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

Comparing libev/ev.html (file contents):
Revision 1.28 by root, Sun Nov 18 03:43:24 2007 UTC vs.
Revision 1.34 by root, Fri Nov 23 15:26:08 2007 UTC

… …
4<head> 4<head>
5 <title>libev</title> 5 <title>libev</title>
6 <meta name="description" content="Pod documentation for libev" /> 6 <meta name="description" content="Pod documentation for libev" />
7 <meta name="inputfile" content="&lt;standard input&gt;" /> 7 <meta name="inputfile" content="&lt;standard input&gt;" />
8 <meta name="outputfile" content="&lt;standard output&gt;" /> 8 <meta name="outputfile" content="&lt;standard output&gt;" />
9 <meta name="created" content="Sun Nov 18 04:43:20 2007" /> 9 <meta name="created" content="Fri Nov 23 16:26:06 2007" />
10 <meta name="generator" content="Pod::Xhtml 1.57" /> 10 <meta name="generator" content="Pod::Xhtml 1.57" />
11<link rel="stylesheet" href="http://res.tst.eu/pod.css"/></head> 11<link rel="stylesheet" href="http://res.tst.eu/pod.css"/></head>
12<body> 12<body>
13<div class="pod"> 13<div class="pod">
14<!-- INDEX START --> 14<!-- INDEX START -->
… …
123 <p>Usually, it's a good idea to terminate if the major versions mismatch, 123 <p>Usually, it's a good idea to terminate if the major versions mismatch,
124as this indicates an incompatible change. Minor versions are usually 124as this indicates an incompatible change. Minor versions are usually
125compatible to older versions, so a larger minor version alone is usually 125compatible to older versions, so a larger minor version alone is usually
126not a problem.</p> 126not a problem.</p>
127 </dd> 127 </dd>
128 <dt>unsigned int ev_supported_backends ()</dt>
129 <dd>
130 <p>Return the set of all backends (i.e. their corresponding <code>EV_BACKEND_*</code>
131value) compiled into this binary of libev (independent of their
132availability on the system you are running on). See <code>ev_default_loop</code> for
133a description of the set values.</p>
134 </dd>
135 <dt>unsigned int ev_recommended_backends ()</dt>
136 <dd>
137 <p>Return the set of all backends compiled into this binary of libev and also
138recommended for this platform. This set is often smaller than the one
139returned by <code>ev_supported_backends</code>, as for example kqueue is broken on
140most BSDs and will not be autodetected unless you explicitly request it
141(assuming you know what you are doing). This is the set of backends that
142libev will probe for if you specify no backends explicitly.</p>
143 </dd>
128 <dt>ev_set_allocator (void *(*cb)(void *ptr, long size))</dt> 144 <dt>ev_set_allocator (void *(*cb)(void *ptr, long size))</dt>
129 <dd> 145 <dd>
130 <p>Sets the allocation function to use (the prototype is similar to the 146 <p>Sets the allocation function to use (the prototype is similar to the
131realloc C function, the semantics are identical). It is used to allocate 147realloc C function, the semantics are identical). It is used to allocate
132and free memory (no surprises here). If it returns zero when memory 148and free memory (no surprises here). If it returns zero when memory
… …
164 <dt>struct ev_loop *ev_default_loop (unsigned int flags)</dt> 180 <dt>struct ev_loop *ev_default_loop (unsigned int flags)</dt>
165 <dd> 181 <dd>
166 <p>This will initialise the default event loop if it hasn't been initialised 182 <p>This will initialise the default event loop if it hasn't been initialised
167yet and return it. If the default loop could not be initialised, returns 183yet and return it. If the default loop could not be initialised, returns
168false. If it already was initialised it simply returns it (and ignores the 184false. If it already was initialised it simply returns it (and ignores the
169flags).</p> 185flags. If that is troubling you, check <code>ev_backend ()</code> afterwards).</p>
170 <p>If you don't know what event loop to use, use the one returned from this 186 <p>If you don't know what event loop to use, use the one returned from this
171function.</p> 187function.</p>
172 <p>The flags argument can be used to specify special behaviour or specific 188 <p>The flags argument can be used to specify special behaviour or specific
173backends to use, and is usually specified as 0 (or EVFLAG_AUTO).</p> 189backends to use, and is usually specified as <code>0</code> (or <code>EVFLAG_AUTO</code>).</p>
174 <p>It supports the following flags:</p> 190 <p>The following flags are supported:</p>
175 <p> 191 <p>
176 <dl> 192 <dl>
177 <dt><code>EVFLAG_AUTO</code></dt> 193 <dt><code>EVFLAG_AUTO</code></dt>
178 <dd> 194 <dd>
179 <p>The default flags value. Use this if you have no clue (it's the right 195 <p>The default flags value. Use this if you have no clue (it's the right
… …
186<code>LIBEV_FLAGS</code>. Otherwise (the default), this environment variable will 202<code>LIBEV_FLAGS</code>. Otherwise (the default), this environment variable will
187override the flags completely if it is found in the environment. This is 203override the flags completely if it is found in the environment. This is
188useful to try out specific backends to test their performance, or to work 204useful to try out specific backends to test their performance, or to work
189around bugs.</p> 205around bugs.</p>
190 </dd> 206 </dd>
191 <dt><code>EVMETHOD_SELECT</code> (portable select backend)</dt> 207 <dt><code>EVBACKEND_SELECT</code> (value 1, portable select backend)</dt>
192 <dt><code>EVMETHOD_POLL</code> (poll backend, available everywhere except on windows)</dt> 208 <dd>
193 <dt><code>EVMETHOD_EPOLL</code> (linux only)</dt> 209 <p>This is your standard select(2) backend. Not <i>completely</i> standard, as
194 <dt><code>EVMETHOD_KQUEUE</code> (some bsds only)</dt> 210libev tries to roll its own fd_set with no limits on the number of fds,
195 <dt><code>EVMETHOD_DEVPOLL</code> (solaris 8 only)</dt> 211but if that fails, expect a fairly low limit on the number of fds when
196 <dt><code>EVMETHOD_PORT</code> (solaris 10 only)</dt> 212using this backend. It doesn't scale too well (O(highest_fd)), but its usually
213the fastest backend for a low number of fds.</p>
197 <dd> 214 </dd>
198 <p>If one or more of these are ored into the flags value, then only these 215 <dt><code>EVBACKEND_POLL</code> (value 2, poll backend, available everywhere except on windows)</dt>
199backends will be tried (in the reverse order as given here). If one are 216 <dd>
200specified, any backend will do.</p> 217 <p>And this is your standard poll(2) backend. It's more complicated than
218select, but handles sparse fds better and has no artificial limit on the
219number of fds you can use (except it will slow down considerably with a
220lot of inactive fds). It scales similarly to select, i.e. O(total_fds).</p>
221 </dd>
222 <dt><code>EVBACKEND_EPOLL</code> (value 4, Linux)</dt>
223 <dd>
224 <p>For few fds, this backend is a bit little slower than poll and select,
225but it scales phenomenally better. While poll and select usually scale like
226O(total_fds) where n is the total number of fds (or the highest fd), epoll scales
227either O(1) or O(active_fds).</p>
228 <p>While stopping and starting an I/O watcher in the same iteration will
229result in some caching, there is still a syscall per such incident
230(because the fd could point to a different file description now), so its
231best to avoid that. Also, dup()ed file descriptors might not work very
232well if you register events for both fds.</p>
233 <p>Please note that epoll sometimes generates spurious notifications, so you
234need to use non-blocking I/O or other means to avoid blocking when no data
235(or space) is available.</p>
236 </dd>
237 <dt><code>EVBACKEND_KQUEUE</code> (value 8, most BSD clones)</dt>
238 <dd>
239 <p>Kqueue deserves special mention, as at the time of this writing, it
240was broken on all BSDs except NetBSD (usually it doesn't work with
241anything but sockets and pipes, except on Darwin, where of course its
242completely useless). For this reason its not being &quot;autodetected&quot;
243unless you explicitly specify it explicitly in the flags (i.e. using
244<code>EVBACKEND_KQUEUE</code>).</p>
245 <p>It scales in the same way as the epoll backend, but the interface to the
246kernel is more efficient (which says nothing about its actual speed, of
247course). While starting and stopping an I/O watcher does not cause an
248extra syscall as with epoll, it still adds up to four event changes per
249incident, so its best to avoid that.</p>
250 </dd>
251 <dt><code>EVBACKEND_DEVPOLL</code> (value 16, Solaris 8)</dt>
252 <dd>
253 <p>This is not implemented yet (and might never be).</p>
254 </dd>
255 <dt><code>EVBACKEND_PORT</code> (value 32, Solaris 10)</dt>
256 <dd>
257 <p>This uses the Solaris 10 port mechanism. As with everything on Solaris,
258it's really slow, but it still scales very well (O(active_fds)).</p>
259 <p>Please note that solaris ports can result in a lot of spurious
260notifications, so you need to use non-blocking I/O or other means to avoid
261blocking when no data (or space) is available.</p>
262 </dd>
263 <dt><code>EVBACKEND_ALL</code></dt>
264 <dd>
265 <p>Try all backends (even potentially broken ones that wouldn't be tried
266with <code>EVFLAG_AUTO</code>). Since this is a mask, you can do stuff such as
267<code>EVBACKEND_ALL &amp; ~EVBACKEND_KQUEUE</code>.</p>
201 </dd> 268 </dd>
202 </dl> 269 </dl>
203 </p> 270 </p>
271 <p>If one or more of these are ored into the flags value, then only these
272backends will be tried (in the reverse order as given here). If none are
273specified, most compiled-in backend will be tried, usually in reverse
274order of their flag values :)</p>
275 <p>The most typical usage is like this:</p>
276<pre> if (!ev_default_loop (0))
277 fatal (&quot;could not initialise libev, bad $LIBEV_FLAGS in environment?&quot;);
278
279</pre>
280 <p>Restrict libev to the select and poll backends, and do not allow
281environment settings to be taken into account:</p>
282<pre> ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
283
284</pre>
285 <p>Use whatever libev has to offer, but make sure that kqueue is used if
286available (warning, breaks stuff, best use only with your own private
287event loop and only if you know the OS supports your types of fds):</p>
288<pre> ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
289
290</pre>
204 </dd> 291 </dd>
205 <dt>struct ev_loop *ev_loop_new (unsigned int flags)</dt> 292 <dt>struct ev_loop *ev_loop_new (unsigned int flags)</dt>
206 <dd> 293 <dd>
207 <p>Similar to <code>ev_default_loop</code>, but always creates a new event loop that is 294 <p>Similar to <code>ev_default_loop</code>, but always creates a new event loop that is
208always distinct from the default loop. Unlike the default loop, it cannot 295always distinct from the default loop. Unlike the default loop, it cannot
… …
224 <dd> 311 <dd>
225 <p>This function reinitialises the kernel state for backends that have 312 <p>This function reinitialises the kernel state for backends that have
226one. Despite the name, you can call it anytime, but it makes most sense 313one. Despite the name, you can call it anytime, but it makes most sense
227after forking, in either the parent or child process (or both, but that 314after forking, in either the parent or child process (or both, but that
228again makes little sense).</p> 315again makes little sense).</p>
229 <p>You <i>must</i> call this function after forking if and only if you want to 316 <p>You <i>must</i> call this function in the child process after forking if and
230use the event library in both processes. If you just fork+exec, you don't 317only if you want to use the event library in both processes. If you just
231have to call it.</p> 318fork+exec, you don't have to call it.</p>
232 <p>The function itself is quite fast and it's usually not a problem to call 319 <p>The function itself is quite fast and it's usually not a problem to call
233it just in case after a fork. To make this easy, the function will fit in 320it just in case after a fork. To make this easy, the function will fit in
234quite nicely into a call to <code>pthread_atfork</code>:</p> 321quite nicely into a call to <code>pthread_atfork</code>:</p>
235<pre> pthread_atfork (0, 0, ev_default_fork); 322<pre> pthread_atfork (0, 0, ev_default_fork);
236 323
237</pre> 324</pre>
325 <p>At the moment, <code>EVBACKEND_SELECT</code> and <code>EVBACKEND_POLL</code> are safe to use
326without calling this function, so if you force one of those backends you
327do not need to care.</p>
238 </dd> 328 </dd>
239 <dt>ev_loop_fork (loop)</dt> 329 <dt>ev_loop_fork (loop)</dt>
240 <dd> 330 <dd>
241 <p>Like <code>ev_default_fork</code>, but acts on an event loop created by 331 <p>Like <code>ev_default_fork</code>, but acts on an event loop created by
242<code>ev_loop_new</code>. Yes, you have to call this on every allocated event loop 332<code>ev_loop_new</code>. Yes, you have to call this on every allocated event loop
243after fork, and how you do this is entirely your own problem.</p> 333after fork, and how you do this is entirely your own problem.</p>
244 </dd> 334 </dd>
245 <dt>unsigned int ev_method (loop)</dt> 335 <dt>unsigned int ev_backend (loop)</dt>
246 <dd> 336 <dd>
247 <p>Returns one of the <code>EVMETHOD_*</code> flags indicating the event backend in 337 <p>Returns one of the <code>EVBACKEND_*</code> flags indicating the event backend in
248use.</p> 338use.</p>
249 </dd> 339 </dd>
250 <dt>ev_tstamp ev_now (loop)</dt> 340 <dt>ev_tstamp ev_now (loop)</dt>
251 <dd> 341 <dd>
252 <p>Returns the current &quot;event loop time&quot;, which is the time the event loop 342 <p>Returns the current &quot;event loop time&quot;, which is the time the event loop
… …
258 <dt>ev_loop (loop, int flags)</dt> 348 <dt>ev_loop (loop, int flags)</dt>
259 <dd> 349 <dd>
260 <p>Finally, this is it, the event handler. This function usually is called 350 <p>Finally, this is it, the event handler. This function usually is called
261after you initialised all your watchers and you want to start handling 351after you initialised all your watchers and you want to start handling
262events.</p> 352events.</p>
263 <p>If the flags argument is specified as 0, it will not return until either 353 <p>If the flags argument is specified as <code>0</code>, it will not return until
264no event watchers are active anymore or <code>ev_unloop</code> was called.</p> 354either no event watchers are active anymore or <code>ev_unloop</code> was called.</p>
265 <p>A flags value of <code>EVLOOP_NONBLOCK</code> will look for new events, will handle 355 <p>A flags value of <code>EVLOOP_NONBLOCK</code> will look for new events, will handle
266those events and any outstanding ones, but will not block your process in 356those events and any outstanding ones, but will not block your process in
267case there are no events and will return after one iteration of the loop.</p> 357case there are no events and will return after one iteration of the loop.</p>
268 <p>A flags value of <code>EVLOOP_ONESHOT</code> will look for new events (waiting if 358 <p>A flags value of <code>EVLOOP_ONESHOT</code> will look for new events (waiting if
269neccessary) and will handle those and any outstanding ones. It will block 359neccessary) and will handle those and any outstanding ones. It will block
270your process until at least one new event arrives, and will return after 360your process until at least one new event arrives, and will return after
271one iteration of the loop.</p> 361one iteration of the loop. This is useful if you are waiting for some
272 <p>This flags value could be used to implement alternative looping 362external event in conjunction with something not expressible using other
273constructs, but the <code>prepare</code> and <code>check</code> watchers provide a better and 363libev watchers. However, a pair of <code>ev_prepare</code>/<code>ev_check</code> watchers is
274more generic mechanism.</p> 364usually a better approach for this kind of thing.</p>
275 <p>Here are the gory details of what ev_loop does:</p> 365 <p>Here are the gory details of what <code>ev_loop</code> does:</p>
276<pre> 1. If there are no active watchers (reference count is zero), return. 366<pre> * If there are no active watchers (reference count is zero), return.
277 2. Queue and immediately call all prepare watchers. 367 - Queue prepare watchers and then call all outstanding watchers.
278 3. If we have been forked, recreate the kernel state. 368 - If we have been forked, recreate the kernel state.
279 4. Update the kernel state with all outstanding changes. 369 - Update the kernel state with all outstanding changes.
280 5. Update the &quot;event loop time&quot;. 370 - Update the &quot;event loop time&quot;.
281 6. Calculate for how long to block. 371 - Calculate for how long to block.
282 7. Block the process, waiting for events. 372 - Block the process, waiting for any events.
373 - Queue all outstanding I/O (fd) events.
283 8. Update the &quot;event loop time&quot; and do time jump handling. 374 - Update the &quot;event loop time&quot; and do time jump handling.
284 9. Queue all outstanding timers. 375 - Queue all outstanding timers.
285 10. Queue all outstanding periodics. 376 - Queue all outstanding periodics.
286 11. If no events are pending now, queue all idle watchers. 377 - If no events are pending now, queue all idle watchers.
287 12. Queue all check watchers. 378 - Queue all check watchers.
288 13. Call all queued watchers in reverse order (i.e. check watchers first). 379 - Call all queued watchers in reverse order (i.e. check watchers first).
380 Signals and child watchers are implemented as I/O watchers, and will
381 be handled here by queueing them when their watcher gets executed.
289 14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK 382 - If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
290 was used, return, otherwise continue with step #1. 383 were used, return, otherwise continue with step *.
291 384
292</pre> 385</pre>
293 </dd> 386 </dd>
294 <dt>ev_unloop (loop, how)</dt> 387 <dt>ev_unloop (loop, how)</dt>
295 <dd> 388 <dd>
… …
350with a watcher-specific start function (<code>ev_&lt;type&gt;_start (loop, watcher 443with a watcher-specific start function (<code>ev_&lt;type&gt;_start (loop, watcher
351*)</code>), and you can stop watching for events at any time by calling the 444*)</code>), and you can stop watching for events at any time by calling the
352corresponding stop function (<code>ev_&lt;type&gt;_stop (loop, watcher *)</code>.</p> 445corresponding stop function (<code>ev_&lt;type&gt;_stop (loop, watcher *)</code>.</p>
353<p>As long as your watcher is active (has been started but not stopped) you 446<p>As long as your watcher is active (has been started but not stopped) you
354must not touch the values stored in it. Most specifically you must never 447must not touch the values stored in it. Most specifically you must never
355reinitialise it or call its set method.</p> 448reinitialise it or call its set macro.</p>
356<p>You can check whether an event is active by calling the <code>ev_is_active 449<p>You can check whether an event is active by calling the <code>ev_is_active
357(watcher *)</code> macro. To see whether an event is outstanding (but the 450(watcher *)</code> macro. To see whether an event is outstanding (but the
358callback for it has not been called yet) you can use the <code>ev_is_pending 451callback for it has not been called yet) you can use the <code>ev_is_pending
359(watcher *)</code> macro.</p> 452(watcher *)</code> macro.</p>
360<p>Each and every callback receives the event loop pointer as first, the 453<p>Each and every callback receives the event loop pointer as first, the
… …
472(the linux epoll backend is a notable example) cannot handle dup'ed file 565(the linux epoll backend is a notable example) cannot handle dup'ed file
473descriptors correctly if you register interest in two or more fds pointing 566descriptors correctly if you register interest in two or more fds pointing
474to the same underlying file/socket etc. description (that is, they share 567to the same underlying file/socket etc. description (that is, they share
475the same underlying &quot;file open&quot;).</p> 568the same underlying &quot;file open&quot;).</p>
476<p>If you must do this, then force the use of a known-to-be-good backend 569<p>If you must do this, then force the use of a known-to-be-good backend
477(at the time of this writing, this includes only EVMETHOD_SELECT and 570(at the time of this writing, this includes only <code>EVBACKEND_SELECT</code> and
478EVMETHOD_POLL).</p> 571<code>EVBACKEND_POLL</code>).</p>
479<dl> 572<dl>
480 <dt>ev_io_init (ev_io *, callback, int fd, int events)</dt> 573 <dt>ev_io_init (ev_io *, callback, int fd, int events)</dt>
481 <dt>ev_io_set (ev_io *, int fd, int events)</dt> 574 <dt>ev_io_set (ev_io *, int fd, int events)</dt>
482 <dd> 575 <dd>
483 <p>Configures an <code>ev_io</code> watcher. The fd is the file descriptor to rceeive 576 <p>Configures an <code>ev_io</code> watcher. The fd is the file descriptor to rceeive
484events for and events is either <code>EV_READ</code>, <code>EV_WRITE</code> or <code>EV_READ | 577events for and events is either <code>EV_READ</code>, <code>EV_WRITE</code> or <code>EV_READ |
485EV_WRITE</code> to receive the given events.</p> 578EV_WRITE</code> to receive the given events.</p>
579 <p>Please note that most of the more scalable backend mechanisms (for example
580epoll and solaris ports) can result in spurious readyness notifications
581for file descriptors, so you practically need to use non-blocking I/O (and
582treat callback invocation as hint only), or retest separately with a safe
583interface before doing I/O (XLib can do this), or force the use of either
584<code>EVBACKEND_SELECT</code> or <code>EVBACKEND_POLL</code>, which don't suffer from this
585problem. Also note that it is quite easy to have your callback invoked
586when the readyness condition is no longer valid even when employing
587typical ways of handling events, so its a good idea to use non-blocking
588I/O unconditionally.</p>
486 </dd> 589 </dd>
487</dl> 590</dl>
488 591
489</div> 592</div>
490<h2 id="code_ev_timer_code_relative_and_opti"><code>ev_timer</code> - relative and optionally recurring timeouts</h2> 593<h2 id="code_ev_timer_code_relative_and_opti"><code>ev_timer</code> - relative and optionally recurring timeouts</h2>

Diff Legend

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