> fast, but my Linux system also has 343without a system call and thus I<very> fast, but my GNU/Linux system also has
291C<pthread_atfork> which is even faster). 344C<pthread_atfork> which is even faster).
292 345
293The big advantage of this flag is that you can forget about fork (and 346The big advantage of this flag is that you can forget about fork (and
294forget about forgetting to tell libev about forking) when you use this 347forget about forgetting to tell libev about forking) when you use this
295flag. 348flag.
296 349
297This flag setting cannot be overriden or specified in the C<LIBEV_FLAGS> 350This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
298environment variable. 351environment variable.
299 352
300=item C<EVBACKEND_SELECT> (value 1, portable select backend) 353=item C<EVBACKEND_SELECT> (value 1, portable select backend)
301 354
302This is your standard select(2) backend. Not I<completely> standard, as 355This is your standard select(2) backend. Not I<completely> standard, as
303libev tries to roll its own fd_set with no limits on the number of fds, 356libev tries to roll its own fd_set with no limits on the number of fds,
304but if that fails, expect a fairly low limit on the number of fds when 357but if that fails, expect a fairly low limit on the number of fds when
305using this backend. It doesn't scale too well (O(highest_fd)), but its usually 358using this backend. It doesn't scale too well (O(highest_fd)), but its
306the fastest backend for a low number of fds. 359usually the fastest backend for a low number of (low-numbered :) fds.
360
361To get good performance out of this backend you need a high amount of
362parallelism (most of the file descriptors should be busy). If you are
363writing a server, you should C<accept ()> in a loop to accept as many
364connections as possible during one iteration. You might also want to have
365a look at C<ev_set_io_collect_interval ()> to increase the amount of
366readiness notifications you get per iteration.
367
368This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
369C<writefds> set (and to work around Microsoft Windows bugs, also onto the
370C<exceptfds> set on that platform).
307 371
308=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows) 372=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
309 373
310And this is your standard poll(2) backend. It's more complicated than 374And this is your standard poll(2) backend. It's more complicated
311select, but handles sparse fds better and has no artificial limit on the 375than select, but handles sparse fds better and has no artificial
312number of fds you can use (except it will slow down considerably with a 376limit on the number of fds you can use (except it will slow down
313lot of inactive fds). It scales similarly to select, i.e. O(total_fds). 377considerably with a lot of inactive fds). It scales similarly to select,
378i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
379performance tips.
380
381This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
382C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
314 383
315=item C<EVBACKEND_EPOLL> (value 4, Linux) 384=item C<EVBACKEND_EPOLL> (value 4, Linux)
316 385
317For few fds, this backend is a bit little slower than poll and select, 386For few fds, this backend is a bit little slower than poll and select,
318but it scales phenomenally better. While poll and select usually scale like 387but it scales phenomenally better. While poll and select usually scale
319O(total_fds) where n is the total number of fds (or the highest fd), epoll scales 388like O(total_fds) where n is the total number of fds (or the highest fd),
320either O(1) or O(active_fds). 389epoll scales either O(1) or O(active_fds). The epoll design has a number
390of shortcomings, such as silently dropping events in some hard-to-detect
391cases and requiring a system call per fd change, no fork support and bad
392support for dup.
321 393
394Epoll is also notoriously buggy - embedding epoll fds should work, but
395of course doesn't, and epoll just loves to report events for totally
396I<different> file descriptors (even already closed ones) than registered
397in the set (especially on SMP systems). Libev tries to counter these
398spurious notifications by employing an additional generation counter and
399comparing that against the events to filter out spurious ones.
400
322While stopping and starting an I/O watcher in the same iteration will 401While stopping, setting and starting an I/O watcher in the same iteration
323result in some caching, there is still a syscall per such incident 402will result in some caching, there is still a system call per such incident
324(because the fd could point to a different file description now), so its 403(because the fd could point to a different file description now), so its
325best to avoid that. Also, dup()ed file descriptors might not work very 404best to avoid that. Also, C<dup ()>'ed file descriptors might not work
326well if you register events for both fds. 405very well if you register events for both fds.
327 406
328Please note that epoll sometimes generates spurious notifications, so you 407Best performance from this backend is achieved by not unregistering all
329need to use non-blocking I/O or other means to avoid blocking when no data 408watchers for a file descriptor until it has been closed, if possible,
330(or space) is available. 409i.e. keep at least one watcher active per fd at all times. Stopping and
410starting a watcher (without re-setting it) also usually doesn't cause
411extra overhead.
412
413While nominally embeddable in other event loops, this feature is broken in
414all kernel versions tested so far.
415
416This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
417C<EVBACKEND_POLL>.
331 418
332=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones) 419=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
333 420
334Kqueue deserves special mention, as at the time of this writing, it 421Kqueue deserves special mention, as at the time of this writing, it was
335was broken on all BSDs except NetBSD (usually it doesn't work with 422broken on all BSDs except NetBSD (usually it doesn't work reliably with
336anything but sockets and pipes, except on Darwin, where of course it's 423anything but sockets and pipes, except on Darwin, where of course it's
337completely useless). For this reason it's not being "autodetected" 424completely useless). For this reason it's not being "auto-detected" unless
338unless you explicitly specify it explicitly in the flags (i.e. using 425you explicitly specify it in the flags (i.e. using C<EVBACKEND_KQUEUE>) or
339C<EVBACKEND_KQUEUE>). 426libev was compiled on a known-to-be-good (-enough) system like NetBSD.
427
428You still can embed kqueue into a normal poll or select backend and use it
429only for sockets (after having made sure that sockets work with kqueue on
430the target platform). See C<ev_embed> watchers for more info.
340 431
341It scales in the same way as the epoll backend, but the interface to the 432It scales in the same way as the epoll backend, but the interface to the
342kernel is more efficient (which says nothing about its actual speed, of 433kernel is more efficient (which says nothing about its actual speed, of
343course). While starting and stopping an I/O watcher does not cause an 434course). While stopping, setting and starting an I/O watcher does never
344extra syscall as with epoll, it still adds up to four event changes per 435cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
345incident, so its best to avoid that. 436two event changes per incident. Support for C<fork ()> is very bad and it
437drops fds silently in similarly hard-to-detect cases.
438
439This backend usually performs well under most conditions.
440
441While nominally embeddable in other event loops, this doesn't work
442everywhere, so you might need to test for this. And since it is broken
443almost everywhere, you should only use it when you have a lot of sockets
444(for which it usually works), by embedding it into another event loop
445(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and, did I mention it,
446using it only for sockets.
447
448This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
449C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
450C<NOTE_EOF>.
346 451
347=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8) 452=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
348 453
349This is not implemented yet (and might never be). 454This is not implemented yet (and might never be, unless you send me an
455implementation). According to reports, C</dev/poll> only supports sockets
456and is not embeddable, which would limit the usefulness of this backend
457immensely.
350 458
351=item C<EVBACKEND_PORT> (value 32, Solaris 10) 459=item C<EVBACKEND_PORT> (value 32, Solaris 10)
352 460
353This uses the Solaris 10 port mechanism. As with everything on Solaris, 461This uses the Solaris 10 event port mechanism. As with everything on Solaris,
354it's really slow, but it still scales very well (O(active_fds)). 462it's really slow, but it still scales very well (O(active_fds)).
355 463
356Please note that solaris ports can result in a lot of spurious 464Please note that Solaris event ports can deliver a lot of spurious
357notifications, so you need to use non-blocking I/O or other means to avoid 465notifications, so you need to use non-blocking I/O or other means to avoid
358blocking when no data (or space) is available. 466blocking when no data (or space) is available.
467
468While this backend scales well, it requires one system call per active
469file descriptor per loop iteration. For small and medium numbers of file
470descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
471might perform better.
472
473On the positive side, with the exception of the spurious readiness
474notifications, this backend actually performed fully to specification
475in all tests and is fully embeddable, which is a rare feat among the
476OS-specific backends.
477
478This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
479C<EVBACKEND_POLL>.
359 480
360=item C<EVBACKEND_ALL> 481=item C<EVBACKEND_ALL>
361 482
362Try all backends (even potentially broken ones that wouldn't be tried 483Try all backends (even potentially broken ones that wouldn't be tried
363with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as 484with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
364C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>. 485C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
365 486
487It is definitely not recommended to use this flag.
488
366=back 489=back
367 490
368If one or more of these are ored into the flags value, then only these 491If one or more of these are or'ed into the flags value, then only these
369backends will be tried (in the reverse order as given here). If none are 492backends will be tried (in the reverse order as listed here). If none are
370specified, most compiled-in backend will be tried, usually in reverse 493specified, all backends in C<ev_recommended_backends ()> will be tried.
371order of their flag values :)
372 494
373The most typical usage is like this: 495Example: This is the most typical usage.
374 496
375 if (!ev_default_loop (0)) 497 if (!ev_default_loop (0))
376 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?"); 498 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
377 499
378Restrict libev to the select and poll backends, and do not allow 500Example: Restrict libev to the select and poll backends, and do not allow
379environment settings to be taken into account: 501environment settings to be taken into account:
380 502
381 ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV); 503 ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
382 504
383Use whatever libev has to offer, but make sure that kqueue is used if 505Example: Use whatever libev has to offer, but make sure that kqueue is
384available (warning, breaks stuff, best use only with your own private 506used if available (warning, breaks stuff, best use only with your own
385event loop and only if you know the OS supports your types of fds): 507private event loop and only if you know the OS supports your types of
508fds):
386 509
387 ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE); 510 ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
388 511
389=item struct ev_loop *ev_loop_new (unsigned int flags) 512=item struct ev_loop *ev_loop_new (unsigned int flags)
390 513
391Similar to C<ev_default_loop>, but always creates a new event loop that is 514Similar to C<ev_default_loop>, but always creates a new event loop that is
392always distinct from the default loop. Unlike the default loop, it cannot 515always distinct from the default loop. Unlike the default loop, it cannot
393handle signal and child watchers, and attempts to do so will be greeted by 516handle signal and child watchers, and attempts to do so will be greeted by
394undefined behaviour (or a failed assertion if assertions are enabled). 517undefined behaviour (or a failed assertion if assertions are enabled).
395 518
519Note that this function I<is> thread-safe, and the recommended way to use
520libev with threads is indeed to create one loop per thread, and using the
521default loop in the "main" or "initial" thread.
522
396Example: Try to create a event loop that uses epoll and nothing else. 523Example: Try to create a event loop that uses epoll and nothing else.
397 524
398 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV); 525 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
399 if (!epoller) 526 if (!epoller)
400 fatal ("no epoll found here, maybe it hides under your chair"); 527 fatal ("no epoll found here, maybe it hides under your chair");
401 528
402=item ev_default_destroy () 529=item ev_default_destroy ()
403 530
404Destroys the default loop again (frees all memory and kernel state 531Destroys the default loop again (frees all memory and kernel state
405etc.). None of the active event watchers will be stopped in the normal 532etc.). None of the active event watchers will be stopped in the normal
406sense, so e.g. C<ev_is_active> might still return true. It is your 533sense, so e.g. C<ev_is_active> might still return true. It is your
407responsibility to either stop all watchers cleanly yoursef I<before> 534responsibility to either stop all watchers cleanly yourself I<before>
408calling this function, or cope with the fact afterwards (which is usually 535calling this function, or cope with the fact afterwards (which is usually
409the easiest thing, you can just ignore the watchers and/or C<free ()> them 536the easiest thing, you can just ignore the watchers and/or C<free ()> them
410for example). 537for example).
411 538
412Note that certain global state, such as signal state, will not be freed by 539Note that certain global state, such as signal state (and installed signal
413this function, and related watchers (such as signal and child watchers) 540handlers), will not be freed by this function, and related watchers (such
414would need to be stopped manually. 541as signal and child watchers) would need to be stopped manually.
415 542
416In general it is not advisable to call this function except in the 543In general it is not advisable to call this function except in the
417rare occasion where you really need to free e.g. the signal handling 544rare occasion where you really need to free e.g. the signal handling
418pipe fds. If you need dynamically allocated loops it is better to use 545pipe fds. If you need dynamically allocated loops it is better to use
419C<ev_loop_new> and C<ev_loop_destroy>). 546C<ev_loop_new> and C<ev_loop_destroy>).
423Like C<ev_default_destroy>, but destroys an event loop created by an 550Like C<ev_default_destroy>, but destroys an event loop created by an
424earlier call to C<ev_loop_new>. 551earlier call to C<ev_loop_new>.
425 552
426=item ev_default_fork () 553=item ev_default_fork ()
427 554
555This function sets a flag that causes subsequent C<ev_loop> iterations
428This function reinitialises the kernel state for backends that have 556to reinitialise the kernel state for backends that have one. Despite the
429one. Despite the name, you can call it anytime, but it makes most sense 557name, you can call it anytime, but it makes most sense after forking, in
430after forking, in either the parent or child process (or both, but that 558the child process (or both child and parent, but that again makes little
431again makes little sense). 559sense). You I<must> call it in the child before using any of the libev
560functions, and it will only take effect at the next C<ev_loop> iteration.
432 561
433You I<must> call this function in the child process after forking if and 562On the other hand, you only need to call this function in the child
434only if you want to use the event library in both processes. If you just 563process if and only if you want to use the event library in the child. If
435fork+exec, you don't have to call it. 564you just fork+exec, you don't have to call it at all.
436 565
437The function itself is quite fast and it's usually not a problem to call 566The function itself is quite fast and it's usually not a problem to call
438it just in case after a fork. To make this easy, the function will fit in 567it just in case after a fork. To make this easy, the function will fit in
439quite nicely into a call to C<pthread_atfork>: 568quite nicely into a call to C<pthread_atfork>:
440 569
441 pthread_atfork (0, 0, ev_default_fork); 570 pthread_atfork (0, 0, ev_default_fork);
442 571
443At the moment, C<EVBACKEND_SELECT> and C<EVBACKEND_POLL> are safe to use
444without calling this function, so if you force one of those backends you
445do not need to care.
446
447=item ev_loop_fork (loop) 572=item ev_loop_fork (loop)
448 573
449Like C<ev_default_fork>, but acts on an event loop created by 574Like C<ev_default_fork>, but acts on an event loop created by
450C<ev_loop_new>. Yes, you have to call this on every allocated event loop 575C<ev_loop_new>. Yes, you have to call this on every allocated event loop
451after fork, and how you do this is entirely your own problem. 576after fork that you want to re-use in the child, and how you do this is
577entirely your own problem.
578
579=item int ev_is_default_loop (loop)
580
581Returns true when the given loop is, in fact, the default loop, and false
582otherwise.
452 583
453=item unsigned int ev_loop_count (loop) 584=item unsigned int ev_loop_count (loop)
454 585
455Returns the count of loop iterations for the loop, which is identical to 586Returns the count of loop iterations for the loop, which is identical to
456the number of times libev did poll for new events. It starts at C<0> and 587the number of times libev did poll for new events. It starts at C<0> and
469 600
470Returns the current "event loop time", which is the time the event loop 601Returns the current "event loop time", which is the time the event loop
471received events and started processing them. This timestamp does not 602received events and started processing them. This timestamp does not
472change as long as callbacks are being processed, and this is also the base 603change as long as callbacks are being processed, and this is also the base
473time used for relative timers. You can treat it as the timestamp of the 604time used for relative timers. You can treat it as the timestamp of the
474event occuring (or more correctly, libev finding out about it). 605event occurring (or more correctly, libev finding out about it).
606
607=item ev_now_update (loop)
608
609Establishes the current time by querying the kernel, updating the time
610returned by C<ev_now ()> in the progress. This is a costly operation and
611is usually done automatically within C<ev_loop ()>.
612
613This function is rarely useful, but when some event callback runs for a
614very long time without entering the event loop, updating libev's idea of
615the current time is a good idea.
616
617See also "The special problem of time updates" in the C<ev_timer> section.
475 618
476=item ev_loop (loop, int flags) 619=item ev_loop (loop, int flags)
477 620
478Finally, this is it, the event handler. This function usually is called 621Finally, this is it, the event handler. This function usually is called
479after you initialised all your watchers and you want to start handling 622after you initialised all your watchers and you want to start handling
482If the flags argument is specified as C<0>, it will not return until 625If the flags argument is specified as C<0>, it will not return until
483either no event watchers are active anymore or C<ev_unloop> was called. 626either no event watchers are active anymore or C<ev_unloop> was called.
484 627
485Please note that an explicit C<ev_unloop> is usually better than 628Please note that an explicit C<ev_unloop> is usually better than
486relying on all watchers to be stopped when deciding when a program has 629relying on all watchers to be stopped when deciding when a program has
487finished (especially in interactive programs), but having a program that 630finished (especially in interactive programs), but having a program
488automatically loops as long as it has to and no longer by virtue of 631that automatically loops as long as it has to and no longer by virtue
489relying on its watchers stopping correctly is a thing of beauty. 632of relying on its watchers stopping correctly, that is truly a thing of
633beauty.
490 634
491A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle 635A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
492those events and any outstanding ones, but will not block your process in 636those events and any already outstanding ones, but will not block your
493case there are no events and will return after one iteration of the loop. 637process in case there are no events and will return after one iteration of
638the loop.
494 639
495A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if 640A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
496neccessary) and will handle those and any outstanding ones. It will block 641necessary) and will handle those and any already outstanding ones. It
497your process until at least one new event arrives, and will return after 642will block your process until at least one new event arrives (which could
498one iteration of the loop. This is useful if you are waiting for some 643be an event internal to libev itself, so there is no guarentee that a
499external event in conjunction with something not expressible using other 644user-registered callback will be called), and will return after one
645iteration of the loop.
646
647This is useful if you are waiting for some external event in conjunction
648with something not expressible using other libev watchers (i.e. "roll your
500libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is 649own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
501usually a better approach for this kind of thing. 650usually a better approach for this kind of thing.
502 651
503Here are the gory details of what C<ev_loop> does: 652Here are the gory details of what C<ev_loop> does:
504 653
505 - Before the first iteration, call any pending watchers. 654 - Before the first iteration, call any pending watchers.
506 * If there are no active watchers (reference count is zero), return. 655 * If EVFLAG_FORKCHECK was used, check for a fork.
507 - Queue all prepare watchers and then call all outstanding watchers. 656 - If a fork was detected (by any means), queue and call all fork watchers.
657 - Queue and call all prepare watchers.
508 - If we have been forked, recreate the kernel state. 658 - If we have been forked, detach and recreate the kernel state
659 as to not disturb the other process.
509 - Update the kernel state with all outstanding changes. 660 - Update the kernel state with all outstanding changes.
510 - Update the "event loop time". 661 - Update the "event loop time" (ev_now ()).
511 - Calculate for how long to block. 662 - Calculate for how long to sleep or block, if at all
663 (active idle watchers, EVLOOP_NONBLOCK or not having
664 any active watchers at all will result in not sleeping).
665 - Sleep if the I/O and timer collect interval say so.
512 - Block the process, waiting for any events. 666 - Block the process, waiting for any events.
513 - Queue all outstanding I/O (fd) events. 667 - Queue all outstanding I/O (fd) events.
514 - Update the "event loop time" and do time jump handling. 668 - Update the "event loop time" (ev_now ()), and do time jump adjustments.
515 - Queue all outstanding timers. 669 - Queue all expired timers.
516 - Queue all outstanding periodics. 670 - Queue all expired periodics.
517 - If no events are pending now, queue all idle watchers. 671 - Unless any events are pending now, queue all idle watchers.
518 - Queue all check watchers. 672 - Queue all check watchers.
519 - Call all queued watchers in reverse order (i.e. check watchers first). 673 - Call all queued watchers in reverse order (i.e. check watchers first).
520 Signals and child watchers are implemented as I/O watchers, and will 674 Signals and child watchers are implemented as I/O watchers, and will
521 be handled here by queueing them when their watcher gets executed. 675 be handled here by queueing them when their watcher gets executed.
522 - If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK 676 - If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
523 were used, return, otherwise continue with step *. 677 were used, or there are no active watchers, return, otherwise
678 continue with step *.
524 679
525Example: Queue some jobs and then loop until no events are outsanding 680Example: Queue some jobs and then loop until no events are outstanding
526anymore. 681anymore.
527 682
528 ... queue jobs here, make sure they register event watchers as long 683 ... queue jobs here, make sure they register event watchers as long
529 ... as they still have work to do (even an idle watcher will do..) 684 ... as they still have work to do (even an idle watcher will do..)
530 ev_loop (my_loop, 0); 685 ev_loop (my_loop, 0);
531 ... jobs done. yeah! 686 ... jobs done or somebody called unloop. yeah!
532 687
533=item ev_unloop (loop, how) 688=item ev_unloop (loop, how)
534 689
535Can be used to make a call to C<ev_loop> return early (but only after it 690Can be used to make a call to C<ev_loop> return early (but only after it
536has processed all outstanding events). The C<how> argument must be either 691has processed all outstanding events). The C<how> argument must be either
537C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or 692C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
538C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return. 693C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
539 694
695This "unloop state" will be cleared when entering C<ev_loop> again.
696
697It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
698
540=item ev_ref (loop) 699=item ev_ref (loop)
541 700
542=item ev_unref (loop) 701=item ev_unref (loop)
543 702
544Ref/unref can be used to add or remove a reference count on the event 703Ref/unref can be used to add or remove a reference count on the event
545loop: Every watcher keeps one reference, and as long as the reference 704loop: Every watcher keeps one reference, and as long as the reference
546count is nonzero, C<ev_loop> will not return on its own. If you have 705count is nonzero, C<ev_loop> will not return on its own.
706
547a watcher you never unregister that should not keep C<ev_loop> from 707If you have a watcher you never unregister that should not keep C<ev_loop>
548returning, ev_unref() after starting, and ev_ref() before stopping it. For 708from returning, call ev_unref() after starting, and ev_ref() before
709stopping it.
710
549example, libev itself uses this for its internal signal pipe: It is not 711As an example, libev itself uses this for its internal signal pipe: It is
550visible to the libev user and should not keep C<ev_loop> from exiting if 712not visible to the libev user and should not keep C<ev_loop> from exiting
551no event watchers registered by it are active. It is also an excellent 713if no event watchers registered by it are active. It is also an excellent
552way to do this for generic recurring timers or from within third-party 714way to do this for generic recurring timers or from within third-party
553libraries. Just remember to I<unref after start> and I<ref before stop>. 715libraries. Just remember to I<unref after start> and I<ref before stop>
716(but only if the watcher wasn't active before, or was active before,
717respectively).
554 718
555Example: Create a signal watcher, but keep it from keeping C<ev_loop> 719Example: Create a signal watcher, but keep it from keeping C<ev_loop>
556running when nothing else is active. 720running when nothing else is active.
557 721
558 struct ev_signal exitsig; 722 ev_signal exitsig;
559 ev_signal_init (&exitsig, sig_cb, SIGINT); 723 ev_signal_init (&exitsig, sig_cb, SIGINT);
560 ev_signal_start (loop, &exitsig); 724 ev_signal_start (loop, &exitsig);
561 evf_unref (loop); 725 evf_unref (loop);
562 726
563Example: For some weird reason, unregister the above signal handler again. 727Example: For some weird reason, unregister the above signal handler again.
564 728
565 ev_ref (loop); 729 ev_ref (loop);
566 ev_signal_stop (loop, &exitsig); 730 ev_signal_stop (loop, &exitsig);
731
732=item ev_set_io_collect_interval (loop, ev_tstamp interval)
733
734=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
735
736These advanced functions influence the time that libev will spend waiting
737for events. Both time intervals are by default C<0>, meaning that libev
738will try to invoke timer/periodic callbacks and I/O callbacks with minimum
739latency.
740
741Setting these to a higher value (the C<interval> I<must> be >= C<0>)
742allows libev to delay invocation of I/O and timer/periodic callbacks
743to increase efficiency of loop iterations (or to increase power-saving
744opportunities).
745
746The idea is that sometimes your program runs just fast enough to handle
747one (or very few) event(s) per loop iteration. While this makes the
748program responsive, it also wastes a lot of CPU time to poll for new
749events, especially with backends like C<select ()> which have a high
750overhead for the actual polling but can deliver many events at once.
751
752By setting a higher I<io collect interval> you allow libev to spend more
753time collecting I/O events, so you can handle more events per iteration,
754at the cost of increasing latency. Timeouts (both C<ev_periodic> and
755C<ev_timer>) will be not affected. Setting this to a non-null value will
756introduce an additional C<ev_sleep ()> call into most loop iterations.
757
758Likewise, by setting a higher I<timeout collect interval> you allow libev
759to spend more time collecting timeouts, at the expense of increased
760latency/jitter/inexactness (the watcher callback will be called
761later). C<ev_io> watchers will not be affected. Setting this to a non-null
762value will not introduce any overhead in libev.
763
764Many (busy) programs can usually benefit by setting the I/O collect
765interval to a value near C<0.1> or so, which is often enough for
766interactive servers (of course not for games), likewise for timeouts. It
767usually doesn't make much sense to set it to a lower value than C<0.01>,
768as this approaches the timing granularity of most systems.
769
770Setting the I<timeout collect interval> can improve the opportunity for
771saving power, as the program will "bundle" timer callback invocations that
772are "near" in time together, by delaying some, thus reducing the number of
773times the process sleeps and wakes up again. Another useful technique to
774reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
775they fire on, say, one-second boundaries only.
776
777=item ev_loop_verify (loop)
778
779This function only does something when C<EV_VERIFY> support has been
780compiled in, which is the default for non-minimal builds. It tries to go
781through all internal structures and checks them for validity. If anything
782is found to be inconsistent, it will print an error message to standard
783error and call C<abort ()>.
784
785This can be used to catch bugs inside libev itself: under normal
786circumstances, this function will never abort as of course libev keeps its
787data structures consistent.
567 788
568=back 789=back
569 790
570 791
571=head1 ANATOMY OF A WATCHER 792=head1 ANATOMY OF A WATCHER
793
794In the following description, uppercase C<TYPE> in names stands for the
795watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
796watchers and C<ev_io_start> for I/O watchers.
572 797
573A watcher is a structure that you create and register to record your 798A watcher is a structure that you create and register to record your
574interest in some event. For instance, if you want to wait for STDIN to 799interest in some event. For instance, if you want to wait for STDIN to
575become readable, you would create an C<ev_io> watcher for that: 800become readable, you would create an C<ev_io> watcher for that:
576 801
577 static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents) 802 static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
578 { 803 {
579 ev_io_stop (w); 804 ev_io_stop (w);
580 ev_unloop (loop, EVUNLOOP_ALL); 805 ev_unloop (loop, EVUNLOOP_ALL);
581 } 806 }
582 807
583 struct ev_loop *loop = ev_default_loop (0); 808 struct ev_loop *loop = ev_default_loop (0);
809
584 struct ev_io stdin_watcher; 810 ev_io stdin_watcher;
811
585 ev_init (&stdin_watcher, my_cb); 812 ev_init (&stdin_watcher, my_cb);
586 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ); 813 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
587 ev_io_start (loop, &stdin_watcher); 814 ev_io_start (loop, &stdin_watcher);
815
588 ev_loop (loop, 0); 816 ev_loop (loop, 0);
589 817
590As you can see, you are responsible for allocating the memory for your 818As you can see, you are responsible for allocating the memory for your
591watcher structures (and it is usually a bad idea to do this on the stack, 819watcher structures (and it is I<usually> a bad idea to do this on the
592although this can sometimes be quite valid). 820stack).
821
822Each watcher has an associated watcher structure (called C<struct ev_TYPE>
823or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
593 824
594Each watcher structure must be initialised by a call to C<ev_init 825Each watcher structure must be initialised by a call to C<ev_init
595(watcher *, callback)>, which expects a callback to be provided. This 826(watcher *, callback)>, which expects a callback to be provided. This
596callback gets invoked each time the event occurs (or, in the case of io 827callback gets invoked each time the event occurs (or, in the case of I/O
597watchers, each time the event loop detects that the file descriptor given 828watchers, each time the event loop detects that the file descriptor given
598is readable and/or writable). 829is readable and/or writable).
599 830
600Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro 831Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
601with arguments specific to this watcher type. There is also a macro 832macro to configure it, with arguments specific to the watcher type. There
602to combine initialisation and setting in one call: C<< ev_<type>_init 833is also a macro to combine initialisation and setting in one call: C<<
603(watcher *, callback, ...) >>. 834ev_TYPE_init (watcher *, callback, ...) >>.
604 835
605To make the watcher actually watch out for events, you have to start it 836To make the watcher actually watch out for events, you have to start it
606with a watcher-specific start function (C<< ev_<type>_start (loop, watcher 837with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
607*) >>), and you can stop watching for events at any time by calling the 838*) >>), and you can stop watching for events at any time by calling the
608corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>. 839corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
609 840
610As long as your watcher is active (has been started but not stopped) you 841As long as your watcher is active (has been started but not stopped) you
611must not touch the values stored in it. Most specifically you must never 842must not touch the values stored in it. Most specifically you must never
612reinitialise it or call its C<set> macro. 843reinitialise it or call its C<ev_TYPE_set> macro.
613 844
614Each and every callback receives the event loop pointer as first, the 845Each and every callback receives the event loop pointer as first, the
615registered watcher structure as second, and a bitset of received events as 846registered watcher structure as second, and a bitset of received events as
616third argument. 847third argument.
617 848
671=item C<EV_FORK> 902=item C<EV_FORK>
672 903
673The event loop has been resumed in the child process after fork (see 904The event loop has been resumed in the child process after fork (see
674C<ev_fork>). 905C<ev_fork>).
675 906
907=item C<EV_ASYNC>
908
909The given async watcher has been asynchronously notified (see C<ev_async>).
910
676=item C<EV_ERROR> 911=item C<EV_ERROR>
677 912
678An unspecified error has occured, the watcher has been stopped. This might 913An unspecified error has occurred, the watcher has been stopped. This might
679happen because the watcher could not be properly started because libev 914happen because the watcher could not be properly started because libev
680ran out of memory, a file descriptor was found to be closed or any other 915ran out of memory, a file descriptor was found to be closed or any other
916problem. Libev considers these application bugs.
917
681problem. You best act on it by reporting the problem and somehow coping 918You best act on it by reporting the problem and somehow coping with the
682with the watcher being stopped. 919watcher being stopped. Note that well-written programs should not receive
920an error ever, so when your watcher receives it, this usually indicates a
921bug in your program.
683 922
684Libev will usually signal a few "dummy" events together with an error, 923Libev will usually signal a few "dummy" events together with an error, for
685for example it might indicate that a fd is readable or writable, and if 924example it might indicate that a fd is readable or writable, and if your
686your callbacks is well-written it can just attempt the operation and cope 925callbacks is well-written it can just attempt the operation and cope with
687with the error from read() or write(). This will not work in multithreaded 926the error from read() or write(). This will not work in multi-threaded
688programs, though, so beware. 927programs, though, as the fd could already be closed and reused for another
928thing, so beware.
689 929
690=back 930=back
691 931
692=head2 GENERIC WATCHER FUNCTIONS 932=head2 GENERIC WATCHER FUNCTIONS
693
694In the following description, C<TYPE> stands for the watcher type,
695e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
696 933
697=over 4 934=over 4
698 935
699=item C<ev_init> (ev_TYPE *watcher, callback) 936=item C<ev_init> (ev_TYPE *watcher, callback)
700 937
706which rolls both calls into one. 943which rolls both calls into one.
707 944
708You can reinitialise a watcher at any time as long as it has been stopped 945You can reinitialise a watcher at any time as long as it has been stopped
709(or never started) and there are no pending events outstanding. 946(or never started) and there are no pending events outstanding.
710 947
711The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher, 948The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
712int revents)>. 949int revents)>.
950
951Example: Initialise an C<ev_io> watcher in two steps.
952
953 ev_io w;
954 ev_init (&w, my_cb);
955 ev_io_set (&w, STDIN_FILENO, EV_READ);
713 956
714=item C<ev_TYPE_set> (ev_TYPE *, [args]) 957=item C<ev_TYPE_set> (ev_TYPE *, [args])
715 958
716This macro initialises the type-specific parts of a watcher. You need to 959This macro initialises the type-specific parts of a watcher. You need to
717call C<ev_init> at least once before you call this macro, but you can 960call C<ev_init> at least once before you call this macro, but you can
720difference to the C<ev_init> macro). 963difference to the C<ev_init> macro).
721 964
722Although some watcher types do not have type-specific arguments 965Although some watcher types do not have type-specific arguments
723(e.g. C<ev_prepare>) you still need to call its C<set> macro. 966(e.g. C<ev_prepare>) you still need to call its C<set> macro.
724 967
968See C<ev_init>, above, for an example.
969
725=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args]) 970=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
726 971
727This convinience macro rolls both C<ev_init> and C<ev_TYPE_set> macro 972This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
728calls into a single call. This is the most convinient method to initialise 973calls into a single call. This is the most convenient method to initialise
729a watcher. The same limitations apply, of course. 974a watcher. The same limitations apply, of course.
975
976Example: Initialise and set an C<ev_io> watcher in one step.
977
978 ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
730 979
731=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher) 980=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
732 981
733Starts (activates) the given watcher. Only active watchers will receive 982Starts (activates) the given watcher. Only active watchers will receive
734events. If the watcher is already active nothing will happen. 983events. If the watcher is already active nothing will happen.
735 984
985Example: Start the C<ev_io> watcher that is being abused as example in this
986whole section.
987
988 ev_io_start (EV_DEFAULT_UC, &w);
989
736=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher) 990=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
737 991
738Stops the given watcher again (if active) and clears the pending 992Stops the given watcher if active, and clears the pending status (whether
993the watcher was active or not).
994
739status. It is possible that stopped watchers are pending (for example, 995It is possible that stopped watchers are pending - for example,
740non-repeating timers are being stopped when they become pending), but 996non-repeating timers are being stopped when they become pending - but
741C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If 997calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
742you want to free or reuse the memory used by the watcher it is therefore a 998pending. If you want to free or reuse the memory used by the watcher it is
743good idea to always call its C<ev_TYPE_stop> function. 999therefore a good idea to always call its C<ev_TYPE_stop> function.
744 1000
745=item bool ev_is_active (ev_TYPE *watcher) 1001=item bool ev_is_active (ev_TYPE *watcher)
746 1002
747Returns a true value iff the watcher is active (i.e. it has been started 1003Returns a true value iff the watcher is active (i.e. it has been started
748and not yet been stopped). As long as a watcher is active you must not modify 1004and not yet been stopped). As long as a watcher is active you must not modify
790The default priority used by watchers when no priority has been set is 1046The default priority used by watchers when no priority has been set is
791always C<0>, which is supposed to not be too high and not be too low :). 1047always C<0>, which is supposed to not be too high and not be too low :).
792 1048
793Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is 1049Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
794fine, as long as you do not mind that the priority value you query might 1050fine, as long as you do not mind that the priority value you query might
795or might not have been adjusted to be within valid range. 1051or might not have been clamped to the valid range.
796 1052
797=item ev_invoke (loop, ev_TYPE *watcher, int revents) 1053=item ev_invoke (loop, ev_TYPE *watcher, int revents)
798 1054
799Invoke the C<watcher> with the given C<loop> and C<revents>. Neither 1055Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
800C<loop> nor C<revents> need to be valid as long as the watcher callback 1056C<loop> nor C<revents> need to be valid as long as the watcher callback
801can deal with that fact. 1057can deal with that fact, as both are simply passed through to the
1058callback.
802 1059
803=item int ev_clear_pending (loop, ev_TYPE *watcher) 1060=item int ev_clear_pending (loop, ev_TYPE *watcher)
804 1061
805If the watcher is pending, this function returns clears its pending status 1062If the watcher is pending, this function clears its pending status and
806and returns its C<revents> bitset (as if its callback was invoked). If the 1063returns its C<revents> bitset (as if its callback was invoked). If the
807watcher isn't pending it does nothing and returns C<0>. 1064watcher isn't pending it does nothing and returns C<0>.
808 1065
1066Sometimes it can be useful to "poll" a watcher instead of waiting for its
1067callback to be invoked, which can be accomplished with this function.
1068
809=back 1069=back
810 1070
811 1071
812=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER 1072=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
813 1073
814Each watcher has, by default, a member C<void *data> that you can change 1074Each watcher has, by default, a member C<void *data> that you can change
815and read at any time, libev will completely ignore it. This can be used 1075and read at any time: libev will completely ignore it. This can be used
816to associate arbitrary data with your watcher. If you need more data and 1076to associate arbitrary data with your watcher. If you need more data and
817don't want to allocate memory and store a pointer to it in that data 1077don't want to allocate memory and store a pointer to it in that data
818member, you can also "subclass" the watcher type and provide your own 1078member, you can also "subclass" the watcher type and provide your own
819data: 1079data:
820 1080
821 struct my_io 1081 struct my_io
822 { 1082 {
823 struct ev_io io; 1083 ev_io io;
824 int otherfd; 1084 int otherfd;
825 void *somedata; 1085 void *somedata;
826 struct whatever *mostinteresting; 1086 struct whatever *mostinteresting;
827 } 1087 };
1088
1089 ...
1090 struct my_io w;
1091 ev_io_init (&w.io, my_cb, fd, EV_READ);
828 1092
829And since your callback will be called with a pointer to the watcher, you 1093And since your callback will be called with a pointer to the watcher, you
830can cast it back to your own type: 1094can cast it back to your own type:
831 1095
832 static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents) 1096 static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
833 { 1097 {
834 struct my_io *w = (struct my_io *)w_; 1098 struct my_io *w = (struct my_io *)w_;
835 ... 1099 ...
836 } 1100 }
837 1101
838More interesting and less C-conformant ways of casting your callback type 1102More interesting and less C-conformant ways of casting your callback type
839instead have been omitted. 1103instead have been omitted.
840 1104
841Another common scenario is having some data structure with multiple 1105Another common scenario is to use some data structure with multiple
842watchers: 1106embedded watchers:
843 1107
844 struct my_biggy 1108 struct my_biggy
845 { 1109 {
846 int some_data; 1110 int some_data;
847 ev_timer t1; 1111 ev_timer t1;
848 ev_timer t2; 1112 ev_timer t2;
849 } 1113 }
850 1114
851In this case getting the pointer to C<my_biggy> is a bit more complicated, 1115In this case getting the pointer to C<my_biggy> is a bit more
852you need to use C<offsetof>: 1116complicated: Either you store the address of your C<my_biggy> struct
1117in the C<data> member of the watcher (for woozies), or you need to use
1118some pointer arithmetic using C<offsetof> inside your watchers (for real
1119programmers):
853 1120
854 #include <stddef.h> 1121 #include <stddef.h>
855 1122
856 static void 1123 static void
857 t1_cb (EV_P_ struct ev_timer *w, int revents) 1124 t1_cb (EV_P_ ev_timer *w, int revents)
858 { 1125 {
859 struct my_biggy big = (struct my_biggy * 1126 struct my_biggy big = (struct my_biggy *
860 (((char *)w) - offsetof (struct my_biggy, t1)); 1127 (((char *)w) - offsetof (struct my_biggy, t1));
861 } 1128 }
862 1129
863 static void 1130 static void
864 t2_cb (EV_P_ struct ev_timer *w, int revents) 1131 t2_cb (EV_P_ ev_timer *w, int revents)
865 { 1132 {
866 struct my_biggy big = (struct my_biggy * 1133 struct my_biggy big = (struct my_biggy *
867 (((char *)w) - offsetof (struct my_biggy, t2)); 1134 (((char *)w) - offsetof (struct my_biggy, t2));
868 } 1135 }
869 1136
870 1137
871=head1 WATCHER TYPES 1138=head1 WATCHER TYPES
872 1139
873This section describes each watcher in detail, but will not repeat 1140This section describes each watcher in detail, but will not repeat
897In general you can register as many read and/or write event watchers per 1164In general you can register as many read and/or write event watchers per
898fd as you want (as long as you don't confuse yourself). Setting all file 1165fd as you want (as long as you don't confuse yourself). Setting all file
899descriptors to non-blocking mode is also usually a good idea (but not 1166descriptors to non-blocking mode is also usually a good idea (but not
900required if you know what you are doing). 1167required if you know what you are doing).
901 1168
902You have to be careful with dup'ed file descriptors, though. Some backends 1169If you cannot use non-blocking mode, then force the use of a
903(the linux epoll backend is a notable example) cannot handle dup'ed file 1170known-to-be-good backend (at the time of this writing, this includes only
904descriptors correctly if you register interest in two or more fds pointing 1171C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).
905to the same underlying file/socket/etc. description (that is, they share
906the same underlying "file open").
907
908If you must do this, then force the use of a known-to-be-good backend
909(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
910C<EVBACKEND_POLL>).
911 1172
912Another thing you have to watch out for is that it is quite easy to 1173Another thing you have to watch out for is that it is quite easy to
913receive "spurious" readyness notifications, that is your callback might 1174receive "spurious" readiness notifications, that is your callback might
914be called with C<EV_READ> but a subsequent C<read>(2) will actually block 1175be called with C<EV_READ> but a subsequent C<read>(2) will actually block
915because there is no data. Not only are some backends known to create a 1176because there is no data. Not only are some backends known to create a
916lot of those (for example solaris ports), it is very easy to get into 1177lot of those (for example Solaris ports), it is very easy to get into
917this situation even with a relatively standard program structure. Thus 1178this situation even with a relatively standard program structure. Thus
918it is best to always use non-blocking I/O: An extra C<read>(2) returning 1179it is best to always use non-blocking I/O: An extra C<read>(2) returning
919C<EAGAIN> is far preferable to a program hanging until some data arrives. 1180C<EAGAIN> is far preferable to a program hanging until some data arrives.
920 1181
921If you cannot run the fd in non-blocking mode (for example you should not 1182If you cannot run the fd in non-blocking mode (for example you should
922play around with an Xlib connection), then you have to seperately re-test 1183not play around with an Xlib connection), then you have to separately
923whether a file descriptor is really ready with a known-to-be good interface 1184re-test whether a file descriptor is really ready with a known-to-be good
924such as poll (fortunately in our Xlib example, Xlib already does this on 1185interface such as poll (fortunately in our Xlib example, Xlib already
925its own, so its quite safe to use). 1186does this on its own, so its quite safe to use). Some people additionally
1187use C<SIGALRM> and an interval timer, just to be sure you won't block
1188indefinitely.
1189
1190But really, best use non-blocking mode.
926 1191
927=head3 The special problem of disappearing file descriptors 1192=head3 The special problem of disappearing file descriptors
928 1193
929Some backends (e.g kqueue, epoll) need to be told about closing a file 1194Some backends (e.g. kqueue, epoll) need to be told about closing a file
930descriptor (either by calling C<close> explicitly or by any other means, 1195descriptor (either due to calling C<close> explicitly or any other means,
931such as C<dup>). The reason is that you register interest in some file 1196such as C<dup2>). The reason is that you register interest in some file
932descriptor, but when it goes away, the operating system will silently drop 1197descriptor, but when it goes away, the operating system will silently drop
933this interest. If another file descriptor with the same number then is 1198this interest. If another file descriptor with the same number then is
934registered with libev, there is no efficient way to see that this is, in 1199registered with libev, there is no efficient way to see that this is, in
935fact, a different file descriptor. 1200fact, a different file descriptor.
936 1201
943 1208
944This is how one would do it normally anyway, the important point is that 1209This is how one would do it normally anyway, the important point is that
945the libev application should not optimise around libev but should leave 1210the libev application should not optimise around libev but should leave
946optimisations to libev. 1211optimisations to libev.
947 1212
1213=head3 The special problem of dup'ed file descriptors
1214
1215Some backends (e.g. epoll), cannot register events for file descriptors,
1216but only events for the underlying file descriptions. That means when you
1217have C<dup ()>'ed file descriptors or weirder constellations, and register
1218events for them, only one file descriptor might actually receive events.
1219
1220There is no workaround possible except not registering events
1221for potentially C<dup ()>'ed file descriptors, or to resort to
1222C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1223
1224=head3 The special problem of fork
1225
1226Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1227useless behaviour. Libev fully supports fork, but needs to be told about
1228it in the child.
1229
1230To support fork in your programs, you either have to call
1231C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
1232enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
1233C<EVBACKEND_POLL>.
1234
1235=head3 The special problem of SIGPIPE
1236
1237While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1238when writing to a pipe whose other end has been closed, your program gets
1239sent a SIGPIPE, which, by default, aborts your program. For most programs
1240this is sensible behaviour, for daemons, this is usually undesirable.
1241
1242So when you encounter spurious, unexplained daemon exits, make sure you
1243ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
1244somewhere, as that would have given you a big clue).
1245
948 1246
949=head3 Watcher-Specific Functions 1247=head3 Watcher-Specific Functions
950 1248
951=over 4 1249=over 4
952 1250
953=item ev_io_init (ev_io *, callback, int fd, int events) 1251=item ev_io_init (ev_io *, callback, int fd, int events)
954 1252
955=item ev_io_set (ev_io *, int fd, int events) 1253=item ev_io_set (ev_io *, int fd, int events)
956 1254
957Configures an C<ev_io> watcher. The C<fd> is the file descriptor to 1255Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
958rceeive events for and events is either C<EV_READ>, C<EV_WRITE> or 1256receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
959C<EV_READ | EV_WRITE> to receive the given events. 1257C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
960 1258
961=item int fd [read-only] 1259=item int fd [read-only]
962 1260
963The file descriptor being watched. 1261The file descriptor being watched.
964 1262
965=item int events [read-only] 1263=item int events [read-only]
966 1264
967The events being watched. 1265The events being watched.
968 1266
969=back 1267=back
1268
1269=head3 Examples
970 1270
971Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well 1271Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
972readable, but only once. Since it is likely line-buffered, you could 1272readable, but only once. Since it is likely line-buffered, you could
973attempt to read a whole line in the callback. 1273attempt to read a whole line in the callback.
974 1274
975 static void 1275 static void
976 stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents) 1276 stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
977 { 1277 {
978 ev_io_stop (loop, w); 1278 ev_io_stop (loop, w);
979 .. read from stdin here (or from w->fd) and haqndle any I/O errors 1279 .. read from stdin here (or from w->fd) and handle any I/O errors
980 } 1280 }
981 1281
982 ... 1282 ...
983 struct ev_loop *loop = ev_default_init (0); 1283 struct ev_loop *loop = ev_default_init (0);
984 struct ev_io stdin_readable; 1284 ev_io stdin_readable;
985 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ); 1285 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
986 ev_io_start (loop, &stdin_readable); 1286 ev_io_start (loop, &stdin_readable);
987 ev_loop (loop, 0); 1287 ev_loop (loop, 0);
988 1288
989 1289
990=head2 C<ev_timer> - relative and optionally repeating timeouts 1290=head2 C<ev_timer> - relative and optionally repeating timeouts
991 1291
992Timer watchers are simple relative timers that generate an event after a 1292Timer watchers are simple relative timers that generate an event after a
993given time, and optionally repeating in regular intervals after that. 1293given time, and optionally repeating in regular intervals after that.
994 1294
995The timers are based on real time, that is, if you register an event that 1295The timers are based on real time, that is, if you register an event that
996times out after an hour and you reset your system clock to last years 1296times out after an hour and you reset your system clock to January last
997time, it will still time out after (roughly) and hour. "Roughly" because 1297year, it will still time out after (roughly) one hour. "Roughly" because
998detecting time jumps is hard, and some inaccuracies are unavoidable (the 1298detecting time jumps is hard, and some inaccuracies are unavoidable (the
999monotonic clock option helps a lot here). 1299monotonic clock option helps a lot here).
1300
1301The callback is guaranteed to be invoked only I<after> its timeout has
1302passed, but if multiple timers become ready during the same loop iteration
1303then order of execution is undefined.
1304
1305=head3 Be smart about timeouts
1306
1307Many real-world problems involve some kind of timeout, usually for error
1308recovery. A typical example is an HTTP request - if the other side hangs,
1309you want to raise some error after a while.
1310
1311What follows are some ways to handle this problem, from obvious and
1312inefficient to smart and efficient.
1313
1314In the following, a 60 second activity timeout is assumed - a timeout that
1315gets reset to 60 seconds each time there is activity (e.g. each time some
1316data or other life sign was received).
1317
1318=over 4
1319
1320=item 1. Use a timer and stop, reinitialise and start it on activity.
1321
1322This is the most obvious, but not the most simple way: In the beginning,
1323start the watcher:
1324
1325 ev_timer_init (timer, callback, 60., 0.);
1326 ev_timer_start (loop, timer);
1327
1328Then, each time there is some activity, C<ev_timer_stop> it, initialise it
1329and start it again:
1330
1331 ev_timer_stop (loop, timer);
1332 ev_timer_set (timer, 60., 0.);
1333 ev_timer_start (loop, timer);
1334
1335This is relatively simple to implement, but means that each time there is
1336some activity, libev will first have to remove the timer from its internal
1337data structure and then add it again. Libev tries to be fast, but it's
1338still not a constant-time operation.
1339
1340=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
1341
1342This is the easiest way, and involves using C<ev_timer_again> instead of
1343C<ev_timer_start>.
1344
1345To implement this, configure an C<ev_timer> with a C<repeat> value
1346of C<60> and then call C<ev_timer_again> at start and each time you
1347successfully read or write some data. If you go into an idle state where
1348you do not expect data to travel on the socket, you can C<ev_timer_stop>
1349the timer, and C<ev_timer_again> will automatically restart it if need be.
1350
1351That means you can ignore both the C<ev_timer_start> function and the
1352C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
1353member and C<ev_timer_again>.
1354
1355At start:
1356
1357 ev_timer_init (timer, callback);
1358 timer->repeat = 60.;
1359 ev_timer_again (loop, timer);
1360
1361Each time there is some activity:
1362
1363 ev_timer_again (loop, timer);
1364
1365It is even possible to change the time-out on the fly, regardless of
1366whether the watcher is active or not:
1367
1368 timer->repeat = 30.;
1369 ev_timer_again (loop, timer);
1370
1371This is slightly more efficient then stopping/starting the timer each time
1372you want to modify its timeout value, as libev does not have to completely
1373remove and re-insert the timer from/into its internal data structure.
1374
1375It is, however, even simpler than the "obvious" way to do it.
1376
1377=item 3. Let the timer time out, but then re-arm it as required.
1378
1379This method is more tricky, but usually most efficient: Most timeouts are
1380relatively long compared to the intervals between other activity - in
1381our example, within 60 seconds, there are usually many I/O events with
1382associated activity resets.
1383
1384In this case, it would be more efficient to leave the C<ev_timer> alone,
1385but remember the time of last activity, and check for a real timeout only
1386within the callback:
1387
1388 ev_tstamp last_activity; // time of last activity
1389
1390 static void
1391 callback (EV_P_ ev_timer *w, int revents)
1392 {
1393 ev_tstamp now = ev_now (EV_A);
1394 ev_tstamp timeout = last_activity + 60.;
1395
1396 // if last_activity + 60. is older than now, we did time out
1397 if (timeout < now)
1398 {
1399 // timeout occured, take action
1400 }
1401 else
1402 {
1403 // callback was invoked, but there was some activity, re-arm
1404 // the watcher to fire in last_activity + 60, which is
1405 // guaranteed to be in the future, so "again" is positive:
1406 w->again = timeout - now;
1407 ev_timer_again (EV_A_ w);
1408 }
1409 }
1410
1411To summarise the callback: first calculate the real timeout (defined
1412as "60 seconds after the last activity"), then check if that time has
1413been reached, which means something I<did>, in fact, time out. Otherwise
1414the callback was invoked too early (C<timeout> is in the future), so
1415re-schedule the timer to fire at that future time, to see if maybe we have
1416a timeout then.
1417
1418Note how C<ev_timer_again> is used, taking advantage of the
1419C<ev_timer_again> optimisation when the timer is already running.
1420
1421This scheme causes more callback invocations (about one every 60 seconds
1422minus half the average time between activity), but virtually no calls to
1423libev to change the timeout.
1424
1425To start the timer, simply initialise the watcher and set C<last_activity>
1426to the current time (meaning we just have some activity :), then call the
1427callback, which will "do the right thing" and start the timer:
1428
1429 ev_timer_init (timer, callback);
1430 last_activity = ev_now (loop);
1431 callback (loop, timer, EV_TIMEOUT);
1432
1433And when there is some activity, simply store the current time in
1434C<last_activity>, no libev calls at all:
1435
1436 last_actiivty = ev_now (loop);
1437
1438This technique is slightly more complex, but in most cases where the
1439time-out is unlikely to be triggered, much more efficient.
1440
1441Changing the timeout is trivial as well (if it isn't hard-coded in the
1442callback :) - just change the timeout and invoke the callback, which will
1443fix things for you.
1444
1445=item 4. Wee, just use a double-linked list for your timeouts.
1446
1447If there is not one request, but many thousands (millions...), all
1448employing some kind of timeout with the same timeout value, then one can
1449do even better:
1450
1451When starting the timeout, calculate the timeout value and put the timeout
1452at the I<end> of the list.
1453
1454Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
1455the list is expected to fire (for example, using the technique #3).
1456
1457When there is some activity, remove the timer from the list, recalculate
1458the timeout, append it to the end of the list again, and make sure to
1459update the C<ev_timer> if it was taken from the beginning of the list.
1460
1461This way, one can manage an unlimited number of timeouts in O(1) time for
1462starting, stopping and updating the timers, at the expense of a major
1463complication, and having to use a constant timeout. The constant timeout
1464ensures that the list stays sorted.
1465
1466=back
1467
1468So which method the best?
1469
1470Method #2 is a simple no-brain-required solution that is adequate in most
1471situations. Method #3 requires a bit more thinking, but handles many cases
1472better, and isn't very complicated either. In most case, choosing either
1473one is fine, with #3 being better in typical situations.
1474
1475Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
1476rather complicated, but extremely efficient, something that really pays
1477off after the first million or so of active timers, i.e. it's usually
1478overkill :)
1479
1480=head3 The special problem of time updates
1481
1482Establishing the current time is a costly operation (it usually takes at
1483least two system calls): EV therefore updates its idea of the current
1484time only before and after C<ev_loop> collects new events, which causes a
1485growing difference between C<ev_now ()> and C<ev_time ()> when handling
1486lots of events in one iteration.
1000 1487
1001The relative timeouts are calculated relative to the C<ev_now ()> 1488The relative timeouts are calculated relative to the C<ev_now ()>
1002time. This is usually the right thing as this timestamp refers to the time 1489time. This is usually the right thing as this timestamp refers to the time
1003of the event triggering whatever timeout you are modifying/starting. If 1490of the event triggering whatever timeout you are modifying/starting. If
1004you suspect event processing to be delayed and you I<need> to base the timeout 1491you suspect event processing to be delayed and you I<need> to base the
1005on the current time, use something like this to adjust for this: 1492timeout on the current time, use something like this to adjust for this:
1006 1493
1007 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.); 1494 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1008 1495
1009The callback is guarenteed to be invoked only when its timeout has passed, 1496If the event loop is suspended for a long time, you can also force an
1010but if multiple timers become ready during the same loop iteration then 1497update of the time returned by C<ev_now ()> by calling C<ev_now_update
1011order of execution is undefined. 1498()>.
1012 1499
1013=head3 Watcher-Specific Functions and Data Members 1500=head3 Watcher-Specific Functions and Data Members
1014 1501
1015=over 4 1502=over 4
1016 1503
1017=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat) 1504=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1018 1505
1019=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat) 1506=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1020 1507
1021Configure the timer to trigger after C<after> seconds. If C<repeat> is 1508Configure the timer to trigger after C<after> seconds. If C<repeat>
1022C<0.>, then it will automatically be stopped. If it is positive, then the 1509is C<0.>, then it will automatically be stopped once the timeout is
1023timer will automatically be configured to trigger again C<repeat> seconds 1510reached. If it is positive, then the timer will automatically be
1024later, again, and again, until stopped manually. 1511configured to trigger again C<repeat> seconds later, again, and again,
1512until stopped manually.
1025 1513
1026The timer itself will do a best-effort at avoiding drift, that is, if you 1514The timer itself will do a best-effort at avoiding drift, that is, if
1027configure a timer to trigger every 10 seconds, then it will trigger at 1515you configure a timer to trigger every 10 seconds, then it will normally
1028exactly 10 second intervals. If, however, your program cannot keep up with 1516trigger at exactly 10 second intervals. If, however, your program cannot
1029the timer (because it takes longer than those 10 seconds to do stuff) the 1517keep up with the timer (because it takes longer than those 10 seconds to
1030timer will not fire more than once per event loop iteration. 1518do stuff) the timer will not fire more than once per event loop iteration.
1031 1519
1032=item ev_timer_again (loop) 1520=item ev_timer_again (loop, ev_timer *)
1033 1521
1034This will act as if the timer timed out and restart it again if it is 1522This will act as if the timer timed out and restart it again if it is
1035repeating. The exact semantics are: 1523repeating. The exact semantics are:
1036 1524
1037If the timer is pending, its pending status is cleared. 1525If the timer is pending, its pending status is cleared.
1038 1526
1039If the timer is started but nonrepeating, stop it (as if it timed out). 1527If the timer is started but non-repeating, stop it (as if it timed out).
1040 1528
1041If the timer is repeating, either start it if necessary (with the 1529If the timer is repeating, either start it if necessary (with the
1042C<repeat> value), or reset the running timer to the C<repeat> value. 1530C<repeat> value), or reset the running timer to the C<repeat> value.
1043 1531
1044This sounds a bit complicated, but here is a useful and typical 1532This sounds a bit complicated, see "Be smart about timeouts", above, for a
1045example: Imagine you have a tcp connection and you want a so-called idle 1533usage example.
1046timeout, that is, you want to be called when there have been, say, 60
1047seconds of inactivity on the socket. The easiest way to do this is to
1048configure an C<ev_timer> with a C<repeat> value of C<60> and then call
1049C<ev_timer_again> each time you successfully read or write some data. If
1050you go into an idle state where you do not expect data to travel on the
1051socket, you can C<ev_timer_stop> the timer, and C<ev_timer_again> will
1052automatically restart it if need be.
1053
1054That means you can ignore the C<after> value and C<ev_timer_start>
1055altogether and only ever use the C<repeat> value and C<ev_timer_again>:
1056
1057 ev_timer_init (timer, callback, 0., 5.);
1058 ev_timer_again (loop, timer);
1059 ...
1060 timer->again = 17.;
1061 ev_timer_again (loop, timer);
1062 ...
1063 timer->again = 10.;
1064 ev_timer_again (loop, timer);
1065
1066This is more slightly efficient then stopping/starting the timer each time
1067you want to modify its timeout value.
1068 1534
1069=item ev_tstamp repeat [read-write] 1535=item ev_tstamp repeat [read-write]
1070 1536
1071The current C<repeat> value. Will be used each time the watcher times out 1537The current C<repeat> value. Will be used each time the watcher times out
1072or C<ev_timer_again> is called and determines the next timeout (if any), 1538or C<ev_timer_again> is called, and determines the next timeout (if any),
1073which is also when any modifications are taken into account. 1539which is also when any modifications are taken into account.
1074 1540
1075=back 1541=back
1076 1542
1543=head3 Examples
1544
1077Example: Create a timer that fires after 60 seconds. 1545Example: Create a timer that fires after 60 seconds.
1078 1546
1079 static void 1547 static void
1080 one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents) 1548 one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
1081 { 1549 {
1082 .. one minute over, w is actually stopped right here 1550 .. one minute over, w is actually stopped right here
1083 } 1551 }
1084 1552
1085 struct ev_timer mytimer; 1553 ev_timer mytimer;
1086 ev_timer_init (&mytimer, one_minute_cb, 60., 0.); 1554 ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1087 ev_timer_start (loop, &mytimer); 1555 ev_timer_start (loop, &mytimer);
1088 1556
1089Example: Create a timeout timer that times out after 10 seconds of 1557Example: Create a timeout timer that times out after 10 seconds of
1090inactivity. 1558inactivity.
1091 1559
1092 static void 1560 static void
1093 timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents) 1561 timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
1094 { 1562 {
1095 .. ten seconds without any activity 1563 .. ten seconds without any activity
1096 } 1564 }
1097 1565
1098 struct ev_timer mytimer; 1566 ev_timer mytimer;
1099 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */ 1567 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1100 ev_timer_again (&mytimer); /* start timer */ 1568 ev_timer_again (&mytimer); /* start timer */
1101 ev_loop (loop, 0); 1569 ev_loop (loop, 0);
1102 1570
1103 // and in some piece of code that gets executed on any "activity": 1571 // and in some piece of code that gets executed on any "activity":
1104 // reset the timeout to start ticking again at 10 seconds 1572 // reset the timeout to start ticking again at 10 seconds
1105 ev_timer_again (&mytimer); 1573 ev_timer_again (&mytimer);
1106 1574
1107 1575
1108=head2 C<ev_periodic> - to cron or not to cron? 1576=head2 C<ev_periodic> - to cron or not to cron?
1109 1577
1110Periodic watchers are also timers of a kind, but they are very versatile 1578Periodic watchers are also timers of a kind, but they are very versatile
1111(and unfortunately a bit complex). 1579(and unfortunately a bit complex).
1112 1580
1113Unlike C<ev_timer>'s, they are not based on real time (or relative time) 1581Unlike C<ev_timer>'s, they are not based on real time (or relative time)
1114but on wallclock time (absolute time). You can tell a periodic watcher 1582but on wall clock time (absolute time). You can tell a periodic watcher
1115to trigger "at" some specific point in time. For example, if you tell a 1583to trigger after some specific point in time. For example, if you tell a
1116periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now () 1584periodic watcher to trigger in 10 seconds (by specifying e.g. C<ev_now ()
1117+ 10.>) and then reset your system clock to the last year, then it will 1585+ 10.>, that is, an absolute time not a delay) and then reset your system
1586clock to January of the previous year, then it will take more than year
1118take a year to trigger the event (unlike an C<ev_timer>, which would trigger 1587to trigger the event (unlike an C<ev_timer>, which would still trigger
1119roughly 10 seconds later). 1588roughly 10 seconds later as it uses a relative timeout).
1120 1589
1121They can also be used to implement vastly more complex timers, such as 1590C<ev_periodic>s can also be used to implement vastly more complex timers,
1122triggering an event on each midnight, local time or other, complicated, 1591such as triggering an event on each "midnight, local time", or other
1123rules. 1592complicated rules.
1124 1593
1125As with timers, the callback is guarenteed to be invoked only when the 1594As with timers, the callback is guaranteed to be invoked only when the
1126time (C<at>) has been passed, but if multiple periodic timers become ready 1595time (C<at>) has passed, but if multiple periodic timers become ready
1127during the same loop iteration then order of execution is undefined. 1596during the same loop iteration, then order of execution is undefined.
1128 1597
1129=head3 Watcher-Specific Functions and Data Members 1598=head3 Watcher-Specific Functions and Data Members
1130 1599
1131=over 4 1600=over 4
1132 1601
1133=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb) 1602=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
1134 1603
1135=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb) 1604=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
1136 1605
1137Lots of arguments, lets sort it out... There are basically three modes of 1606Lots of arguments, lets sort it out... There are basically three modes of
1138operation, and we will explain them from simplest to complex: 1607operation, and we will explain them from simplest to most complex:
1139 1608
1140=over 4 1609=over 4
1141 1610
1142=item * absolute timer (at = time, interval = reschedule_cb = 0) 1611=item * absolute timer (at = time, interval = reschedule_cb = 0)
1143 1612
1144In this configuration the watcher triggers an event at the wallclock time 1613In this configuration the watcher triggers an event after the wall clock
1145C<at> and doesn't repeat. It will not adjust when a time jump occurs, 1614time C<at> has passed. It will not repeat and will not adjust when a time
1146that is, if it is to be run at January 1st 2011 then it will run when the 1615jump occurs, that is, if it is to be run at January 1st 2011 then it will
1147system time reaches or surpasses this time. 1616only run when the system clock reaches or surpasses this time.
1148 1617
1149=item * non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0) 1618=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1150 1619
1151In this mode the watcher will always be scheduled to time out at the next 1620In this mode the watcher will always be scheduled to time out at the next
1152C<at + N * interval> time (for some integer N, which can also be negative) 1621C<at + N * interval> time (for some integer N, which can also be negative)
1153and then repeat, regardless of any time jumps. 1622and then repeat, regardless of any time jumps.
1154 1623
1155This can be used to create timers that do not drift with respect to system 1624This can be used to create timers that do not drift with respect to the
1156time: 1625system clock, for example, here is a C<ev_periodic> that triggers each
1626hour, on the hour:
1157 1627
1158 ev_periodic_set (&periodic, 0., 3600., 0); 1628 ev_periodic_set (&periodic, 0., 3600., 0);
1159 1629
1160This doesn't mean there will always be 3600 seconds in between triggers, 1630This doesn't mean there will always be 3600 seconds in between triggers,
1161but only that the the callback will be called when the system time shows a 1631but only that the callback will be called when the system time shows a
1162full hour (UTC), or more correctly, when the system time is evenly divisible 1632full hour (UTC), or more correctly, when the system time is evenly divisible
1163by 3600. 1633by 3600.
1164 1634
1165Another way to think about it (for the mathematically inclined) is that 1635Another way to think about it (for the mathematically inclined) is that
1166C<ev_periodic> will try to run the callback in this mode at the next possible 1636C<ev_periodic> will try to run the callback in this mode at the next possible
1167time where C<time = at (mod interval)>, regardless of any time jumps. 1637time where C<time = at (mod interval)>, regardless of any time jumps.
1168 1638
1169For numerical stability it is preferable that the C<at> value is near 1639For numerical stability it is preferable that the C<at> value is near
1170C<ev_now ()> (the current time), but there is no range requirement for 1640C<ev_now ()> (the current time), but there is no range requirement for
1171this value. 1641this value, and in fact is often specified as zero.
1642
1643Note also that there is an upper limit to how often a timer can fire (CPU
1644speed for example), so if C<interval> is very small then timing stability
1645will of course deteriorate. Libev itself tries to be exact to be about one
1646millisecond (if the OS supports it and the machine is fast enough).
1172 1647
1173=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback) 1648=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1174 1649
1175In this mode the values for C<interval> and C<at> are both being 1650In this mode the values for C<interval> and C<at> are both being
1176ignored. Instead, each time the periodic watcher gets scheduled, the 1651ignored. Instead, each time the periodic watcher gets scheduled, the
1177reschedule callback will be called with the watcher as first, and the 1652reschedule callback will be called with the watcher as first, and the
1178current time as second argument. 1653current time as second argument.
1179 1654
1180NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, 1655NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
1181ever, or make any event loop modifications>. If you need to stop it, 1656ever, or make ANY event loop modifications whatsoever>.
1182return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
1183starting an C<ev_prepare> watcher, which is legal).
1184 1657
1658If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
1659it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
1660only event loop modification you are allowed to do).
1661
1185Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w, 1662The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
1186ev_tstamp now)>, e.g.: 1663*w, ev_tstamp now)>, e.g.:
1187 1664
1665 static ev_tstamp
1188 static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now) 1666 my_rescheduler (ev_periodic *w, ev_tstamp now)
1189 { 1667 {
1190 return now + 60.; 1668 return now + 60.;
1191 } 1669 }
1192 1670
1193It must return the next time to trigger, based on the passed time value 1671It must return the next time to trigger, based on the passed time value
1194(that is, the lowest time value larger than to the second argument). It 1672(that is, the lowest time value larger than to the second argument). It
1195will usually be called just before the callback will be triggered, but 1673will usually be called just before the callback will be triggered, but
1196might be called at other times, too. 1674might be called at other times, too.
1197 1675
1198NOTE: I<< This callback must always return a time that is later than the 1676NOTE: I<< This callback must always return a time that is higher than or
1199passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger. 1677equal to the passed C<now> value >>.
1200 1678
1201This can be used to create very complex timers, such as a timer that 1679This can be used to create very complex timers, such as a timer that
1202triggers on each midnight, local time. To do this, you would calculate the 1680triggers on "next midnight, local time". To do this, you would calculate the
1203next midnight after C<now> and return the timestamp value for this. How 1681next midnight after C<now> and return the timestamp value for this. How
1204you do this is, again, up to you (but it is not trivial, which is the main 1682you do this is, again, up to you (but it is not trivial, which is the main
1205reason I omitted it as an example). 1683reason I omitted it as an example).
1206 1684
1207=back 1685=back
1211Simply stops and restarts the periodic watcher again. This is only useful 1689Simply stops and restarts the periodic watcher again. This is only useful
1212when you changed some parameters or the reschedule callback would return 1690when you changed some parameters or the reschedule callback would return
1213a different time than the last time it was called (e.g. in a crond like 1691a different time than the last time it was called (e.g. in a crond like
1214program when the crontabs have changed). 1692program when the crontabs have changed).
1215 1693
1694=item ev_tstamp ev_periodic_at (ev_periodic *)
1695
1696When active, returns the absolute time that the watcher is supposed to
1697trigger next.
1698
1216=item ev_tstamp offset [read-write] 1699=item ev_tstamp offset [read-write]
1217 1700
1218When repeating, this contains the offset value, otherwise this is the 1701When repeating, this contains the offset value, otherwise this is the
1219absolute point in time (the C<at> value passed to C<ev_periodic_set>). 1702absolute point in time (the C<at> value passed to C<ev_periodic_set>).
1220 1703
1225 1708
1226The current interval value. Can be modified any time, but changes only 1709The current interval value. Can be modified any time, but changes only
1227take effect when the periodic timer fires or C<ev_periodic_again> is being 1710take effect when the periodic timer fires or C<ev_periodic_again> is being
1228called. 1711called.
1229 1712
1230=item ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write] 1713=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
1231 1714
1232The current reschedule callback, or C<0>, if this functionality is 1715The current reschedule callback, or C<0>, if this functionality is
1233switched off. Can be changed any time, but changes only take effect when 1716switched off. Can be changed any time, but changes only take effect when
1234the periodic timer fires or C<ev_periodic_again> is being called. 1717the periodic timer fires or C<ev_periodic_again> is being called.
1235 1718
1236=item ev_tstamp at [read-only]
1237
1238When active, contains the absolute time that the watcher is supposed to
1239trigger next.
1240
1241=back 1719=back
1242 1720
1721=head3 Examples
1722
1243Example: Call a callback every hour, or, more precisely, whenever the 1723Example: Call a callback every hour, or, more precisely, whenever the
1244system clock is divisible by 3600. The callback invocation times have 1724system time is divisible by 3600. The callback invocation times have
1245potentially a lot of jittering, but good long-term stability. 1725potentially a lot of jitter, but good long-term stability.
1246 1726
1247 static void 1727 static void
1248 clock_cb (struct ev_loop *loop, struct ev_io *w, int revents) 1728 clock_cb (struct ev_loop *loop, ev_io *w, int revents)
1249 { 1729 {
1250 ... its now a full hour (UTC, or TAI or whatever your clock follows) 1730 ... its now a full hour (UTC, or TAI or whatever your clock follows)
1251 } 1731 }
1252 1732
1253 struct ev_periodic hourly_tick; 1733 ev_periodic hourly_tick;
1254 ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0); 1734 ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1255 ev_periodic_start (loop, &hourly_tick); 1735 ev_periodic_start (loop, &hourly_tick);
1256 1736
1257Example: The same as above, but use a reschedule callback to do it: 1737Example: The same as above, but use a reschedule callback to do it:
1258 1738
1259 #include <math.h> 1739 #include <math.h>
1260 1740
1261 static ev_tstamp 1741 static ev_tstamp
1262 my_scheduler_cb (struct ev_periodic *w, ev_tstamp now) 1742 my_scheduler_cb (ev_periodic *w, ev_tstamp now)
1263 { 1743 {
1264 return fmod (now, 3600.) + 3600.; 1744 return now + (3600. - fmod (now, 3600.));
1265 } 1745 }
1266 1746
1267 ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb); 1747 ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1268 1748
1269Example: Call a callback every hour, starting now: 1749Example: Call a callback every hour, starting now:
1270 1750
1271 struct ev_periodic hourly_tick; 1751 ev_periodic hourly_tick;
1272 ev_periodic_init (&hourly_tick, clock_cb, 1752 ev_periodic_init (&hourly_tick, clock_cb,
1273 fmod (ev_now (loop), 3600.), 3600., 0); 1753 fmod (ev_now (loop), 3600.), 3600., 0);
1274 ev_periodic_start (loop, &hourly_tick); 1754 ev_periodic_start (loop, &hourly_tick);
1275 1755
1276 1756
1277=head2 C<ev_signal> - signal me when a signal gets signalled! 1757=head2 C<ev_signal> - signal me when a signal gets signalled!
1278 1758
1279Signal watchers will trigger an event when the process receives a specific 1759Signal watchers will trigger an event when the process receives a specific
1280signal one or more times. Even though signals are very asynchronous, libev 1760signal one or more times. Even though signals are very asynchronous, libev
1281will try it's best to deliver signals synchronously, i.e. as part of the 1761will try it's best to deliver signals synchronously, i.e. as part of the
1282normal event processing, like any other event. 1762normal event processing, like any other event.
1283 1763
1764If you want signals asynchronously, just use C<sigaction> as you would
1765do without libev and forget about sharing the signal. You can even use
1766C<ev_async> from a signal handler to synchronously wake up an event loop.
1767
1284You can configure as many watchers as you like per signal. Only when the 1768You can configure as many watchers as you like per signal. Only when the
1285first watcher gets started will libev actually register a signal watcher 1769first watcher gets started will libev actually register a signal handler
1286with the kernel (thus it coexists with your own signal handlers as long 1770with the kernel (thus it coexists with your own signal handlers as long as
1287as you don't register any with libev). Similarly, when the last signal 1771you don't register any with libev for the same signal). Similarly, when
1288watcher for a signal is stopped libev will reset the signal handler to 1772the last signal watcher for a signal is stopped, libev will reset the
1289SIG_DFL (regardless of what it was set to before). 1773signal handler to SIG_DFL (regardless of what it was set to before).
1774
1775If possible and supported, libev will install its handlers with
1776C<SA_RESTART> behaviour enabled, so system calls should not be unduly
1777interrupted. If you have a problem with system calls getting interrupted by
1778signals you can block all signals in an C<ev_check> watcher and unblock
1779them in an C<ev_prepare> watcher.
1290 1780
1291=head3 Watcher-Specific Functions and Data Members 1781=head3 Watcher-Specific Functions and Data Members
1292 1782
1293=over 4 1783=over 4
1294 1784
1303 1793
1304The signal the watcher watches out for. 1794The signal the watcher watches out for.
1305 1795
1306=back 1796=back
1307 1797
1798=head3 Examples
1799
1800Example: Try to exit cleanly on SIGINT.
1801
1802 static void
1803 sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
1804 {
1805 ev_unloop (loop, EVUNLOOP_ALL);
1806 }
1807
1808 ev_signal signal_watcher;
1809 ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1810 ev_signal_start (loop, &signal_watcher);
1811
1308 1812
1309=head2 C<ev_child> - watch out for process status changes 1813=head2 C<ev_child> - watch out for process status changes
1310 1814
1311Child watchers trigger when your process receives a SIGCHLD in response to 1815Child watchers trigger when your process receives a SIGCHLD in response to
1312some child status changes (most typically when a child of yours dies). 1816some child status changes (most typically when a child of yours dies or
1817exits). It is permissible to install a child watcher I<after> the child
1818has been forked (which implies it might have already exited), as long
1819as the event loop isn't entered (or is continued from a watcher), i.e.,
1820forking and then immediately registering a watcher for the child is fine,
1821but forking and registering a watcher a few event loop iterations later is
1822not.
1823
1824Only the default event loop is capable of handling signals, and therefore
1825you can only register child watchers in the default event loop.
1826
1827=head3 Process Interaction
1828
1829Libev grabs C<SIGCHLD> as soon as the default event loop is
1830initialised. This is necessary to guarantee proper behaviour even if
1831the first child watcher is started after the child exits. The occurrence
1832of C<SIGCHLD> is recorded asynchronously, but child reaping is done
1833synchronously as part of the event loop processing. Libev always reaps all
1834children, even ones not watched.
1835
1836=head3 Overriding the Built-In Processing
1837
1838Libev offers no special support for overriding the built-in child
1839processing, but if your application collides with libev's default child
1840handler, you can override it easily by installing your own handler for
1841C<SIGCHLD> after initialising the default loop, and making sure the
1842default loop never gets destroyed. You are encouraged, however, to use an
1843event-based approach to child reaping and thus use libev's support for
1844that, so other libev users can use C<ev_child> watchers freely.
1845
1846=head3 Stopping the Child Watcher
1847
1848Currently, the child watcher never gets stopped, even when the
1849child terminates, so normally one needs to stop the watcher in the
1850callback. Future versions of libev might stop the watcher automatically
1851when a child exit is detected.
1313 1852
1314=head3 Watcher-Specific Functions and Data Members 1853=head3 Watcher-Specific Functions and Data Members
1315 1854
1316=over 4 1855=over 4
1317 1856
1318=item ev_child_init (ev_child *, callback, int pid) 1857=item ev_child_init (ev_child *, callback, int pid, int trace)
1319 1858
1320=item ev_child_set (ev_child *, int pid) 1859=item ev_child_set (ev_child *, int pid, int trace)
1321 1860
1322Configures the watcher to wait for status changes of process C<pid> (or 1861Configures the watcher to wait for status changes of process C<pid> (or
1323I<any> process if C<pid> is specified as C<0>). The callback can look 1862I<any> process if C<pid> is specified as C<0>). The callback can look
1324at the C<rstatus> member of the C<ev_child> watcher structure to see 1863at the C<rstatus> member of the C<ev_child> watcher structure to see
1325the status word (use the macros from C<sys/wait.h> and see your systems 1864the status word (use the macros from C<sys/wait.h> and see your systems
1326C<waitpid> documentation). The C<rpid> member contains the pid of the 1865C<waitpid> documentation). The C<rpid> member contains the pid of the
1327process causing the status change. 1866process causing the status change. C<trace> must be either C<0> (only
1867activate the watcher when the process terminates) or C<1> (additionally
1868activate the watcher when the process is stopped or continued).
1328 1869
1329=item int pid [read-only] 1870=item int pid [read-only]
1330 1871
1331The process id this watcher watches out for, or C<0>, meaning any process id. 1872The process id this watcher watches out for, or C<0>, meaning any process id.
1332 1873
1339The process exit/trace status caused by C<rpid> (see your systems 1880The process exit/trace status caused by C<rpid> (see your systems
1340C<waitpid> and C<sys/wait.h> documentation for details). 1881C<waitpid> and C<sys/wait.h> documentation for details).
1341 1882
1342=back 1883=back
1343 1884
1344Example: Try to exit cleanly on SIGINT and SIGTERM. 1885=head3 Examples
1345 1886
1887Example: C<fork()> a new process and install a child handler to wait for
1888its completion.
1889
1890 ev_child cw;
1891
1346 static void 1892 static void
1347 sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents) 1893 child_cb (EV_P_ ev_child *w, int revents)
1348 { 1894 {
1349 ev_unloop (loop, EVUNLOOP_ALL); 1895 ev_child_stop (EV_A_ w);
1896 printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
1350 } 1897 }
1351 1898
1352 struct ev_signal signal_watcher; 1899 pid_t pid = fork ();
1353 ev_signal_init (&signal_watcher, sigint_cb, SIGINT); 1900
1354 ev_signal_start (loop, &sigint_cb); 1901 if (pid < 0)
1902 // error
1903 else if (pid == 0)
1904 {
1905 // the forked child executes here
1906 exit (1);
1907 }
1908 else
1909 {
1910 ev_child_init (&cw, child_cb, pid, 0);
1911 ev_child_start (EV_DEFAULT_ &cw);
1912 }
1355 1913
1356 1914
1357=head2 C<ev_stat> - did the file attributes just change? 1915=head2 C<ev_stat> - did the file attributes just change?
1358 1916
1359This watches a filesystem path for attribute changes. That is, it calls 1917This watches a file system path for attribute changes. That is, it calls
1360C<stat> regularly (or when the OS says it changed) and sees if it changed 1918C<stat> regularly (or when the OS says it changed) and sees if it changed
1361compared to the last time, invoking the callback if it did. 1919compared to the last time, invoking the callback if it did.
1362 1920
1363The path does not need to exist: changing from "path exists" to "path does 1921The path does not need to exist: changing from "path exists" to "path does
1364not exist" is a status change like any other. The condition "path does 1922not exist" is a status change like any other. The condition "path does
1367the stat buffer having unspecified contents. 1925the stat buffer having unspecified contents.
1368 1926
1369The path I<should> be absolute and I<must not> end in a slash. If it is 1927The path I<should> be absolute and I<must not> end in a slash. If it is
1370relative and your working directory changes, the behaviour is undefined. 1928relative and your working directory changes, the behaviour is undefined.
1371 1929
1372Since there is no standard to do this, the portable implementation simply 1930Since there is no standard kernel interface to do this, the portable
1373calls C<stat (2)> regularly on the path to see if it changed somehow. You 1931implementation simply calls C<stat (2)> regularly on the path to see if
1374can specify a recommended polling interval for this case. If you specify 1932it changed somehow. You can specify a recommended polling interval for
1375a polling interval of C<0> (highly recommended!) then a I<suitable, 1933this case. If you specify a polling interval of C<0> (highly recommended!)
1376unspecified default> value will be used (which you can expect to be around 1934then a I<suitable, unspecified default> value will be used (which
1377five seconds, although this might change dynamically). Libev will also 1935you can expect to be around five seconds, although this might change
1378impose a minimum interval which is currently around C<0.1>, but thats 1936dynamically). Libev will also impose a minimum interval which is currently
1379usually overkill. 1937around C<0.1>, but thats usually overkill.
1380 1938
1381This watcher type is not meant for massive numbers of stat watchers, 1939This watcher type is not meant for massive numbers of stat watchers,
1382as even with OS-supported change notifications, this can be 1940as even with OS-supported change notifications, this can be
1383resource-intensive. 1941resource-intensive.
1384 1942
1385At the time of this writing, only the Linux inotify interface is 1943At the time of this writing, the only OS-specific interface implemented
1386implemented (implementing kqueue support is left as an exercise for the 1944is the Linux inotify interface (implementing kqueue support is left as
1387reader). Inotify will be used to give hints only and should not change the 1945an exercise for the reader. Note, however, that the author sees no way
1388semantics of C<ev_stat> watchers, which means that libev sometimes needs 1946of implementing C<ev_stat> semantics with kqueue).
1389to fall back to regular polling again even with inotify, but changes are 1947
1390usually detected immediately, and if the file exists there will be no 1948=head3 ABI Issues (Largefile Support)
1391polling. 1949
1950Libev by default (unless the user overrides this) uses the default
1951compilation environment, which means that on systems with large file
1952support disabled by default, you get the 32 bit version of the stat
1953structure. When using the library from programs that change the ABI to
1954use 64 bit file offsets the programs will fail. In that case you have to
1955compile libev with the same flags to get binary compatibility. This is
1956obviously the case with any flags that change the ABI, but the problem is
1957most noticeably disabled with ev_stat and large file support.
1958
1959The solution for this is to lobby your distribution maker to make large
1960file interfaces available by default (as e.g. FreeBSD does) and not
1961optional. Libev cannot simply switch on large file support because it has
1962to exchange stat structures with application programs compiled using the
1963default compilation environment.
1964
1965=head3 Inotify and Kqueue
1966
1967When C<inotify (7)> support has been compiled into libev (generally
1968only available with Linux 2.6.25 or above due to bugs in earlier
1969implementations) and present at runtime, it will be used to speed up
1970change detection where possible. The inotify descriptor will be created
1971lazily when the first C<ev_stat> watcher is being started.
1972
1973Inotify presence does not change the semantics of C<ev_stat> watchers
1974except that changes might be detected earlier, and in some cases, to avoid
1975making regular C<stat> calls. Even in the presence of inotify support
1976there are many cases where libev has to resort to regular C<stat> polling,
1977but as long as the path exists, libev usually gets away without polling.
1978
1979There is no support for kqueue, as apparently it cannot be used to
1980implement this functionality, due to the requirement of having a file
1981descriptor open on the object at all times, and detecting renames, unlinks
1982etc. is difficult.
1983
1984=head3 The special problem of stat time resolution
1985
1986The C<stat ()> system call only supports full-second resolution portably, and
1987even on systems where the resolution is higher, most file systems still
1988only support whole seconds.
1989
1990That means that, if the time is the only thing that changes, you can
1991easily miss updates: on the first update, C<ev_stat> detects a change and
1992calls your callback, which does something. When there is another update
1993within the same second, C<ev_stat> will be unable to detect unless the
1994stat data does change in other ways (e.g. file size).
1995
1996The solution to this is to delay acting on a change for slightly more
1997than a second (or till slightly after the next full second boundary), using
1998a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
1999ev_timer_again (loop, w)>).
2000
2001The C<.02> offset is added to work around small timing inconsistencies
2002of some operating systems (where the second counter of the current time
2003might be be delayed. One such system is the Linux kernel, where a call to
2004C<gettimeofday> might return a timestamp with a full second later than
2005a subsequent C<time> call - if the equivalent of C<time ()> is used to
2006update file times then there will be a small window where the kernel uses
2007the previous second to update file times but libev might already execute
2008the timer callback).
1392 2009
1393=head3 Watcher-Specific Functions and Data Members 2010=head3 Watcher-Specific Functions and Data Members
1394 2011
1395=over 4 2012=over 4
1396 2013
1402C<path>. The C<interval> is a hint on how quickly a change is expected to 2019C<path>. The C<interval> is a hint on how quickly a change is expected to
1403be detected and should normally be specified as C<0> to let libev choose 2020be detected and should normally be specified as C<0> to let libev choose
1404a suitable value. The memory pointed to by C<path> must point to the same 2021a suitable value. The memory pointed to by C<path> must point to the same
1405path for as long as the watcher is active. 2022path for as long as the watcher is active.
1406 2023
1407The callback will be receive C<EV_STAT> when a change was detected, 2024The callback will receive an C<EV_STAT> event when a change was detected,
1408relative to the attributes at the time the watcher was started (or the 2025relative to the attributes at the time the watcher was started (or the
1409last change was detected). 2026last change was detected).
1410 2027
1411=item ev_stat_stat (ev_stat *) 2028=item ev_stat_stat (loop, ev_stat *)
1412 2029
1413Updates the stat buffer immediately with new values. If you change the 2030Updates the stat buffer immediately with new values. If you change the
1414watched path in your callback, you could call this fucntion to avoid 2031watched path in your callback, you could call this function to avoid
1415detecting this change (while introducing a race condition). Can also be 2032detecting this change (while introducing a race condition if you are not
1416useful simply to find out the new values. 2033the only one changing the path). Can also be useful simply to find out the
2034new values.
1417 2035
1418=item ev_statdata attr [read-only] 2036=item ev_statdata attr [read-only]
1419 2037
1420The most-recently detected attributes of the file. Although the type is of 2038The most-recently detected attributes of the file. Although the type is
1421C<ev_statdata>, this is usually the (or one of the) C<struct stat> types 2039C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
1422suitable for your system. If the C<st_nlink> member is C<0>, then there 2040suitable for your system, but you can only rely on the POSIX-standardised
2041members to be present. If the C<st_nlink> member is C<0>, then there was
1423was some error while C<stat>ing the file. 2042some error while C<stat>ing the file.
1424 2043
1425=item ev_statdata prev [read-only] 2044=item ev_statdata prev [read-only]
1426 2045
1427The previous attributes of the file. The callback gets invoked whenever 2046The previous attributes of the file. The callback gets invoked whenever
1428C<prev> != C<attr>. 2047C<prev> != C<attr>, or, more precisely, one or more of these members
2048differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
2049C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
1429 2050
1430=item ev_tstamp interval [read-only] 2051=item ev_tstamp interval [read-only]
1431 2052
1432The specified interval. 2053The specified interval.
1433 2054
1434=item const char *path [read-only] 2055=item const char *path [read-only]
1435 2056
1436The filesystem path that is being watched. 2057The file system path that is being watched.
1437 2058
1438=back 2059=back
1439 2060
2061=head3 Examples
2062
1440Example: Watch C</etc/passwd> for attribute changes. 2063Example: Watch C</etc/passwd> for attribute changes.
1441 2064
1442 static void 2065 static void
1443 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents) 2066 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
1444 { 2067 {
1445 /* /etc/passwd changed in some way */ 2068 /* /etc/passwd changed in some way */
1446 if (w->attr.st_nlink) 2069 if (w->attr.st_nlink)
1447 { 2070 {
1448 printf ("passwd current size %ld\n", (long)w->attr.st_size); 2071 printf ("passwd current size %ld\n", (long)w->attr.st_size);
1449 printf ("passwd current atime %ld\n", (long)w->attr.st_mtime); 2072 printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
1450 printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime); 2073 printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
1451 } 2074 }
1452 else 2075 else
1453 /* you shalt not abuse printf for puts */ 2076 /* you shalt not abuse printf for puts */
1454 puts ("wow, /etc/passwd is not there, expect problems. " 2077 puts ("wow, /etc/passwd is not there, expect problems. "
1455 "if this is windows, they already arrived\n"); 2078 "if this is windows, they already arrived\n");
1456 } 2079 }
1457 2080
1458 ... 2081 ...
1459 ev_stat passwd; 2082 ev_stat passwd;
1460 2083
1461 ev_stat_init (&passwd, passwd_cb, "/etc/passwd"); 2084 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
1462 ev_stat_start (loop, &passwd); 2085 ev_stat_start (loop, &passwd);
2086
2087Example: Like above, but additionally use a one-second delay so we do not
2088miss updates (however, frequent updates will delay processing, too, so
2089one might do the work both on C<ev_stat> callback invocation I<and> on
2090C<ev_timer> callback invocation).
2091
2092 static ev_stat passwd;
2093 static ev_timer timer;
2094
2095 static void
2096 timer_cb (EV_P_ ev_timer *w, int revents)
2097 {
2098 ev_timer_stop (EV_A_ w);
2099
2100 /* now it's one second after the most recent passwd change */
2101 }
2102
2103 static void
2104 stat_cb (EV_P_ ev_stat *w, int revents)
2105 {
2106 /* reset the one-second timer */
2107 ev_timer_again (EV_A_ &timer);
2108 }
2109
2110 ...
2111 ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
2112 ev_stat_start (loop, &passwd);
2113 ev_timer_init (&timer, timer_cb, 0., 1.02);
1463 2114
1464 2115
1465=head2 C<ev_idle> - when you've got nothing better to do... 2116=head2 C<ev_idle> - when you've got nothing better to do...
1466 2117
1467Idle watchers trigger events when no other events of the same or higher 2118Idle watchers trigger events when no other events of the same or higher
1468priority are pending (prepare, check and other idle watchers do not 2119priority are pending (prepare, check and other idle watchers do not count
1469count). 2120as receiving "events").
1470 2121
1471That is, as long as your process is busy handling sockets or timeouts 2122That is, as long as your process is busy handling sockets or timeouts
1472(or even signals, imagine) of the same or higher priority it will not be 2123(or even signals, imagine) of the same or higher priority it will not be
1473triggered. But when your process is idle (or only lower-priority watchers 2124triggered. But when your process is idle (or only lower-priority watchers
1474are pending), the idle watchers are being called once per event loop 2125are pending), the idle watchers are being called once per event loop
1493kind. There is a C<ev_idle_set> macro, but using it is utterly pointless, 2144kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1494believe me. 2145believe me.
1495 2146
1496=back 2147=back
1497 2148
2149=head3 Examples
2150
1498Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the 2151Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1499callback, free it. Also, use no error checking, as usual. 2152callback, free it. Also, use no error checking, as usual.
1500 2153
1501 static void 2154 static void
1502 idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents) 2155 idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
1503 { 2156 {
1504 free (w); 2157 free (w);
1505 // now do something you wanted to do when the program has 2158 // now do something you wanted to do when the program has
1506 // no longer asnything immediate to do. 2159 // no longer anything immediate to do.
1507 } 2160 }
1508 2161
1509 struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle)); 2162 ev_idle *idle_watcher = malloc (sizeof (ev_idle));
1510 ev_idle_init (idle_watcher, idle_cb); 2163 ev_idle_init (idle_watcher, idle_cb);
1511 ev_idle_start (loop, idle_cb); 2164 ev_idle_start (loop, idle_cb);
1512 2165
1513 2166
1514=head2 C<ev_prepare> and C<ev_check> - customise your event loop! 2167=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
1515 2168
1516Prepare and check watchers are usually (but not always) used in tandem: 2169Prepare and check watchers are usually (but not always) used in pairs:
1517prepare watchers get invoked before the process blocks and check watchers 2170prepare watchers get invoked before the process blocks and check watchers
1518afterwards. 2171afterwards.
1519 2172
1520You I<must not> call C<ev_loop> or similar functions that enter 2173You I<must not> call C<ev_loop> or similar functions that enter
1521the current event loop from either C<ev_prepare> or C<ev_check> 2174the current event loop from either C<ev_prepare> or C<ev_check>
1524those watchers, i.e. the sequence will always be C<ev_prepare>, blocking, 2177those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
1525C<ev_check> so if you have one watcher of each kind they will always be 2178C<ev_check> so if you have one watcher of each kind they will always be
1526called in pairs bracketing the blocking call. 2179called in pairs bracketing the blocking call.
1527 2180
1528Their main purpose is to integrate other event mechanisms into libev and 2181Their main purpose is to integrate other event mechanisms into libev and
1529their use is somewhat advanced. This could be used, for example, to track 2182their use is somewhat advanced. They could be used, for example, to track
1530variable changes, implement your own watchers, integrate net-snmp or a 2183variable changes, implement your own watchers, integrate net-snmp or a
1531coroutine library and lots more. They are also occasionally useful if 2184coroutine library and lots more. They are also occasionally useful if
1532you cache some data and want to flush it before blocking (for example, 2185you cache some data and want to flush it before blocking (for example,
1533in X programs you might want to do an C<XFlush ()> in an C<ev_prepare> 2186in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
1534watcher). 2187watcher).
1535 2188
1536This is done by examining in each prepare call which file descriptors need 2189This is done by examining in each prepare call which file descriptors
1537to be watched by the other library, registering C<ev_io> watchers for 2190need to be watched by the other library, registering C<ev_io> watchers
1538them and starting an C<ev_timer> watcher for any timeouts (many libraries 2191for them and starting an C<ev_timer> watcher for any timeouts (many
1539provide just this functionality). Then, in the check watcher you check for 2192libraries provide exactly this functionality). Then, in the check watcher,
1540any events that occured (by checking the pending status of all watchers 2193you check for any events that occurred (by checking the pending status
1541and stopping them) and call back into the library. The I/O and timer 2194of all watchers and stopping them) and call back into the library. The
1542callbacks will never actually be called (but must be valid nevertheless, 2195I/O and timer callbacks will never actually be called (but must be valid
1543because you never know, you know?). 2196nevertheless, because you never know, you know?).
1544 2197
1545As another example, the Perl Coro module uses these hooks to integrate 2198As another example, the Perl Coro module uses these hooks to integrate
1546coroutines into libev programs, by yielding to other active coroutines 2199coroutines into libev programs, by yielding to other active coroutines
1547during each prepare and only letting the process block if no coroutines 2200during each prepare and only letting the process block if no coroutines
1548are ready to run (it's actually more complicated: it only runs coroutines 2201are ready to run (it's actually more complicated: it only runs coroutines
1551loop from blocking if lower-priority coroutines are active, thus mapping 2204loop from blocking if lower-priority coroutines are active, thus mapping
1552low-priority coroutines to idle/background tasks). 2205low-priority coroutines to idle/background tasks).
1553 2206
1554It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>) 2207It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1555priority, to ensure that they are being run before any other watchers 2208priority, to ensure that they are being run before any other watchers
2209after the poll (this doesn't matter for C<ev_prepare> watchers).
2210
1556after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers, 2211Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
1557too) should not activate ("feed") events into libev. While libev fully 2212activate ("feed") events into libev. While libev fully supports this, they
1558supports this, they will be called before other C<ev_check> watchers did 2213might get executed before other C<ev_check> watchers did their job. As
1559their job. As C<ev_check> watchers are often used to embed other event 2214C<ev_check> watchers are often used to embed other (non-libev) event
1560loops those other event loops might be in an unusable state until their 2215loops those other event loops might be in an unusable state until their
1561C<ev_check> watcher ran (always remind yourself to coexist peacefully with 2216C<ev_check> watcher ran (always remind yourself to coexist peacefully with
1562others). 2217others).
1563 2218
1564=head3 Watcher-Specific Functions and Data Members 2219=head3 Watcher-Specific Functions and Data Members
1569 2224
1570=item ev_check_init (ev_check *, callback) 2225=item ev_check_init (ev_check *, callback)
1571 2226
1572Initialises and configures the prepare or check watcher - they have no 2227Initialises and configures the prepare or check watcher - they have no
1573parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set> 2228parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1574macros, but using them is utterly, utterly and completely pointless. 2229macros, but using them is utterly, utterly, utterly and completely
2230pointless.
1575 2231
1576=back 2232=back
2233
2234=head3 Examples
1577 2235
1578There are a number of principal ways to embed other event loops or modules 2236There are a number of principal ways to embed other event loops or modules
1579into libev. Here are some ideas on how to include libadns into libev 2237into libev. Here are some ideas on how to include libadns into libev
1580(there is a Perl module named C<EV::ADNS> that does this, which you could 2238(there is a Perl module named C<EV::ADNS> that does this, which you could
1581use for an actually working example. Another Perl module named C<EV::Glib> 2239use as a working example. Another Perl module named C<EV::Glib> embeds a
1582embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV 2240Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
1583into the Glib event loop). 2241Glib event loop).
1584 2242
1585Method 1: Add IO watchers and a timeout watcher in a prepare handler, 2243Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1586and in a check watcher, destroy them and call into libadns. What follows 2244and in a check watcher, destroy them and call into libadns. What follows
1587is pseudo-code only of course. This requires you to either use a low 2245is pseudo-code only of course. This requires you to either use a low
1588priority for the check watcher or use C<ev_clear_pending> explicitly, as 2246priority for the check watcher or use C<ev_clear_pending> explicitly, as
1589the callbacks for the IO/timeout watchers might not have been called yet. 2247the callbacks for the IO/timeout watchers might not have been called yet.
1590 2248
1591 static ev_io iow [nfd]; 2249 static ev_io iow [nfd];
1592 static ev_timer tw; 2250 static ev_timer tw;
1593 2251
1594 static void 2252 static void
1595 io_cb (ev_loop *loop, ev_io *w, int revents) 2253 io_cb (struct ev_loop *loop, ev_io *w, int revents)
1596 { 2254 {
1597 } 2255 }
1598 2256
1599 // create io watchers for each fd and a timer before blocking 2257 // create io watchers for each fd and a timer before blocking
1600 static void 2258 static void
1601 adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents) 2259 adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
1602 { 2260 {
1603 int timeout = 3600000; 2261 int timeout = 3600000;
1604 struct pollfd fds [nfd]; 2262 struct pollfd fds [nfd];
1605 // actual code will need to loop here and realloc etc. 2263 // actual code will need to loop here and realloc etc.
1606 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ())); 2264 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1607 2265
1608 /* the callback is illegal, but won't be called as we stop during check */ 2266 /* the callback is illegal, but won't be called as we stop during check */
1609 ev_timer_init (&tw, 0, timeout * 1e-3); 2267 ev_timer_init (&tw, 0, timeout * 1e-3);
1610 ev_timer_start (loop, &tw); 2268 ev_timer_start (loop, &tw);
1611 2269
1612 // create one ev_io per pollfd 2270 // create one ev_io per pollfd
1613 for (int i = 0; i < nfd; ++i) 2271 for (int i = 0; i < nfd; ++i)
1614 { 2272 {
1615 ev_io_init (iow + i, io_cb, fds [i].fd, 2273 ev_io_init (iow + i, io_cb, fds [i].fd,
1616 ((fds [i].events & POLLIN ? EV_READ : 0) 2274 ((fds [i].events & POLLIN ? EV_READ : 0)
1617 | (fds [i].events & POLLOUT ? EV_WRITE : 0))); 2275 | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
1618 2276
1619 fds [i].revents = 0; 2277 fds [i].revents = 0;
1620 ev_io_start (loop, iow + i); 2278 ev_io_start (loop, iow + i);
1621 } 2279 }
1622 } 2280 }
1623 2281
1624 // stop all watchers after blocking 2282 // stop all watchers after blocking
1625 static void 2283 static void
1626 adns_check_cb (ev_loop *loop, ev_check *w, int revents) 2284 adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
1627 { 2285 {
1628 ev_timer_stop (loop, &tw); 2286 ev_timer_stop (loop, &tw);
1629 2287
1630 for (int i = 0; i < nfd; ++i) 2288 for (int i = 0; i < nfd; ++i)
1631 { 2289 {
1632 // set the relevant poll flags 2290 // set the relevant poll flags
1633 // could also call adns_processreadable etc. here 2291 // could also call adns_processreadable etc. here
1634 struct pollfd *fd = fds + i; 2292 struct pollfd *fd = fds + i;
1635 int revents = ev_clear_pending (iow + i); 2293 int revents = ev_clear_pending (iow + i);
1636 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN; 2294 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1637 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT; 2295 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1638 2296
1639 // now stop the watcher 2297 // now stop the watcher
1640 ev_io_stop (loop, iow + i); 2298 ev_io_stop (loop, iow + i);
1641 } 2299 }
1642 2300
1643 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop)); 2301 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
1644 } 2302 }
1645 2303
1646Method 2: This would be just like method 1, but you run C<adns_afterpoll> 2304Method 2: This would be just like method 1, but you run C<adns_afterpoll>
1647in the prepare watcher and would dispose of the check watcher. 2305in the prepare watcher and would dispose of the check watcher.
1648 2306
1649Method 3: If the module to be embedded supports explicit event 2307Method 3: If the module to be embedded supports explicit event
1650notification (adns does), you can also make use of the actual watcher 2308notification (libadns does), you can also make use of the actual watcher
1651callbacks, and only destroy/create the watchers in the prepare watcher. 2309callbacks, and only destroy/create the watchers in the prepare watcher.
1652 2310
1653 static void 2311 static void
1654 timer_cb (EV_P_ ev_timer *w, int revents) 2312 timer_cb (EV_P_ ev_timer *w, int revents)
1655 { 2313 {
1656 adns_state ads = (adns_state)w->data; 2314 adns_state ads = (adns_state)w->data;
1657 update_now (EV_A); 2315 update_now (EV_A);
1658 2316
1659 adns_processtimeouts (ads, &tv_now); 2317 adns_processtimeouts (ads, &tv_now);
1660 } 2318 }
1661 2319
1662 static void 2320 static void
1663 io_cb (EV_P_ ev_io *w, int revents) 2321 io_cb (EV_P_ ev_io *w, int revents)
1664 { 2322 {
1665 adns_state ads = (adns_state)w->data; 2323 adns_state ads = (adns_state)w->data;
1666 update_now (EV_A); 2324 update_now (EV_A);
1667 2325
1668 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now); 2326 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
1669 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now); 2327 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
1670 } 2328 }
1671 2329
1672 // do not ever call adns_afterpoll 2330 // do not ever call adns_afterpoll
1673 2331
1674Method 4: Do not use a prepare or check watcher because the module you 2332Method 4: Do not use a prepare or check watcher because the module you
1675want to embed is too inflexible to support it. Instead, youc na override 2333want to embed is not flexible enough to support it. Instead, you can
1676their poll function. The drawback with this solution is that the main 2334override their poll function. The drawback with this solution is that the
1677loop is now no longer controllable by EV. The C<Glib::EV> module does 2335main loop is now no longer controllable by EV. The C<Glib::EV> module uses
1678this. 2336this approach, effectively embedding EV as a client into the horrible
2337libglib event loop.
1679 2338
1680 static gint 2339 static gint
1681 event_poll_func (GPollFD *fds, guint nfds, gint timeout) 2340 event_poll_func (GPollFD *fds, guint nfds, gint timeout)
1682 { 2341 {
1683 int got_events = 0; 2342 int got_events = 0;
1684 2343
1685 for (n = 0; n < nfds; ++n) 2344 for (n = 0; n < nfds; ++n)
1686 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events 2345 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
1687 2346
1688 if (timeout >= 0) 2347 if (timeout >= 0)
1689 // create/start timer 2348 // create/start timer
1690 2349
1691 // poll 2350 // poll
1692 ev_loop (EV_A_ 0); 2351 ev_loop (EV_A_ 0);
1693 2352
1694 // stop timer again 2353 // stop timer again
1695 if (timeout >= 0) 2354 if (timeout >= 0)
1696 ev_timer_stop (EV_A_ &to); 2355 ev_timer_stop (EV_A_ &to);
1697 2356
1698 // stop io watchers again - their callbacks should have set 2357 // stop io watchers again - their callbacks should have set
1699 for (n = 0; n < nfds; ++n) 2358 for (n = 0; n < nfds; ++n)
1700 ev_io_stop (EV_A_ iow [n]); 2359 ev_io_stop (EV_A_ iow [n]);
1701 2360
1702 return got_events; 2361 return got_events;
1703 } 2362 }
1704 2363
1705 2364
1706=head2 C<ev_embed> - when one backend isn't enough... 2365=head2 C<ev_embed> - when one backend isn't enough...
1707 2366
1708This is a rather advanced watcher type that lets you embed one event loop 2367This is a rather advanced watcher type that lets you embed one event loop
1714prioritise I/O. 2373prioritise I/O.
1715 2374
1716As an example for a bug workaround, the kqueue backend might only support 2375As an example for a bug workaround, the kqueue backend might only support
1717sockets on some platform, so it is unusable as generic backend, but you 2376sockets on some platform, so it is unusable as generic backend, but you
1718still want to make use of it because you have many sockets and it scales 2377still want to make use of it because you have many sockets and it scales
1719so nicely. In this case, you would create a kqueue-based loop and embed it 2378so nicely. In this case, you would create a kqueue-based loop and embed
1720into your default loop (which might use e.g. poll). Overall operation will 2379it into your default loop (which might use e.g. poll). Overall operation
1721be a bit slower because first libev has to poll and then call kevent, but 2380will be a bit slower because first libev has to call C<poll> and then
1722at least you can use both at what they are best. 2381C<kevent>, but at least you can use both mechanisms for what they are
2382best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
1723 2383
1724As for prioritising I/O: rarely you have the case where some fds have 2384As for prioritising I/O: under rare circumstances you have the case where
1725to be watched and handled very quickly (with low latency), and even 2385some fds have to be watched and handled very quickly (with low latency),
1726priorities and idle watchers might have too much overhead. In this case 2386and even priorities and idle watchers might have too much overhead. In
1727you would put all the high priority stuff in one loop and all the rest in 2387this case you would put all the high priority stuff in one loop and all
1728a second one, and embed the second one in the first. 2388the rest in a second one, and embed the second one in the first.
1729 2389
1730As long as the watcher is active, the callback will be invoked every time 2390As long as the watcher is active, the callback will be invoked every time
1731there might be events pending in the embedded loop. The callback must then 2391there might be events pending in the embedded loop. The callback must then
1732call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke 2392call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
1733their callbacks (you could also start an idle watcher to give the embedded 2393their callbacks (you could also start an idle watcher to give the embedded
1741interested in that. 2401interested in that.
1742 2402
1743Also, there have not currently been made special provisions for forking: 2403Also, there have not currently been made special provisions for forking:
1744when you fork, you not only have to call C<ev_loop_fork> on both loops, 2404when you fork, you not only have to call C<ev_loop_fork> on both loops,
1745but you will also have to stop and restart any C<ev_embed> watchers 2405but you will also have to stop and restart any C<ev_embed> watchers
1746yourself. 2406yourself - but you can use a fork watcher to handle this automatically,
2407and future versions of libev might do just that.
1747 2408
1748Unfortunately, not all backends are embeddable, only the ones returned by 2409Unfortunately, not all backends are embeddable: only the ones returned by
1749C<ev_embeddable_backends> are, which, unfortunately, does not include any 2410C<ev_embeddable_backends> are, which, unfortunately, does not include any
1750portable one. 2411portable one.
1751 2412
1752So when you want to use this feature you will always have to be prepared 2413So when you want to use this feature you will always have to be prepared
1753that you cannot get an embeddable loop. The recommended way to get around 2414that you cannot get an embeddable loop. The recommended way to get around
1754this is to have a separate variables for your embeddable loop, try to 2415this is to have a separate variables for your embeddable loop, try to
1755create it, and if that fails, use the normal loop for everything: 2416create it, and if that fails, use the normal loop for everything.
1756 2417
1757 struct ev_loop *loop_hi = ev_default_init (0); 2418=head3 C<ev_embed> and fork
1758 struct ev_loop *loop_lo = 0;
1759 struct ev_embed embed;
1760
1761 // see if there is a chance of getting one that works
1762 // (remember that a flags value of 0 means autodetection)
1763 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
1764 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
1765 : 0;
1766 2419
1767 // if we got one, then embed it, otherwise default to loop_hi 2420While the C<ev_embed> watcher is running, forks in the embedding loop will
1768 if (loop_lo) 2421automatically be applied to the embedded loop as well, so no special
1769 { 2422fork handling is required in that case. When the watcher is not running,
1770 ev_embed_init (&embed, 0, loop_lo); 2423however, it is still the task of the libev user to call C<ev_loop_fork ()>
1771 ev_embed_start (loop_hi, &embed); 2424as applicable.
1772 }
1773 else
1774 loop_lo = loop_hi;
1775 2425
1776=head3 Watcher-Specific Functions and Data Members 2426=head3 Watcher-Specific Functions and Data Members
1777 2427
1778=over 4 2428=over 4
1779 2429
1783 2433
1784Configures the watcher to embed the given loop, which must be 2434Configures the watcher to embed the given loop, which must be
1785embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be 2435embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1786invoked automatically, otherwise it is the responsibility of the callback 2436invoked automatically, otherwise it is the responsibility of the callback
1787to invoke it (it will continue to be called until the sweep has been done, 2437to invoke it (it will continue to be called until the sweep has been done,
1788if you do not want thta, you need to temporarily stop the embed watcher). 2438if you do not want that, you need to temporarily stop the embed watcher).
1789 2439
1790=item ev_embed_sweep (loop, ev_embed *) 2440=item ev_embed_sweep (loop, ev_embed *)
1791 2441
1792Make a single, non-blocking sweep over the embedded loop. This works 2442Make a single, non-blocking sweep over the embedded loop. This works
1793similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most 2443similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
1794apropriate way for embedded loops. 2444appropriate way for embedded loops.
1795 2445
1796=item struct ev_loop *other [read-only] 2446=item struct ev_loop *other [read-only]
1797 2447
1798The embedded event loop. 2448The embedded event loop.
1799 2449
1800=back 2450=back
2451
2452=head3 Examples
2453
2454Example: Try to get an embeddable event loop and embed it into the default
2455event loop. If that is not possible, use the default loop. The default
2456loop is stored in C<loop_hi>, while the embeddable loop is stored in
2457C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
2458used).
2459
2460 struct ev_loop *loop_hi = ev_default_init (0);
2461 struct ev_loop *loop_lo = 0;
2462 ev_embed embed;
2463
2464 // see if there is a chance of getting one that works
2465 // (remember that a flags value of 0 means autodetection)
2466 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
2467 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
2468 : 0;
2469
2470 // if we got one, then embed it, otherwise default to loop_hi
2471 if (loop_lo)
2472 {
2473 ev_embed_init (&embed, 0, loop_lo);
2474 ev_embed_start (loop_hi, &embed);
2475 }
2476 else
2477 loop_lo = loop_hi;
2478
2479Example: Check if kqueue is available but not recommended and create
2480a kqueue backend for use with sockets (which usually work with any
2481kqueue implementation). Store the kqueue/socket-only event loop in
2482C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
2483
2484 struct ev_loop *loop = ev_default_init (0);
2485 struct ev_loop *loop_socket = 0;
2486 ev_embed embed;
2487
2488 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2489 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2490 {
2491 ev_embed_init (&embed, 0, loop_socket);
2492 ev_embed_start (loop, &embed);
2493 }
2494
2495 if (!loop_socket)
2496 loop_socket = loop;
2497
2498 // now use loop_socket for all sockets, and loop for everything else
1801 2499
1802 2500
1803=head2 C<ev_fork> - the audacity to resume the event loop after a fork 2501=head2 C<ev_fork> - the audacity to resume the event loop after a fork
1804 2502
1805Fork watchers are called when a C<fork ()> was detected (usually because 2503Fork watchers are called when a C<fork ()> was detected (usually because
1821believe me. 2519believe me.
1822 2520
1823=back 2521=back
1824 2522
1825 2523
2524=head2 C<ev_async> - how to wake up another event loop
2525
2526In general, you cannot use an C<ev_loop> from multiple threads or other
2527asynchronous sources such as signal handlers (as opposed to multiple event
2528loops - those are of course safe to use in different threads).
2529
2530Sometimes, however, you need to wake up another event loop you do not
2531control, for example because it belongs to another thread. This is what
2532C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
2533can signal it by calling C<ev_async_send>, which is thread- and signal
2534safe.
2535
2536This functionality is very similar to C<ev_signal> watchers, as signals,
2537too, are asynchronous in nature, and signals, too, will be compressed
2538(i.e. the number of callback invocations may be less than the number of
2539C<ev_async_sent> calls).
2540
2541Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
2542just the default loop.
2543
2544=head3 Queueing
2545
2546C<ev_async> does not support queueing of data in any way. The reason
2547is that the author does not know of a simple (or any) algorithm for a
2548multiple-writer-single-reader queue that works in all cases and doesn't
2549need elaborate support such as pthreads.
2550
2551That means that if you want to queue data, you have to provide your own
2552queue. But at least I can tell you how to implement locking around your
2553queue:
2554
2555=over 4
2556
2557=item queueing from a signal handler context
2558
2559To implement race-free queueing, you simply add to the queue in the signal
2560handler but you block the signal handler in the watcher callback. Here is
2561an example that does that for some fictitious SIGUSR1 handler:
2562
2563 static ev_async mysig;
2564
2565 static void
2566 sigusr1_handler (void)
2567 {
2568 sometype data;
2569
2570 // no locking etc.
2571 queue_put (data);
2572 ev_async_send (EV_DEFAULT_ &mysig);
2573 }
2574
2575 static void
2576 mysig_cb (EV_P_ ev_async *w, int revents)
2577 {
2578 sometype data;
2579 sigset_t block, prev;
2580
2581 sigemptyset (&block);
2582 sigaddset (&block, SIGUSR1);
2583 sigprocmask (SIG_BLOCK, &block, &prev);
2584
2585 while (queue_get (&data))
2586 process (data);
2587
2588 if (sigismember (&prev, SIGUSR1)
2589 sigprocmask (SIG_UNBLOCK, &block, 0);
2590 }
2591
2592(Note: pthreads in theory requires you to use C<pthread_setmask>
2593instead of C<sigprocmask> when you use threads, but libev doesn't do it
2594either...).
2595
2596=item queueing from a thread context
2597
2598The strategy for threads is different, as you cannot (easily) block
2599threads but you can easily preempt them, so to queue safely you need to
2600employ a traditional mutex lock, such as in this pthread example:
2601
2602 static ev_async mysig;
2603 static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
2604
2605 static void
2606 otherthread (void)
2607 {
2608 // only need to lock the actual queueing operation
2609 pthread_mutex_lock (&mymutex);
2610 queue_put (data);
2611 pthread_mutex_unlock (&mymutex);
2612
2613 ev_async_send (EV_DEFAULT_ &mysig);
2614 }
2615
2616 static void
2617 mysig_cb (EV_P_ ev_async *w, int revents)
2618 {
2619 pthread_mutex_lock (&mymutex);
2620
2621 while (queue_get (&data))
2622 process (data);
2623
2624 pthread_mutex_unlock (&mymutex);
2625 }
2626
2627=back
2628
2629
2630=head3 Watcher-Specific Functions and Data Members
2631
2632=over 4
2633
2634=item ev_async_init (ev_async *, callback)
2635
2636Initialises and configures the async watcher - it has no parameters of any
2637kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
2638trust me.
2639
2640=item ev_async_send (loop, ev_async *)
2641
2642Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2643an C<EV_ASYNC> event on the watcher into the event loop. Unlike
2644C<ev_feed_event>, this call is safe to do from other threads, signal or
2645similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
2646section below on what exactly this means).
2647
2648This call incurs the overhead of a system call only once per loop iteration,
2649so while the overhead might be noticeable, it doesn't apply to repeated
2650calls to C<ev_async_send>.
2651
2652=item bool = ev_async_pending (ev_async *)
2653
2654Returns a non-zero value when C<ev_async_send> has been called on the
2655watcher but the event has not yet been processed (or even noted) by the
2656event loop.
2657
2658C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
2659the loop iterates next and checks for the watcher to have become active,
2660it will reset the flag again. C<ev_async_pending> can be used to very
2661quickly check whether invoking the loop might be a good idea.
2662
2663Not that this does I<not> check whether the watcher itself is pending, only
2664whether it has been requested to make this watcher pending.
2665
2666=back
2667
2668
1826=head1 OTHER FUNCTIONS 2669=head1 OTHER FUNCTIONS
1827 2670
1828There are some other functions of possible interest. Described. Here. Now. 2671There are some other functions of possible interest. Described. Here. Now.
1829 2672
1830=over 4 2673=over 4
1831 2674
1832=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback) 2675=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
1833 2676
1834This function combines a simple timer and an I/O watcher, calls your 2677This function combines a simple timer and an I/O watcher, calls your
1835callback on whichever event happens first and automatically stop both 2678callback on whichever event happens first and automatically stops both
1836watchers. This is useful if you want to wait for a single event on an fd 2679watchers. This is useful if you want to wait for a single event on an fd
1837or timeout without having to allocate/configure/start/stop/free one or 2680or timeout without having to allocate/configure/start/stop/free one or
1838more watchers yourself. 2681more watchers yourself.
1839 2682
1840If C<fd> is less than 0, then no I/O watcher will be started and events 2683If C<fd> is less than 0, then no I/O watcher will be started and the
1841is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and 2684C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
1842C<events> set will be craeted and started. 2685the given C<fd> and C<events> set will be created and started.
1843 2686
1844If C<timeout> is less than 0, then no timeout watcher will be 2687If C<timeout> is less than 0, then no timeout watcher will be
1845started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and 2688started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
1846repeat = 0) will be started. While C<0> is a valid timeout, it is of 2689repeat = 0) will be started. C<0> is a valid timeout.
1847dubious value.
1848 2690
1849The callback has the type C<void (*cb)(int revents, void *arg)> and gets 2691The callback has the type C<void (*cb)(int revents, void *arg)> and gets
1850passed an C<revents> set like normal event callbacks (a combination of 2692passed an C<revents> set like normal event callbacks (a combination of
1851C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg> 2693C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
1852value passed to C<ev_once>: 2694value passed to C<ev_once>. Note that it is possible to receive I<both>
2695a timeout and an io event at the same time - you probably should give io
2696events precedence.
1853 2697
2698Example: wait up to ten seconds for data to appear on STDIN_FILENO.
2699
1854 static void stdin_ready (int revents, void *arg) 2700 static void stdin_ready (int revents, void *arg)
1855 { 2701 {
1856 if (revents & EV_TIMEOUT)
1857 /* doh, nothing entered */;
1858 else if (revents & EV_READ) 2702 if (revents & EV_READ)
1859 /* stdin might have data for us, joy! */; 2703 /* stdin might have data for us, joy! */;
2704 else if (revents & EV_TIMEOUT)
2705 /* doh, nothing entered */;
1860 } 2706 }
1861 2707
1862 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0); 2708 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
1863 2709
1864=item ev_feed_event (ev_loop *, watcher *, int revents) 2710=item ev_feed_event (struct ev_loop *, watcher *, int revents)
1865 2711
1866Feeds the given event set into the event loop, as if the specified event 2712Feeds the given event set into the event loop, as if the specified event
1867had happened for the specified watcher (which must be a pointer to an 2713had happened for the specified watcher (which must be a pointer to an
1868initialised but not necessarily started event watcher). 2714initialised but not necessarily started event watcher).
1869 2715
1870=item ev_feed_fd_event (ev_loop *, int fd, int revents) 2716=item ev_feed_fd_event (struct ev_loop *, int fd, int revents)
1871 2717
1872Feed an event on the given fd, as if a file descriptor backend detected 2718Feed an event on the given fd, as if a file descriptor backend detected
1873the given events it. 2719the given events it.
1874 2720
1875=item ev_feed_signal_event (ev_loop *loop, int signum) 2721=item ev_feed_signal_event (struct ev_loop *loop, int signum)
1876 2722
1877Feed an event as if the given signal occured (C<loop> must be the default 2723Feed an event as if the given signal occurred (C<loop> must be the default
1878loop!). 2724loop!).
1879 2725
1880=back 2726=back
1881 2727
1882 2728
1898 2744
1899=item * Priorities are not currently supported. Initialising priorities 2745=item * Priorities are not currently supported. Initialising priorities
1900will fail and all watchers will have the same priority, even though there 2746will fail and all watchers will have the same priority, even though there
1901is an ev_pri field. 2747is an ev_pri field.
1902 2748
2749=item * In libevent, the last base created gets the signals, in libev, the
2750first base created (== the default loop) gets the signals.
2751
1903=item * Other members are not supported. 2752=item * Other members are not supported.
1904 2753
1905=item * The libev emulation is I<not> ABI compatible to libevent, you need 2754=item * The libev emulation is I<not> ABI compatible to libevent, you need
1906to use the libev header file and library. 2755to use the libev header file and library.
1907 2756
1908=back 2757=back
1909 2758
1910=head1 C++ SUPPORT 2759=head1 C++ SUPPORT
1911 2760
1912Libev comes with some simplistic wrapper classes for C++ that mainly allow 2761Libev comes with some simplistic wrapper classes for C++ that mainly allow
1913you to use some convinience methods to start/stop watchers and also change 2762you to use some convenience methods to start/stop watchers and also change
1914the callback model to a model using method callbacks on objects. 2763the callback model to a model using method callbacks on objects.
1915 2764
1916To use it, 2765To use it,
1917 2766
1918 #include <ev++.h> 2767 #include <ev++.h>
1919 2768
1920This automatically includes F<ev.h> and puts all of its definitions (many 2769This automatically includes F<ev.h> and puts all of its definitions (many
1921of them macros) into the global namespace. All C++ specific things are 2770of them macros) into the global namespace. All C++ specific things are
1922put into the C<ev> namespace. It should support all the same embedding 2771put into the C<ev> namespace. It should support all the same embedding
1923options as F<ev.h>, most notably C<EV_MULTIPLICITY>. 2772options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
1990your compiler is good :), then the method will be fully inlined into the 2839your compiler is good :), then the method will be fully inlined into the
1991thunking function, making it as fast as a direct C callback. 2840thunking function, making it as fast as a direct C callback.
1992 2841
1993Example: simple class declaration and watcher initialisation 2842Example: simple class declaration and watcher initialisation
1994 2843
1995 struct myclass 2844 struct myclass
1996 { 2845 {
1997 void io_cb (ev::io &w, int revents) { } 2846 void io_cb (ev::io &w, int revents) { }
1998 } 2847 }
1999 2848
2000 myclass obj; 2849 myclass obj;
2001 ev::io iow; 2850 ev::io iow;
2002 iow.set <myclass, &myclass::io_cb> (&obj); 2851 iow.set <myclass, &myclass::io_cb> (&obj);
2003 2852
2004=item w->set<function> (void *data = 0) 2853=item w->set<function> (void *data = 0)
2005 2854
2006Also sets a callback, but uses a static method or plain function as 2855Also sets a callback, but uses a static method or plain function as
2007callback. The optional C<data> argument will be stored in the watcher's 2856callback. The optional C<data> argument will be stored in the watcher's
2009 2858
2010The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>. 2859The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
2011 2860
2012See the method-C<set> above for more details. 2861See the method-C<set> above for more details.
2013 2862
2014Example: 2863Example: Use a plain function as callback.
2015 2864
2016 static void io_cb (ev::io &w, int revents) { } 2865 static void io_cb (ev::io &w, int revents) { }
2017 iow.set <io_cb> (); 2866 iow.set <io_cb> ();
2018 2867
2019=item w->set (struct ev_loop *) 2868=item w->set (struct ev_loop *)
2020 2869
2021Associates a different C<struct ev_loop> with this watcher. You can only 2870Associates a different C<struct ev_loop> with this watcher. You can only
2022do this when the watcher is inactive (and not pending either). 2871do this when the watcher is inactive (and not pending either).
2023 2872
2024=item w->set ([args]) 2873=item w->set ([arguments])
2025 2874
2026Basically the same as C<ev_TYPE_set>, with the same args. Must be 2875Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
2027called at least once. Unlike the C counterpart, an active watcher gets 2876called at least once. Unlike the C counterpart, an active watcher gets
2028automatically stopped and restarted when reconfiguring it with this 2877automatically stopped and restarted when reconfiguring it with this
2029method. 2878method.
2030 2879
2031=item w->start () 2880=item w->start ()
2055=back 2904=back
2056 2905
2057Example: Define a class with an IO and idle watcher, start one of them in 2906Example: Define a class with an IO and idle watcher, start one of them in
2058the constructor. 2907the constructor.
2059 2908
2060 class myclass 2909 class myclass
2061 { 2910 {
2062 ev_io io; void io_cb (ev::io &w, int revents); 2911 ev::io io ; void io_cb (ev::io &w, int revents);
2063 ev_idle idle void idle_cb (ev::idle &w, int revents); 2912 ev::idle idle; void idle_cb (ev::idle &w, int revents);
2064 2913
2065 myclass (); 2914 myclass (int fd)
2066 } 2915 {
2067
2068 myclass::myclass (int fd)
2069 {
2070 io .set <myclass, &myclass::io_cb > (this); 2916 io .set <myclass, &myclass::io_cb > (this);
2071 idle.set <myclass, &myclass::idle_cb> (this); 2917 idle.set <myclass, &myclass::idle_cb> (this);
2072 2918
2073 io.start (fd, ev::READ); 2919 io.start (fd, ev::READ);
2920 }
2074 } 2921 };
2922
2923
2924=head1 OTHER LANGUAGE BINDINGS
2925
2926Libev does not offer other language bindings itself, but bindings for a
2927number of languages exist in the form of third-party packages. If you know
2928any interesting language binding in addition to the ones listed here, drop
2929me a note.
2930
2931=over 4
2932
2933=item Perl
2934
2935The EV module implements the full libev API and is actually used to test
2936libev. EV is developed together with libev. Apart from the EV core module,
2937there are additional modules that implement libev-compatible interfaces
2938to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
2939C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
2940and C<EV::Glib>).
2941
2942It can be found and installed via CPAN, its homepage is at
2943L<http://software.schmorp.de/pkg/EV>.
2944
2945=item Python
2946
2947Python bindings can be found at L<http://code.google.com/p/pyev/>. It
2948seems to be quite complete and well-documented. Note, however, that the
2949patch they require for libev is outright dangerous as it breaks the ABI
2950for everybody else, and therefore, should never be applied in an installed
2951libev (if python requires an incompatible ABI then it needs to embed
2952libev).
2953
2954=item Ruby
2955
2956Tony Arcieri has written a ruby extension that offers access to a subset
2957of the libev API and adds file handle abstractions, asynchronous DNS and
2958more on top of it. It can be found via gem servers. Its homepage is at
2959L<http://rev.rubyforge.org/>.
2960
2961=item D
2962
2963Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
2964be found at L<http://proj.llucax.com.ar/wiki/evd>.
2965
2966=item Ocaml
2967
2968Erkki Seppala has written Ocaml bindings for libev, to be found at
2969L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
2970
2971=back
2075 2972
2076 2973
2077=head1 MACRO MAGIC 2974=head1 MACRO MAGIC
2078 2975
2079Libev can be compiled with a variety of options, the most fundamantal 2976Libev can be compiled with a variety of options, the most fundamental
2080of which is C<EV_MULTIPLICITY>. This option determines whether (most) 2977of which is C<EV_MULTIPLICITY>. This option determines whether (most)
2081functions and callbacks have an initial C<struct ev_loop *> argument. 2978functions and callbacks have an initial C<struct ev_loop *> argument.
2082 2979
2083To make it easier to write programs that cope with either variant, the 2980To make it easier to write programs that cope with either variant, the
2084following macros are defined: 2981following macros are defined:
2089 2986
2090This provides the loop I<argument> for functions, if one is required ("ev 2987This provides the loop I<argument> for functions, if one is required ("ev
2091loop argument"). The C<EV_A> form is used when this is the sole argument, 2988loop argument"). The C<EV_A> form is used when this is the sole argument,
2092C<EV_A_> is used when other arguments are following. Example: 2989C<EV_A_> is used when other arguments are following. Example:
2093 2990
2094 ev_unref (EV_A); 2991 ev_unref (EV_A);
2095 ev_timer_add (EV_A_ watcher); 2992 ev_timer_add (EV_A_ watcher);
2096 ev_loop (EV_A_ 0); 2993 ev_loop (EV_A_ 0);
2097 2994
2098It assumes the variable C<loop> of type C<struct ev_loop *> is in scope, 2995It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
2099which is often provided by the following macro. 2996which is often provided by the following macro.
2100 2997
2101=item C<EV_P>, C<EV_P_> 2998=item C<EV_P>, C<EV_P_>
2102 2999
2103This provides the loop I<parameter> for functions, if one is required ("ev 3000This provides the loop I<parameter> for functions, if one is required ("ev
2104loop parameter"). The C<EV_P> form is used when this is the sole parameter, 3001loop parameter"). The C<EV_P> form is used when this is the sole parameter,
2105C<EV_P_> is used when other parameters are following. Example: 3002C<EV_P_> is used when other parameters are following. Example:
2106 3003
2107 // this is how ev_unref is being declared 3004 // this is how ev_unref is being declared
2108 static void ev_unref (EV_P); 3005 static void ev_unref (EV_P);
2109 3006
2110 // this is how you can declare your typical callback 3007 // this is how you can declare your typical callback
2111 static void cb (EV_P_ ev_timer *w, int revents) 3008 static void cb (EV_P_ ev_timer *w, int revents)
2112 3009
2113It declares a parameter C<loop> of type C<struct ev_loop *>, quite 3010It declares a parameter C<loop> of type C<struct ev_loop *>, quite
2114suitable for use with C<EV_A>. 3011suitable for use with C<EV_A>.
2115 3012
2116=item C<EV_DEFAULT>, C<EV_DEFAULT_> 3013=item C<EV_DEFAULT>, C<EV_DEFAULT_>
2117 3014
2118Similar to the other two macros, this gives you the value of the default 3015Similar to the other two macros, this gives you the value of the default
2119loop, if multiple loops are supported ("ev loop default"). 3016loop, if multiple loops are supported ("ev loop default").
3017
3018=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
3019
3020Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
3021default loop has been initialised (C<UC> == unchecked). Their behaviour
3022is undefined when the default loop has not been initialised by a previous
3023execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
3024
3025It is often prudent to use C<EV_DEFAULT> when initialising the first
3026watcher in a function but use C<EV_DEFAULT_UC> afterwards.
2120 3027
2121=back 3028=back
2122 3029
2123Example: Declare and initialise a check watcher, utilising the above 3030Example: Declare and initialise a check watcher, utilising the above
2124macros so it will work regardless of whether multiple loops are supported 3031macros so it will work regardless of whether multiple loops are supported
2125or not. 3032or not.
2126 3033
2127 static void 3034 static void
2128 check_cb (EV_P_ ev_timer *w, int revents) 3035 check_cb (EV_P_ ev_timer *w, int revents)
2129 { 3036 {
2130 ev_check_stop (EV_A_ w); 3037 ev_check_stop (EV_A_ w);
2131 } 3038 }
2132 3039
2133 ev_check check; 3040 ev_check check;
2134 ev_check_init (&check, check_cb); 3041 ev_check_init (&check, check_cb);
2135 ev_check_start (EV_DEFAULT_ &check); 3042 ev_check_start (EV_DEFAULT_ &check);
2136 ev_loop (EV_DEFAULT_ 0); 3043 ev_loop (EV_DEFAULT_ 0);
2137 3044
2138=head1 EMBEDDING 3045=head1 EMBEDDING
2139 3046
2140Libev can (and often is) directly embedded into host 3047Libev can (and often is) directly embedded into host
2141applications. Examples of applications that embed it include the Deliantra 3048applications. Examples of applications that embed it include the Deliantra
2148libev somewhere in your source tree). 3055libev somewhere in your source tree).
2149 3056
2150=head2 FILESETS 3057=head2 FILESETS
2151 3058
2152Depending on what features you need you need to include one or more sets of files 3059Depending on what features you need you need to include one or more sets of files
2153in your app. 3060in your application.
2154 3061
2155=head3 CORE EVENT LOOP 3062=head3 CORE EVENT LOOP
2156 3063
2157To include only the libev core (all the C<ev_*> functions), with manual 3064To include only the libev core (all the C<ev_*> functions), with manual
2158configuration (no autoconf): 3065configuration (no autoconf):
2159 3066
2160 #define EV_STANDALONE 1 3067 #define EV_STANDALONE 1
2161 #include "ev.c" 3068 #include "ev.c"
2162 3069
2163This will automatically include F<ev.h>, too, and should be done in a 3070This will automatically include F<ev.h>, too, and should be done in a
2164single C source file only to provide the function implementations. To use 3071single C source file only to provide the function implementations. To use
2165it, do the same for F<ev.h> in all files wishing to use this API (best 3072it, do the same for F<ev.h> in all files wishing to use this API (best
2166done by writing a wrapper around F<ev.h> that you can include instead and 3073done by writing a wrapper around F<ev.h> that you can include instead and
2167where you can put other configuration options): 3074where you can put other configuration options):
2168 3075
2169 #define EV_STANDALONE 1 3076 #define EV_STANDALONE 1
2170 #include "ev.h" 3077 #include "ev.h"
2171 3078
2172Both header files and implementation files can be compiled with a C++ 3079Both header files and implementation files can be compiled with a C++
2173compiler (at least, thats a stated goal, and breakage will be treated 3080compiler (at least, thats a stated goal, and breakage will be treated
2174as a bug). 3081as a bug).
2175 3082
2176You need the following files in your source tree, or in a directory 3083You need the following files in your source tree, or in a directory
2177in your include path (e.g. in libev/ when using -Ilibev): 3084in your include path (e.g. in libev/ when using -Ilibev):
2178 3085
2179 ev.h 3086 ev.h
2180 ev.c 3087 ev.c
2181 ev_vars.h 3088 ev_vars.h
2182 ev_wrap.h 3089 ev_wrap.h
2183 3090
2184 ev_win32.c required on win32 platforms only 3091 ev_win32.c required on win32 platforms only
2185 3092
2186 ev_select.c only when select backend is enabled (which is enabled by default) 3093 ev_select.c only when select backend is enabled (which is enabled by default)
2187 ev_poll.c only when poll backend is enabled (disabled by default) 3094 ev_poll.c only when poll backend is enabled (disabled by default)
2188 ev_epoll.c only when the epoll backend is enabled (disabled by default) 3095 ev_epoll.c only when the epoll backend is enabled (disabled by default)
2189 ev_kqueue.c only when the kqueue backend is enabled (disabled by default) 3096 ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
2190 ev_port.c only when the solaris port backend is enabled (disabled by default) 3097 ev_port.c only when the solaris port backend is enabled (disabled by default)
2191 3098
2192F<ev.c> includes the backend files directly when enabled, so you only need 3099F<ev.c> includes the backend files directly when enabled, so you only need
2193to compile this single file. 3100to compile this single file.
2194 3101
2195=head3 LIBEVENT COMPATIBILITY API 3102=head3 LIBEVENT COMPATIBILITY API
2196 3103
2197To include the libevent compatibility API, also include: 3104To include the libevent compatibility API, also include:
2198 3105
2199 #include "event.c" 3106 #include "event.c"
2200 3107
2201in the file including F<ev.c>, and: 3108in the file including F<ev.c>, and:
2202 3109
2203 #include "event.h" 3110 #include "event.h"
2204 3111
2205in the files that want to use the libevent API. This also includes F<ev.h>. 3112in the files that want to use the libevent API. This also includes F<ev.h>.
2206 3113
2207You need the following additional files for this: 3114You need the following additional files for this:
2208 3115
2209 event.h 3116 event.h
2210 event.c 3117 event.c
2211 3118
2212=head3 AUTOCONF SUPPORT 3119=head3 AUTOCONF SUPPORT
2213 3120
2214Instead of using C<EV_STANDALONE=1> and providing your config in 3121Instead of using C<EV_STANDALONE=1> and providing your configuration in
2215whatever way you want, you can also C<m4_include([libev.m4])> in your 3122whatever way you want, you can also C<m4_include([libev.m4])> in your
2216F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then 3123F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
2217include F<config.h> and configure itself accordingly. 3124include F<config.h> and configure itself accordingly.
2218 3125
2219For this of course you need the m4 file: 3126For this of course you need the m4 file:
2220 3127
2221 libev.m4 3128 libev.m4
2222 3129
2223=head2 PREPROCESSOR SYMBOLS/MACROS 3130=head2 PREPROCESSOR SYMBOLS/MACROS
2224 3131
2225Libev can be configured via a variety of preprocessor symbols you have to define 3132Libev can be configured via a variety of preprocessor symbols you have to
2226before including any of its files. The default is not to build for multiplicity 3133define before including any of its files. The default in the absence of
2227and only include the select backend. 3134autoconf is documented for every option.
2228 3135
2229=over 4 3136=over 4
2230 3137
2231=item EV_STANDALONE 3138=item EV_STANDALONE
2232 3139
2237F<event.h> that are not directly supported by the libev core alone. 3144F<event.h> that are not directly supported by the libev core alone.
2238 3145
2239=item EV_USE_MONOTONIC 3146=item EV_USE_MONOTONIC
2240 3147
2241If defined to be C<1>, libev will try to detect the availability of the 3148If defined to be C<1>, libev will try to detect the availability of the
2242monotonic clock option at both compiletime and runtime. Otherwise no use 3149monotonic clock option at both compile time and runtime. Otherwise no use
2243of the monotonic clock option will be attempted. If you enable this, you 3150of the monotonic clock option will be attempted. If you enable this, you
2244usually have to link against librt or something similar. Enabling it when 3151usually have to link against librt or something similar. Enabling it when
2245the functionality isn't available is safe, though, althoguh you have 3152the functionality isn't available is safe, though, although you have
2246to make sure you link against any libraries where the C<clock_gettime> 3153to make sure you link against any libraries where the C<clock_gettime>
2247function is hiding in (often F<-lrt>). 3154function is hiding in (often F<-lrt>).
2248 3155
2249=item EV_USE_REALTIME 3156=item EV_USE_REALTIME
2250 3157
2251If defined to be C<1>, libev will try to detect the availability of the 3158If defined to be C<1>, libev will try to detect the availability of the
2252realtime clock option at compiletime (and assume its availability at 3159real-time clock option at compile time (and assume its availability at
2253runtime if successful). Otherwise no use of the realtime clock option will 3160runtime if successful). Otherwise no use of the real-time clock option will
2254be attempted. This effectively replaces C<gettimeofday> by C<clock_get 3161be attempted. This effectively replaces C<gettimeofday> by C<clock_get
2255(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the 3162(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
2256note about libraries in the description of C<EV_USE_MONOTONIC>, though. 3163note about libraries in the description of C<EV_USE_MONOTONIC>, though.
2257 3164
3165=item EV_USE_NANOSLEEP
3166
3167If defined to be C<1>, libev will assume that C<nanosleep ()> is available
3168and will use it for delays. Otherwise it will use C<select ()>.
3169
3170=item EV_USE_EVENTFD
3171
3172If defined to be C<1>, then libev will assume that C<eventfd ()> is
3173available and will probe for kernel support at runtime. This will improve
3174C<ev_signal> and C<ev_async> performance and reduce resource consumption.
3175If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
31762.7 or newer, otherwise disabled.
3177
2258=item EV_USE_SELECT 3178=item EV_USE_SELECT
2259 3179
2260If undefined or defined to be C<1>, libev will compile in support for the 3180If undefined or defined to be C<1>, libev will compile in support for the
2261C<select>(2) backend. No attempt at autodetection will be done: if no 3181C<select>(2) backend. No attempt at auto-detection will be done: if no
2262other method takes over, select will be it. Otherwise the select backend 3182other method takes over, select will be it. Otherwise the select backend
2263will not be compiled in. 3183will not be compiled in.
2264 3184
2265=item EV_SELECT_USE_FD_SET 3185=item EV_SELECT_USE_FD_SET
2266 3186
2267If defined to C<1>, then the select backend will use the system C<fd_set> 3187If defined to C<1>, then the select backend will use the system C<fd_set>
2268structure. This is useful if libev doesn't compile due to a missing 3188structure. This is useful if libev doesn't compile due to a missing
2269C<NFDBITS> or C<fd_mask> definition or it misguesses the bitset layout on 3189C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout on
2270exotic systems. This usually limits the range of file descriptors to some 3190exotic systems. This usually limits the range of file descriptors to some
2271low limit such as 1024 or might have other limitations (winsocket only 3191low limit such as 1024 or might have other limitations (winsocket only
2272allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might 3192allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
2273influence the size of the C<fd_set> used. 3193influence the size of the C<fd_set> used.
2274 3194
2280be used is the winsock select). This means that it will call 3200be used is the winsock select). This means that it will call
2281C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise, 3201C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2282it is assumed that all these functions actually work on fds, even 3202it is assumed that all these functions actually work on fds, even
2283on win32. Should not be defined on non-win32 platforms. 3203on win32. Should not be defined on non-win32 platforms.
2284 3204
3205=item EV_FD_TO_WIN32_HANDLE
3206
3207If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
3208file descriptors to socket handles. When not defining this symbol (the
3209default), then libev will call C<_get_osfhandle>, which is usually
3210correct. In some cases, programs use their own file descriptor management,
3211in which case they can provide this function to map fds to socket handles.
3212
2285=item EV_USE_POLL 3213=item EV_USE_POLL
2286 3214
2287If defined to be C<1>, libev will compile in support for the C<poll>(2) 3215If defined to be C<1>, libev will compile in support for the C<poll>(2)
2288backend. Otherwise it will be enabled on non-win32 platforms. It 3216backend. Otherwise it will be enabled on non-win32 platforms. It
2289takes precedence over select. 3217takes precedence over select.
2290 3218
2291=item EV_USE_EPOLL 3219=item EV_USE_EPOLL
2292 3220
2293If defined to be C<1>, libev will compile in support for the Linux 3221If defined to be C<1>, libev will compile in support for the Linux
2294C<epoll>(7) backend. Its availability will be detected at runtime, 3222C<epoll>(7) backend. Its availability will be detected at runtime,
2295otherwise another method will be used as fallback. This is the 3223otherwise another method will be used as fallback. This is the preferred
2296preferred backend for GNU/Linux systems. 3224backend for GNU/Linux systems. If undefined, it will be enabled if the
3225headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
2297 3226
2298=item EV_USE_KQUEUE 3227=item EV_USE_KQUEUE
2299 3228
2300If defined to be C<1>, libev will compile in support for the BSD style 3229If defined to be C<1>, libev will compile in support for the BSD style
2301C<kqueue>(2) backend. Its actual availability will be detected at runtime, 3230C<kqueue>(2) backend. Its actual availability will be detected at runtime,
2314otherwise another method will be used as fallback. This is the preferred 3243otherwise another method will be used as fallback. This is the preferred
2315backend for Solaris 10 systems. 3244backend for Solaris 10 systems.
2316 3245
2317=item EV_USE_DEVPOLL 3246=item EV_USE_DEVPOLL
2318 3247
2319reserved for future expansion, works like the USE symbols above. 3248Reserved for future expansion, works like the USE symbols above.
2320 3249
2321=item EV_USE_INOTIFY 3250=item EV_USE_INOTIFY
2322 3251
2323If defined to be C<1>, libev will compile in support for the Linux inotify 3252If defined to be C<1>, libev will compile in support for the Linux inotify
2324interface to speed up C<ev_stat> watchers. Its actual availability will 3253interface to speed up C<ev_stat> watchers. Its actual availability will
2325be detected at runtime. 3254be detected at runtime. If undefined, it will be enabled if the headers
3255indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
3256
3257=item EV_ATOMIC_T
3258
3259Libev requires an integer type (suitable for storing C<0> or C<1>) whose
3260access is atomic with respect to other threads or signal contexts. No such
3261type is easily found in the C language, so you can provide your own type
3262that you know is safe for your purposes. It is used both for signal handler "locking"
3263as well as for signal and thread safety in C<ev_async> watchers.
3264
3265In the absence of this define, libev will use C<sig_atomic_t volatile>
3266(from F<signal.h>), which is usually good enough on most platforms.
2326 3267
2327=item EV_H 3268=item EV_H
2328 3269
2329The name of the F<ev.h> header file used to include it. The default if 3270The name of the F<ev.h> header file used to include it. The default if
2330undefined is C<< <ev.h> >> in F<event.h> and C<"ev.h"> in F<ev.c>. This 3271undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2331can be used to virtually rename the F<ev.h> header file in case of conflicts. 3272used to virtually rename the F<ev.h> header file in case of conflicts.
2332 3273
2333=item EV_CONFIG_H 3274=item EV_CONFIG_H
2334 3275
2335If C<EV_STANDALONE> isn't C<1>, this variable can be used to override 3276If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
2336F<ev.c>'s idea of where to find the F<config.h> file, similarly to 3277F<ev.c>'s idea of where to find the F<config.h> file, similarly to
2337C<EV_H>, above. 3278C<EV_H>, above.
2338 3279
2339=item EV_EVENT_H 3280=item EV_EVENT_H
2340 3281
2341Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea 3282Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
2342of how the F<event.h> header can be found. 3283of how the F<event.h> header can be found, the default is C<"event.h">.
2343 3284
2344=item EV_PROTOTYPES 3285=item EV_PROTOTYPES
2345 3286
2346If defined to be C<0>, then F<ev.h> will not define any function 3287If defined to be C<0>, then F<ev.h> will not define any function
2347prototypes, but still define all the structs and other symbols. This is 3288prototypes, but still define all the structs and other symbols. This is
2368When doing priority-based operations, libev usually has to linearly search 3309When doing priority-based operations, libev usually has to linearly search
2369all the priorities, so having many of them (hundreds) uses a lot of space 3310all the priorities, so having many of them (hundreds) uses a lot of space
2370and time, so using the defaults of five priorities (-2 .. +2) is usually 3311and time, so using the defaults of five priorities (-2 .. +2) is usually
2371fine. 3312fine.
2372 3313
2373If your embedding app does not need any priorities, defining these both to 3314If your embedding application does not need any priorities, defining these
2374C<0> will save some memory and cpu. 3315both to C<0> will save some memory and CPU.
2375 3316
2376=item EV_PERIODIC_ENABLE 3317=item EV_PERIODIC_ENABLE
2377 3318
2378If undefined or defined to be C<1>, then periodic timers are supported. If 3319If undefined or defined to be C<1>, then periodic timers are supported. If
2379defined to be C<0>, then they are not. Disabling them saves a few kB of 3320defined to be C<0>, then they are not. Disabling them saves a few kB of
2386code. 3327code.
2387 3328
2388=item EV_EMBED_ENABLE 3329=item EV_EMBED_ENABLE
2389 3330
2390If undefined or defined to be C<1>, then embed watchers are supported. If 3331If undefined or defined to be C<1>, then embed watchers are supported. If
2391defined to be C<0>, then they are not. 3332defined to be C<0>, then they are not. Embed watchers rely on most other
3333watcher types, which therefore must not be disabled.
2392 3334
2393=item EV_STAT_ENABLE 3335=item EV_STAT_ENABLE
2394 3336
2395If undefined or defined to be C<1>, then stat watchers are supported. If 3337If undefined or defined to be C<1>, then stat watchers are supported. If
2396defined to be C<0>, then they are not. 3338defined to be C<0>, then they are not.
2398=item EV_FORK_ENABLE 3340=item EV_FORK_ENABLE
2399 3341
2400If undefined or defined to be C<1>, then fork watchers are supported. If 3342If undefined or defined to be C<1>, then fork watchers are supported. If
2401defined to be C<0>, then they are not. 3343defined to be C<0>, then they are not.
2402 3344
3345=item EV_ASYNC_ENABLE
3346
3347If undefined or defined to be C<1>, then async watchers are supported. If
3348defined to be C<0>, then they are not.
3349
2403=item EV_MINIMAL 3350=item EV_MINIMAL
2404 3351
2405If you need to shave off some kilobytes of code at the expense of some 3352If you need to shave off some kilobytes of code at the expense of some
2406speed, define this symbol to C<1>. Currently only used for gcc to override 3353speed, define this symbol to C<1>. Currently this is used to override some
2407some inlining decisions, saves roughly 30% codesize of amd64. 3354inlining decisions, saves roughly 30% code size on amd64. It also selects a
3355much smaller 2-heap for timer management over the default 4-heap.
2408 3356
2409=item EV_PID_HASHSIZE 3357=item EV_PID_HASHSIZE
2410 3358
2411C<ev_child> watchers use a small hash table to distribute workload by 3359C<ev_child> watchers use a small hash table to distribute workload by
2412pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more 3360pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
2413than enough. If you need to manage thousands of children you might want to 3361than enough. If you need to manage thousands of children you might want to
2414increase this value (I<must> be a power of two). 3362increase this value (I<must> be a power of two).
2415 3363
2416=item EV_INOTIFY_HASHSIZE 3364=item EV_INOTIFY_HASHSIZE
2417 3365
2418C<ev_staz> watchers use a small hash table to distribute workload by 3366C<ev_stat> watchers use a small hash table to distribute workload by
2419inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>), 3367inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
2420usually more than enough. If you need to manage thousands of C<ev_stat> 3368usually more than enough. If you need to manage thousands of C<ev_stat>
2421watchers you might want to increase this value (I<must> be a power of 3369watchers you might want to increase this value (I<must> be a power of
2422two). 3370two).
2423 3371
3372=item EV_USE_4HEAP
3373
3374Heaps are not very cache-efficient. To improve the cache-efficiency of the
3375timer and periodics heaps, libev uses a 4-heap when this symbol is defined
3376to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
3377faster performance with many (thousands) of watchers.
3378
3379The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
3380(disabled).
3381
3382=item EV_HEAP_CACHE_AT
3383
3384Heaps are not very cache-efficient. To improve the cache-efficiency of the
3385timer and periodics heaps, libev can cache the timestamp (I<at>) within
3386the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
3387which uses 8-12 bytes more per watcher and a few hundred bytes more code,
3388but avoids random read accesses on heap changes. This improves performance
3389noticeably with many (hundreds) of watchers.
3390
3391The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
3392(disabled).
3393
3394=item EV_VERIFY
3395
3396Controls how much internal verification (see C<ev_loop_verify ()>) will
3397be done: If set to C<0>, no internal verification code will be compiled
3398in. If set to C<1>, then verification code will be compiled in, but not
3399called. If set to C<2>, then the internal verification code will be
3400called once per loop, which can slow down libev. If set to C<3>, then the
3401verification code will be called very frequently, which will slow down
3402libev considerably.
3403
3404The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
3405C<0>.
3406
2424=item EV_COMMON 3407=item EV_COMMON
2425 3408
2426By default, all watchers have a C<void *data> member. By redefining 3409By default, all watchers have a C<void *data> member. By redefining
2427this macro to a something else you can include more and other types of 3410this macro to a something else you can include more and other types of
2428members. You have to define it each time you include one of the files, 3411members. You have to define it each time you include one of the files,
2429though, and it must be identical each time. 3412though, and it must be identical each time.
2430 3413
2431For example, the perl EV module uses something like this: 3414For example, the perl EV module uses something like this:
2432 3415
2433 #define EV_COMMON \ 3416 #define EV_COMMON \
2434 SV *self; /* contains this struct */ \ 3417 SV *self; /* contains this struct */ \
2435 SV *cb_sv, *fh /* note no trailing ";" */ 3418 SV *cb_sv, *fh /* note no trailing ";" */
2436 3419
2437=item EV_CB_DECLARE (type) 3420=item EV_CB_DECLARE (type)
2438 3421
2439=item EV_CB_INVOKE (watcher, revents) 3422=item EV_CB_INVOKE (watcher, revents)
2440 3423
2441=item ev_set_cb (ev, cb) 3424=item ev_set_cb (ev, cb)
2442 3425
2443Can be used to change the callback member declaration in each watcher, 3426Can be used to change the callback member declaration in each watcher,
2444and the way callbacks are invoked and set. Must expand to a struct member 3427and the way callbacks are invoked and set. Must expand to a struct member
2445definition and a statement, respectively. See the F<ev.v> header file for 3428definition and a statement, respectively. See the F<ev.h> header file for
2446their default definitions. One possible use for overriding these is to 3429their default definitions. One possible use for overriding these is to
2447avoid the C<struct ev_loop *> as first argument in all cases, or to use 3430avoid the C<struct ev_loop *> as first argument in all cases, or to use
2448method calls instead of plain function calls in C++. 3431method calls instead of plain function calls in C++.
2449 3432
3433=back
3434
2450=head2 EXPORTED API SYMBOLS 3435=head2 EXPORTED API SYMBOLS
2451 3436
2452If you need to re-export the API (e.g. via a dll) and you need a list of 3437If you need to re-export the API (e.g. via a DLL) and you need a list of
2453exported symbols, you can use the provided F<Symbol.*> files which list 3438exported symbols, you can use the provided F<Symbol.*> files which list
2454all public symbols, one per line: 3439all public symbols, one per line:
2455 3440
2456 Symbols.ev for libev proper 3441 Symbols.ev for libev proper
2457 Symbols.event for the libevent emulation 3442 Symbols.event for the libevent emulation
2458 3443
2459This can also be used to rename all public symbols to avoid clashes with 3444This can also be used to rename all public symbols to avoid clashes with
2460multiple versions of libev linked together (which is obviously bad in 3445multiple versions of libev linked together (which is obviously bad in
2461itself, but sometimes it is inconvinient to avoid this). 3446itself, but sometimes it is inconvenient to avoid this).
2462 3447
2463A sed comamnd like this will create wrapper C<#define>'s that you need to 3448A sed command like this will create wrapper C<#define>'s that you need to
2464include before including F<ev.h>: 3449include before including F<ev.h>:
2465 3450
2466 <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h 3451 <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
2467 3452
2468This would create a file F<wrap.h> which essentially looks like this: 3453This would create a file F<wrap.h> which essentially looks like this:
2483file. 3468file.
2484 3469
2485The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file 3470The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
2486that everybody includes and which overrides some configure choices: 3471that everybody includes and which overrides some configure choices:
2487 3472
2488 #define EV_MINIMAL 1 3473 #define EV_MINIMAL 1
2489 #define EV_USE_POLL 0 3474 #define EV_USE_POLL 0
2490 #define EV_MULTIPLICITY 0 3475 #define EV_MULTIPLICITY 0
2491 #define EV_PERIODIC_ENABLE 0 3476 #define EV_PERIODIC_ENABLE 0
2492 #define EV_STAT_ENABLE 0 3477 #define EV_STAT_ENABLE 0
2493 #define EV_FORK_ENABLE 0 3478 #define EV_FORK_ENABLE 0
2494 #define EV_CONFIG_H <config.h> 3479 #define EV_CONFIG_H <config.h>
2495 #define EV_MINPRI 0 3480 #define EV_MINPRI 0
2496 #define EV_MAXPRI 0 3481 #define EV_MAXPRI 0
2497 3482
2498 #include "ev++.h" 3483 #include "ev++.h"
2499 3484
2500And a F<ev_cpp.C> implementation file that contains libev proper and is compiled: 3485And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
2501 3486
2502 #include "ev_cpp.h" 3487 #include "ev_cpp.h"
2503 #include "ev.c" 3488 #include "ev.c"
2504 3489
3490=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
2505 3491
3492=head2 THREADS AND COROUTINES
3493
3494=head3 THREADS
3495
3496All libev functions are reentrant and thread-safe unless explicitly
3497documented otherwise, but libev implements no locking itself. This means
3498that you can use as many loops as you want in parallel, as long as there
3499are no concurrent calls into any libev function with the same loop
3500parameter (C<ev_default_*> calls have an implicit default loop parameter,
3501of course): libev guarantees that different event loops share no data
3502structures that need any locking.
3503
3504Or to put it differently: calls with different loop parameters can be done
3505concurrently from multiple threads, calls with the same loop parameter
3506must be done serially (but can be done from different threads, as long as
3507only one thread ever is inside a call at any point in time, e.g. by using
3508a mutex per loop).
3509
3510Specifically to support threads (and signal handlers), libev implements
3511so-called C<ev_async> watchers, which allow some limited form of
3512concurrency on the same event loop, namely waking it up "from the
3513outside".
3514
3515If you want to know which design (one loop, locking, or multiple loops
3516without or something else still) is best for your problem, then I cannot
3517help you, but here is some generic advice:
3518
3519=over 4
3520
3521=item * most applications have a main thread: use the default libev loop
3522in that thread, or create a separate thread running only the default loop.
3523
3524This helps integrating other libraries or software modules that use libev
3525themselves and don't care/know about threading.
3526
3527=item * one loop per thread is usually a good model.
3528
3529Doing this is almost never wrong, sometimes a better-performance model
3530exists, but it is always a good start.
3531
3532=item * other models exist, such as the leader/follower pattern, where one
3533loop is handed through multiple threads in a kind of round-robin fashion.
3534
3535Choosing a model is hard - look around, learn, know that usually you can do
3536better than you currently do :-)
3537
3538=item * often you need to talk to some other thread which blocks in the
3539event loop.
3540
3541C<ev_async> watchers can be used to wake them up from other threads safely
3542(or from signal contexts...).
3543
3544An example use would be to communicate signals or other events that only
3545work in the default loop by registering the signal watcher with the
3546default loop and triggering an C<ev_async> watcher from the default loop
3547watcher callback into the event loop interested in the signal.
3548
3549=back
3550
3551=head3 COROUTINES
3552
3553Libev is very accommodating to coroutines ("cooperative threads"):
3554libev fully supports nesting calls to its functions from different
3555coroutines (e.g. you can call C<ev_loop> on the same loop from two
3556different coroutines, and switch freely between both coroutines running the
3557loop, as long as you don't confuse yourself). The only exception is that
3558you must not do this from C<ev_periodic> reschedule callbacks.
3559
3560Care has been taken to ensure that libev does not keep local state inside
3561C<ev_loop>, and other calls do not usually allow for coroutine switches as
3562they do not clal any callbacks.
3563
3564=head2 COMPILER WARNINGS
3565
3566Depending on your compiler and compiler settings, you might get no or a
3567lot of warnings when compiling libev code. Some people are apparently
3568scared by this.
3569
3570However, these are unavoidable for many reasons. For one, each compiler
3571has different warnings, and each user has different tastes regarding
3572warning options. "Warn-free" code therefore cannot be a goal except when
3573targeting a specific compiler and compiler-version.
3574
3575Another reason is that some compiler warnings require elaborate
3576workarounds, or other changes to the code that make it less clear and less
3577maintainable.
3578
3579And of course, some compiler warnings are just plain stupid, or simply
3580wrong (because they don't actually warn about the condition their message
3581seems to warn about). For example, certain older gcc versions had some
3582warnings that resulted an extreme number of false positives. These have
3583been fixed, but some people still insist on making code warn-free with
3584such buggy versions.
3585
3586While libev is written to generate as few warnings as possible,
3587"warn-free" code is not a goal, and it is recommended not to build libev
3588with any compiler warnings enabled unless you are prepared to cope with
3589them (e.g. by ignoring them). Remember that warnings are just that:
3590warnings, not errors, or proof of bugs.
3591
3592
3593=head2 VALGRIND
3594
3595Valgrind has a special section here because it is a popular tool that is
3596highly useful. Unfortunately, valgrind reports are very hard to interpret.
3597
3598If you think you found a bug (memory leak, uninitialised data access etc.)
3599in libev, then check twice: If valgrind reports something like:
3600
3601 ==2274== definitely lost: 0 bytes in 0 blocks.
3602 ==2274== possibly lost: 0 bytes in 0 blocks.
3603 ==2274== still reachable: 256 bytes in 1 blocks.
3604
3605Then there is no memory leak, just as memory accounted to global variables
3606is not a memleak - the memory is still being refernced, and didn't leak.
3607
3608Similarly, under some circumstances, valgrind might report kernel bugs
3609as if it were a bug in libev (e.g. in realloc or in the poll backend,
3610although an acceptable workaround has been found here), or it might be
3611confused.
3612
3613Keep in mind that valgrind is a very good tool, but only a tool. Don't
3614make it into some kind of religion.
3615
3616If you are unsure about something, feel free to contact the mailing list
3617with the full valgrind report and an explanation on why you think this
3618is a bug in libev (best check the archives, too :). However, don't be
3619annoyed when you get a brisk "this is no bug" answer and take the chance
3620of learning how to interpret valgrind properly.
3621
3622If you need, for some reason, empty reports from valgrind for your project
3623I suggest using suppression lists.
3624
3625
3626=head1 PORTABILITY NOTES
3627
3628=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
3629
3630Win32 doesn't support any of the standards (e.g. POSIX) that libev
3631requires, and its I/O model is fundamentally incompatible with the POSIX
3632model. Libev still offers limited functionality on this platform in
3633the form of the C<EVBACKEND_SELECT> backend, and only supports socket
3634descriptors. This only applies when using Win32 natively, not when using
3635e.g. cygwin.
3636
3637Lifting these limitations would basically require the full
3638re-implementation of the I/O system. If you are into these kinds of
3639things, then note that glib does exactly that for you in a very portable
3640way (note also that glib is the slowest event library known to man).
3641
3642There is no supported compilation method available on windows except
3643embedding it into other applications.
3644
3645Not a libev limitation but worth mentioning: windows apparently doesn't
3646accept large writes: instead of resulting in a partial write, windows will
3647either accept everything or return C<ENOBUFS> if the buffer is too large,
3648so make sure you only write small amounts into your sockets (less than a
3649megabyte seems safe, but this apparently depends on the amount of memory
3650available).
3651
3652Due to the many, low, and arbitrary limits on the win32 platform and
3653the abysmal performance of winsockets, using a large number of sockets
3654is not recommended (and not reasonable). If your program needs to use
3655more than a hundred or so sockets, then likely it needs to use a totally
3656different implementation for windows, as libev offers the POSIX readiness
3657notification model, which cannot be implemented efficiently on windows
3658(Microsoft monopoly games).
3659
3660A typical way to use libev under windows is to embed it (see the embedding
3661section for details) and use the following F<evwrap.h> header file instead
3662of F<ev.h>:
3663
3664 #define EV_STANDALONE /* keeps ev from requiring config.h */
3665 #define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
3666
3667 #include "ev.h"
3668
3669And compile the following F<evwrap.c> file into your project (make sure
3670you do I<not> compile the F<ev.c> or any other embedded source files!):
3671
3672 #include "evwrap.h"
3673 #include "ev.c"
3674
3675=over 4
3676
3677=item The winsocket select function
3678
3679The winsocket C<select> function doesn't follow POSIX in that it
3680requires socket I<handles> and not socket I<file descriptors> (it is
3681also extremely buggy). This makes select very inefficient, and also
3682requires a mapping from file descriptors to socket handles (the Microsoft
3683C runtime provides the function C<_open_osfhandle> for this). See the
3684discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
3685C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
3686
3687The configuration for a "naked" win32 using the Microsoft runtime
3688libraries and raw winsocket select is:
3689
3690 #define EV_USE_SELECT 1
3691 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
3692
3693Note that winsockets handling of fd sets is O(n), so you can easily get a
3694complexity in the O(n²) range when using win32.
3695
3696=item Limited number of file descriptors
3697
3698Windows has numerous arbitrary (and low) limits on things.
3699
3700Early versions of winsocket's select only supported waiting for a maximum
3701of C<64> handles (probably owning to the fact that all windows kernels
3702can only wait for C<64> things at the same time internally; Microsoft
3703recommends spawning a chain of threads and wait for 63 handles and the
3704previous thread in each. Great).
3705
3706Newer versions support more handles, but you need to define C<FD_SETSIZE>
3707to some high number (e.g. C<2048>) before compiling the winsocket select
3708call (which might be in libev or elsewhere, for example, perl does its own
3709select emulation on windows).
3710
3711Another limit is the number of file descriptors in the Microsoft runtime
3712libraries, which by default is C<64> (there must be a hidden I<64> fetish
3713or something like this inside Microsoft). You can increase this by calling
3714C<_setmaxstdio>, which can increase this limit to C<2048> (another
3715arbitrary limit), but is broken in many versions of the Microsoft runtime
3716libraries.
3717
3718This might get you to about C<512> or C<2048> sockets (depending on
3719windows version and/or the phase of the moon). To get more, you need to
3720wrap all I/O functions and provide your own fd management, but the cost of
3721calling select (O(n²)) will likely make this unworkable.
3722
3723=back
3724
3725=head2 PORTABILITY REQUIREMENTS
3726
3727In addition to a working ISO-C implementation and of course the
3728backend-specific APIs, libev relies on a few additional extensions:
3729
3730=over 4
3731
3732=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
3733calling conventions regardless of C<ev_watcher_type *>.
3734
3735Libev assumes not only that all watcher pointers have the same internal
3736structure (guaranteed by POSIX but not by ISO C for example), but it also
3737assumes that the same (machine) code can be used to call any watcher
3738callback: The watcher callbacks have different type signatures, but libev
3739calls them using an C<ev_watcher *> internally.
3740
3741=item C<sig_atomic_t volatile> must be thread-atomic as well
3742
3743The type C<sig_atomic_t volatile> (or whatever is defined as
3744C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
3745threads. This is not part of the specification for C<sig_atomic_t>, but is
3746believed to be sufficiently portable.
3747
3748=item C<sigprocmask> must work in a threaded environment
3749
3750Libev uses C<sigprocmask> to temporarily block signals. This is not
3751allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
3752pthread implementations will either allow C<sigprocmask> in the "main
3753thread" or will block signals process-wide, both behaviours would
3754be compatible with libev. Interaction between C<sigprocmask> and
3755C<pthread_sigmask> could complicate things, however.
3756
3757The most portable way to handle signals is to block signals in all threads
3758except the initial one, and run the default loop in the initial thread as
3759well.
3760
3761=item C<long> must be large enough for common memory allocation sizes
3762
3763To improve portability and simplify its API, libev uses C<long> internally
3764instead of C<size_t> when allocating its data structures. On non-POSIX
3765systems (Microsoft...) this might be unexpectedly low, but is still at
3766least 31 bits everywhere, which is enough for hundreds of millions of
3767watchers.
3768
3769=item C<double> must hold a time value in seconds with enough accuracy
3770
3771The type C<double> is used to represent timestamps. It is required to
3772have at least 51 bits of mantissa (and 9 bits of exponent), which is good
3773enough for at least into the year 4000. This requirement is fulfilled by
3774implementations implementing IEEE 754 (basically all existing ones).
3775
3776=back
3777
3778If you know of other additional requirements drop me a note.
3779
3780
2506=head1 COMPLEXITIES 3781=head1 ALGORITHMIC COMPLEXITIES
2507 3782
2508In this section the complexities of (many of) the algorithms used inside 3783In this section the complexities of (many of) the algorithms used inside
2509libev will be explained. For complexity discussions about backends see the 3784libev will be documented. For complexity discussions about backends see
2510documentation for C<ev_default_init>. 3785the documentation for C<ev_default_init>.
2511 3786
2512All of the following are about amortised time: If an array needs to be 3787All of the following are about amortised time: If an array needs to be
2513extended, libev needs to realloc and move the whole array, but this 3788extended, libev needs to realloc and move the whole array, but this
2514happens asymptotically never with higher number of elements, so O(1) might 3789happens asymptotically rarer with higher number of elements, so O(1) might
2515mean it might do a lengthy realloc operation in rare cases, but on average 3790mean that libev does a lengthy realloc operation in rare cases, but on
2516it is much faster and asymptotically approaches constant time. 3791average it is much faster and asymptotically approaches constant time.
2517 3792
2518=over 4 3793=over 4
2519 3794
2520=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers) 3795=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
2521 3796
2522This means that, when you have a watcher that triggers in one hour and 3797This means that, when you have a watcher that triggers in one hour and
2523there are 100 watchers that would trigger before that then inserting will 3798there are 100 watchers that would trigger before that, then inserting will
2524have to skip those 100 watchers. 3799have to skip roughly seven (C<ld 100>) of these watchers.
2525 3800
2526=item Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers) 3801=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
2527 3802
2528That means that for changing a timer costs less than removing/adding them 3803That means that changing a timer costs less than removing/adding them,
2529as only the relative motion in the event queue has to be paid for. 3804as only the relative motion in the event queue has to be paid for.
2530 3805
2531=item Starting io/check/prepare/idle/signal/child watchers: O(1) 3806=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
2532 3807
2533These just add the watcher into an array or at the head of a list. 3808These just add the watcher into an array or at the head of a list.
3809
2534=item Stopping check/prepare/idle watchers: O(1) 3810=item Stopping check/prepare/idle/fork/async watchers: O(1)
2535 3811
2536=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE)) 3812=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
2537 3813
2538These watchers are stored in lists then need to be walked to find the 3814These watchers are stored in lists, so they need to be walked to find the
2539correct watcher to remove. The lists are usually short (you don't usually 3815correct watcher to remove. The lists are usually short (you don't usually
2540have many watchers waiting for the same fd or signal). 3816have many watchers waiting for the same fd or signal: one is typical, two
3817is rare).
2541 3818
2542=item Finding the next timer per loop iteration: O(1) 3819=item Finding the next timer in each loop iteration: O(1)
3820
3821By virtue of using a binary or 4-heap, the next timer is always found at a
3822fixed position in the storage array.
2543 3823
2544=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd) 3824=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
2545 3825
2546A change means an I/O watcher gets started or stopped, which requires 3826A change means an I/O watcher gets started or stopped, which requires
2547libev to recalculate its status (and possibly tell the kernel). 3827libev to recalculate its status (and possibly tell the kernel, depending
3828on backend and whether C<ev_io_set> was used).
2548 3829
2549=item Activating one watcher: O(1) 3830=item Activating one watcher (putting it into the pending state): O(1)
2550 3831
2551=item Priority handling: O(number_of_priorities) 3832=item Priority handling: O(number_of_priorities)
2552 3833
2553Priorities are implemented by allocating some space for each 3834Priorities are implemented by allocating some space for each
2554priority. When doing priority-based operations, libev usually has to 3835priority. When doing priority-based operations, libev usually has to
2555linearly search all the priorities. 3836linearly search all the priorities, but starting/stopping and activating
3837watchers becomes O(1) with respect to priority handling.
3838
3839=item Sending an ev_async: O(1)
3840
3841=item Processing ev_async_send: O(number_of_async_watchers)
3842
3843=item Processing signals: O(max_signal_number)
3844
3845Sending involves a system call I<iff> there were no other C<ev_async_send>
3846calls in the current loop iteration. Checking for async and signal events
3847involves iterating over all running async watchers or all signal numbers.
2556 3848
2557=back 3849=back
2558 3850
2559 3851
2560=head1 AUTHOR 3852=head1 AUTHOR

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines