ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev.3
(Generate patch)

Comparing libev/ev.3 (file contents):
Revision 1.29 by root, Tue Nov 27 20:38:07 2007 UTC vs.
Revision 1.54 by root, Fri Dec 21 04:38:45 2007 UTC

… …
126. ds Ae AE 126. ds Ae AE
127.\} 127.\}
128.rm #[ #] #H #V #F C 128.rm #[ #] #H #V #F C
129.\" ======================================================================== 129.\" ========================================================================
130.\" 130.\"
131.IX Title ""<STANDARD INPUT>" 1" 131.IX Title "EV 1"
132.TH "<STANDARD INPUT>" 1 "2007-11-27" "perl v5.8.8" "User Contributed Perl Documentation" 132.TH EV 1 "2007-12-21" "perl v5.8.8" "User Contributed Perl Documentation"
133.SH "NAME" 133.SH "NAME"
134libev \- a high performance full\-featured event loop written in C 134libev \- a high performance full\-featured event loop written in C
135.SH "SYNOPSIS" 135.SH "SYNOPSIS"
136.IX Header "SYNOPSIS" 136.IX Header "SYNOPSIS"
137.Vb 1 137.Vb 1
… …
196\& return 0; 196\& return 0;
197\& } 197\& }
198.Ve 198.Ve
199.SH "DESCRIPTION" 199.SH "DESCRIPTION"
200.IX Header "DESCRIPTION" 200.IX Header "DESCRIPTION"
201The newest version of this document is also available as a html-formatted
202web page you might find easier to navigate when reading it for the first
203time: <http://cvs.schmorp.de/libev/ev.html>.
204.PP
201Libev is an event loop: you register interest in certain events (such as a 205Libev is an event loop: you register interest in certain events (such as a
202file descriptor being readable or a timeout occuring), and it will manage 206file descriptor being readable or a timeout occurring), and it will manage
203these event sources and provide your program with events. 207these event sources and provide your program with events.
204.PP 208.PP
205To do this, it must take more or less complete control over your process 209To do this, it must take more or less complete control over your process
206(or thread) by executing the \fIevent loop\fR handler, and will then 210(or thread) by executing the \fIevent loop\fR handler, and will then
207communicate events via a callback mechanism. 211communicate events via a callback mechanism.
… …
210watchers\fR, which are relatively small C structures you initialise with the 214watchers\fR, which are relatively small C structures you initialise with the
211details of the event, and then hand it over to libev by \fIstarting\fR the 215details of the event, and then hand it over to libev by \fIstarting\fR the
212watcher. 216watcher.
213.SH "FEATURES" 217.SH "FEATURES"
214.IX Header "FEATURES" 218.IX Header "FEATURES"
215Libev supports \f(CW\*(C`select\*(C'\fR, \f(CW\*(C`poll\*(C'\fR, the linux-specific \f(CW\*(C`epoll\*(C'\fR, the 219Libev supports \f(CW\*(C`select\*(C'\fR, \f(CW\*(C`poll\*(C'\fR, the Linux-specific \f(CW\*(C`epoll\*(C'\fR, the
216bsd-specific \f(CW\*(C`kqueue\*(C'\fR and the solaris-specific event port mechanisms 220BSD-specific \f(CW\*(C`kqueue\*(C'\fR and the Solaris-specific event port mechanisms
217for file descriptor events (\f(CW\*(C`ev_io\*(C'\fR), relative timers (\f(CW\*(C`ev_timer\*(C'\fR), 221for file descriptor events (\f(CW\*(C`ev_io\*(C'\fR), the Linux \f(CW\*(C`inotify\*(C'\fR interface
222(for \f(CW\*(C`ev_stat\*(C'\fR), relative timers (\f(CW\*(C`ev_timer\*(C'\fR), absolute timers
218absolute timers with customised rescheduling (\f(CW\*(C`ev_periodic\*(C'\fR), synchronous 223with customised rescheduling (\f(CW\*(C`ev_periodic\*(C'\fR), synchronous signals
219signals (\f(CW\*(C`ev_signal\*(C'\fR), process status change events (\f(CW\*(C`ev_child\*(C'\fR), and 224(\f(CW\*(C`ev_signal\*(C'\fR), process status change events (\f(CW\*(C`ev_child\*(C'\fR), and event
220event watchers dealing with the event loop mechanism itself (\f(CW\*(C`ev_idle\*(C'\fR, 225watchers dealing with the event loop mechanism itself (\f(CW\*(C`ev_idle\*(C'\fR,
221\&\f(CW\*(C`ev_embed\*(C'\fR, \f(CW\*(C`ev_prepare\*(C'\fR and \f(CW\*(C`ev_check\*(C'\fR watchers) as well as 226\&\f(CW\*(C`ev_embed\*(C'\fR, \f(CW\*(C`ev_prepare\*(C'\fR and \f(CW\*(C`ev_check\*(C'\fR watchers) as well as
222file watchers (\f(CW\*(C`ev_stat\*(C'\fR) and even limited support for fork events 227file watchers (\f(CW\*(C`ev_stat\*(C'\fR) and even limited support for fork events
223(\f(CW\*(C`ev_fork\*(C'\fR). 228(\f(CW\*(C`ev_fork\*(C'\fR).
224.PP 229.PP
225It also is quite fast (see this 230It also is quite fast (see this
… …
238Libev represents time as a single floating point number, representing the 243Libev represents time as a single floating point number, representing the
239(fractional) number of seconds since the (\s-1POSIX\s0) epoch (somewhere near 244(fractional) number of seconds since the (\s-1POSIX\s0) epoch (somewhere near
240the beginning of 1970, details are complicated, don't ask). This type is 245the beginning of 1970, details are complicated, don't ask). This type is
241called \f(CW\*(C`ev_tstamp\*(C'\fR, which is what you should use too. It usually aliases 246called \f(CW\*(C`ev_tstamp\*(C'\fR, which is what you should use too. It usually aliases
242to the \f(CW\*(C`double\*(C'\fR type in C, and when you need to do any calculations on 247to the \f(CW\*(C`double\*(C'\fR type in C, and when you need to do any calculations on
243it, you should treat it as such. 248it, you should treat it as some floatingpoint value. Unlike the name
249component \f(CW\*(C`stamp\*(C'\fR might indicate, it is also used for time differences
250throughout libev.
244.SH "GLOBAL FUNCTIONS" 251.SH "GLOBAL FUNCTIONS"
245.IX Header "GLOBAL FUNCTIONS" 252.IX Header "GLOBAL FUNCTIONS"
246These functions can be called anytime, even before initialising the 253These functions can be called anytime, even before initialising the
247library in any way. 254library in any way.
248.IP "ev_tstamp ev_time ()" 4 255.IP "ev_tstamp ev_time ()" 4
… …
254.IX Item "int ev_version_major ()" 261.IX Item "int ev_version_major ()"
255.PD 0 262.PD 0
256.IP "int ev_version_minor ()" 4 263.IP "int ev_version_minor ()" 4
257.IX Item "int ev_version_minor ()" 264.IX Item "int ev_version_minor ()"
258.PD 265.PD
259You can find out the major and minor version numbers of the library 266You can find out the major and minor \s-1ABI\s0 version numbers of the library
260you linked against by calling the functions \f(CW\*(C`ev_version_major\*(C'\fR and 267you linked against by calling the functions \f(CW\*(C`ev_version_major\*(C'\fR and
261\&\f(CW\*(C`ev_version_minor\*(C'\fR. If you want, you can compare against the global 268\&\f(CW\*(C`ev_version_minor\*(C'\fR. If you want, you can compare against the global
262symbols \f(CW\*(C`EV_VERSION_MAJOR\*(C'\fR and \f(CW\*(C`EV_VERSION_MINOR\*(C'\fR, which specify the 269symbols \f(CW\*(C`EV_VERSION_MAJOR\*(C'\fR and \f(CW\*(C`EV_VERSION_MINOR\*(C'\fR, which specify the
263version of the library your program was compiled against. 270version of the library your program was compiled against.
264.Sp 271.Sp
272These version numbers refer to the \s-1ABI\s0 version of the library, not the
273release version.
274.Sp
265Usually, it's a good idea to terminate if the major versions mismatch, 275Usually, it's a good idea to terminate if the major versions mismatch,
266as this indicates an incompatible change. Minor versions are usually 276as this indicates an incompatible change. Minor versions are usually
267compatible to older versions, so a larger minor version alone is usually 277compatible to older versions, so a larger minor version alone is usually
268not a problem. 278not a problem.
269.Sp 279.Sp
270Example: Make sure we haven't accidentally been linked against the wrong 280Example: Make sure we haven't accidentally been linked against the wrong
271version. 281version.
… …
304might be supported on the current system, you would need to look at 314might be supported on the current system, you would need to look at
305\&\f(CW\*(C`ev_embeddable_backends () & ev_supported_backends ()\*(C'\fR, likewise for 315\&\f(CW\*(C`ev_embeddable_backends () & ev_supported_backends ()\*(C'\fR, likewise for
306recommended ones. 316recommended ones.
307.Sp 317.Sp
308See the description of \f(CW\*(C`ev_embed\*(C'\fR watchers for more info. 318See the description of \f(CW\*(C`ev_embed\*(C'\fR watchers for more info.
309.IP "ev_set_allocator (void *(*cb)(void *ptr, size_t size))" 4 319.IP "ev_set_allocator (void *(*cb)(void *ptr, long size))" 4
310.IX Item "ev_set_allocator (void *(*cb)(void *ptr, size_t size))" 320.IX Item "ev_set_allocator (void *(*cb)(void *ptr, long size))"
311Sets the allocation function to use (the prototype and semantics are 321Sets the allocation function to use (the prototype is similar \- the
312identical to the realloc C function). It is used to allocate and free 322semantics is identical \- to the realloc C function). It is used to
313memory (no surprises here). If it returns zero when memory needs to be 323allocate and free memory (no surprises here). If it returns zero when
314allocated, the library might abort or take some potentially destructive 324memory needs to be allocated, the library might abort or take some
315action. The default is your system realloc function. 325potentially destructive action. The default is your system realloc
326function.
316.Sp 327.Sp
317You could override this function in high-availability programs to, say, 328You could override this function in high-availability programs to, say,
318free some memory if it cannot allocate memory, to use a special allocator, 329free some memory if it cannot allocate memory, to use a special allocator,
319or even to sleep a while and retry until some memory is available. 330or even to sleep a while and retry until some memory is available.
320.Sp 331.Sp
… …
409or setgid) then libev will \fInot\fR look at the environment variable 420or setgid) then libev will \fInot\fR look at the environment variable
410\&\f(CW\*(C`LIBEV_FLAGS\*(C'\fR. Otherwise (the default), this environment variable will 421\&\f(CW\*(C`LIBEV_FLAGS\*(C'\fR. Otherwise (the default), this environment variable will
411override the flags completely if it is found in the environment. This is 422override the flags completely if it is found in the environment. This is
412useful to try out specific backends to test their performance, or to work 423useful to try out specific backends to test their performance, or to work
413around bugs. 424around bugs.
425.ie n .IP """EVFLAG_FORKCHECK""" 4
426.el .IP "\f(CWEVFLAG_FORKCHECK\fR" 4
427.IX Item "EVFLAG_FORKCHECK"
428Instead of calling \f(CW\*(C`ev_default_fork\*(C'\fR or \f(CW\*(C`ev_loop_fork\*(C'\fR manually after
429a fork, you can also make libev check for a fork in each iteration by
430enabling this flag.
431.Sp
432This works by calling \f(CW\*(C`getpid ()\*(C'\fR on every iteration of the loop,
433and thus this might slow down your event loop if you do a lot of loop
434iterations and little real work, but is usually not noticeable (on my
435Linux system for example, \f(CW\*(C`getpid\*(C'\fR is actually a simple 5\-insn sequence
436without a syscall and thus \fIvery\fR fast, but my Linux system also has
437\&\f(CW\*(C`pthread_atfork\*(C'\fR which is even faster).
438.Sp
439The big advantage of this flag is that you can forget about fork (and
440forget about forgetting to tell libev about forking) when you use this
441flag.
442.Sp
443This flag setting cannot be overriden or specified in the \f(CW\*(C`LIBEV_FLAGS\*(C'\fR
444environment variable.
414.ie n .IP """EVBACKEND_SELECT"" (value 1, portable select backend)" 4 445.ie n .IP """EVBACKEND_SELECT"" (value 1, portable select backend)" 4
415.el .IP "\f(CWEVBACKEND_SELECT\fR (value 1, portable select backend)" 4 446.el .IP "\f(CWEVBACKEND_SELECT\fR (value 1, portable select backend)" 4
416.IX Item "EVBACKEND_SELECT (value 1, portable select backend)" 447.IX Item "EVBACKEND_SELECT (value 1, portable select backend)"
417This is your standard \fIselect\fR\|(2) backend. Not \fIcompletely\fR standard, as 448This is your standard \fIselect\fR\|(2) backend. Not \fIcompletely\fR standard, as
418libev tries to roll its own fd_set with no limits on the number of fds, 449libev tries to roll its own fd_set with no limits on the number of fds,
… …
428lot of inactive fds). It scales similarly to select, i.e. O(total_fds). 459lot of inactive fds). It scales similarly to select, i.e. O(total_fds).
429.ie n .IP """EVBACKEND_EPOLL"" (value 4, Linux)" 4 460.ie n .IP """EVBACKEND_EPOLL"" (value 4, Linux)" 4
430.el .IP "\f(CWEVBACKEND_EPOLL\fR (value 4, Linux)" 4 461.el .IP "\f(CWEVBACKEND_EPOLL\fR (value 4, Linux)" 4
431.IX Item "EVBACKEND_EPOLL (value 4, Linux)" 462.IX Item "EVBACKEND_EPOLL (value 4, Linux)"
432For few fds, this backend is a bit little slower than poll and select, 463For few fds, this backend is a bit little slower than poll and select,
433but it scales phenomenally better. While poll and select usually scale like 464but it scales phenomenally better. While poll and select usually scale
434O(total_fds) where n is the total number of fds (or the highest fd), epoll scales 465like O(total_fds) where n is the total number of fds (or the highest fd),
435either O(1) or O(active_fds). 466epoll scales either O(1) or O(active_fds). The epoll design has a number
467of shortcomings, such as silently dropping events in some hard-to-detect
468cases and rewuiring a syscall per fd change, no fork support and bad
469support for dup:
436.Sp 470.Sp
437While stopping and starting an I/O watcher in the same iteration will 471While stopping, setting and starting an I/O watcher in the same iteration
438result in some caching, there is still a syscall per such incident 472will result in some caching, there is still a syscall per such incident
439(because the fd could point to a different file description now), so its 473(because the fd could point to a different file description now), so its
440best to avoid that. Also, \fIdup()\fRed file descriptors might not work very 474best to avoid that. Also, \f(CW\*(C`dup ()\*(C'\fR'ed file descriptors might not work
441well if you register events for both fds. 475very well if you register events for both fds.
442.Sp 476.Sp
443Please note that epoll sometimes generates spurious notifications, so you 477Please note that epoll sometimes generates spurious notifications, so you
444need to use non-blocking I/O or other means to avoid blocking when no data 478need to use non-blocking I/O or other means to avoid blocking when no data
445(or space) is available. 479(or space) is available.
446.ie n .IP """EVBACKEND_KQUEUE"" (value 8, most \s-1BSD\s0 clones)" 4 480.ie n .IP """EVBACKEND_KQUEUE"" (value 8, most \s-1BSD\s0 clones)" 4
447.el .IP "\f(CWEVBACKEND_KQUEUE\fR (value 8, most \s-1BSD\s0 clones)" 4 481.el .IP "\f(CWEVBACKEND_KQUEUE\fR (value 8, most \s-1BSD\s0 clones)" 4
448.IX Item "EVBACKEND_KQUEUE (value 8, most BSD clones)" 482.IX Item "EVBACKEND_KQUEUE (value 8, most BSD clones)"
449Kqueue deserves special mention, as at the time of this writing, it 483Kqueue deserves special mention, as at the time of this writing, it
450was broken on all BSDs except NetBSD (usually it doesn't work with 484was broken on \fIall\fR BSDs (usually it doesn't work with anything but
451anything but sockets and pipes, except on Darwin, where of course its 485sockets and pipes, except on Darwin, where of course it's completely
486useless. On NetBSD, it seems to work for all the \s-1FD\s0 types I tested, so it
452completely useless). For this reason its not being \*(L"autodetected\*(R" 487is used by default there). For this reason it's not being \*(L"autodetected\*(R"
453unless you explicitly specify it explicitly in the flags (i.e. using 488unless you explicitly specify it explicitly in the flags (i.e. using
454\&\f(CW\*(C`EVBACKEND_KQUEUE\*(C'\fR). 489\&\f(CW\*(C`EVBACKEND_KQUEUE\*(C'\fR) or libev was compiled on a known-to-be-good (\-enough)
490system like NetBSD.
455.Sp 491.Sp
456It scales in the same way as the epoll backend, but the interface to the 492It scales in the same way as the epoll backend, but the interface to the
457kernel is more efficient (which says nothing about its actual speed, of 493kernel is more efficient (which says nothing about its actual speed,
458course). While starting and stopping an I/O watcher does not cause an 494of course). While stopping, setting and starting an I/O watcher does
459extra syscall as with epoll, it still adds up to four event changes per 495never cause an extra syscall as with epoll, it still adds up to two event
460incident, so its best to avoid that. 496changes per incident, support for \f(CW\*(C`fork ()\*(C'\fR is very bad and it drops fds
497silently in similarly hard-to-detetc cases.
461.ie n .IP """EVBACKEND_DEVPOLL"" (value 16, Solaris 8)" 4 498.ie n .IP """EVBACKEND_DEVPOLL"" (value 16, Solaris 8)" 4
462.el .IP "\f(CWEVBACKEND_DEVPOLL\fR (value 16, Solaris 8)" 4 499.el .IP "\f(CWEVBACKEND_DEVPOLL\fR (value 16, Solaris 8)" 4
463.IX Item "EVBACKEND_DEVPOLL (value 16, Solaris 8)" 500.IX Item "EVBACKEND_DEVPOLL (value 16, Solaris 8)"
464This is not implemented yet (and might never be). 501This is not implemented yet (and might never be).
465.ie n .IP """EVBACKEND_PORT"" (value 32, Solaris 10)" 4 502.ie n .IP """EVBACKEND_PORT"" (value 32, Solaris 10)" 4
466.el .IP "\f(CWEVBACKEND_PORT\fR (value 32, Solaris 10)" 4 503.el .IP "\f(CWEVBACKEND_PORT\fR (value 32, Solaris 10)" 4
467.IX Item "EVBACKEND_PORT (value 32, Solaris 10)" 504.IX Item "EVBACKEND_PORT (value 32, Solaris 10)"
468This uses the Solaris 10 port mechanism. As with everything on Solaris, 505This uses the Solaris 10 event port mechanism. As with everything on Solaris,
469it's really slow, but it still scales very well (O(active_fds)). 506it's really slow, but it still scales very well (O(active_fds)).
470.Sp 507.Sp
471Please note that solaris ports can result in a lot of spurious 508Please note that solaris event ports can deliver a lot of spurious
472notifications, so you need to use non-blocking I/O or other means to avoid 509notifications, so you need to use non-blocking I/O or other means to avoid
473blocking when no data (or space) is available. 510blocking when no data (or space) is available.
474.ie n .IP """EVBACKEND_ALL""" 4 511.ie n .IP """EVBACKEND_ALL""" 4
475.el .IP "\f(CWEVBACKEND_ALL\fR" 4 512.el .IP "\f(CWEVBACKEND_ALL\fR" 4
476.IX Item "EVBACKEND_ALL" 513.IX Item "EVBACKEND_ALL"
… …
526Destroys the default loop again (frees all memory and kernel state 563Destroys the default loop again (frees all memory and kernel state
527etc.). None of the active event watchers will be stopped in the normal 564etc.). None of the active event watchers will be stopped in the normal
528sense, so e.g. \f(CW\*(C`ev_is_active\*(C'\fR might still return true. It is your 565sense, so e.g. \f(CW\*(C`ev_is_active\*(C'\fR might still return true. It is your
529responsibility to either stop all watchers cleanly yoursef \fIbefore\fR 566responsibility to either stop all watchers cleanly yoursef \fIbefore\fR
530calling this function, or cope with the fact afterwards (which is usually 567calling this function, or cope with the fact afterwards (which is usually
531the easiest thing, youc na just ignore the watchers and/or \f(CW\*(C`free ()\*(C'\fR them 568the easiest thing, you can just ignore the watchers and/or \f(CW\*(C`free ()\*(C'\fR them
532for example). 569for example).
570.Sp
571Note that certain global state, such as signal state, will not be freed by
572this function, and related watchers (such as signal and child watchers)
573would need to be stopped manually.
574.Sp
575In general it is not advisable to call this function except in the
576rare occasion where you really need to free e.g. the signal handling
577pipe fds. If you need dynamically allocated loops it is better to use
578\&\f(CW\*(C`ev_loop_new\*(C'\fR and \f(CW\*(C`ev_loop_destroy\*(C'\fR).
533.IP "ev_loop_destroy (loop)" 4 579.IP "ev_loop_destroy (loop)" 4
534.IX Item "ev_loop_destroy (loop)" 580.IX Item "ev_loop_destroy (loop)"
535Like \f(CW\*(C`ev_default_destroy\*(C'\fR, but destroys an event loop created by an 581Like \f(CW\*(C`ev_default_destroy\*(C'\fR, but destroys an event loop created by an
536earlier call to \f(CW\*(C`ev_loop_new\*(C'\fR. 582earlier call to \f(CW\*(C`ev_loop_new\*(C'\fR.
537.IP "ev_default_fork ()" 4 583.IP "ev_default_fork ()" 4
… …
559.IP "ev_loop_fork (loop)" 4 605.IP "ev_loop_fork (loop)" 4
560.IX Item "ev_loop_fork (loop)" 606.IX Item "ev_loop_fork (loop)"
561Like \f(CW\*(C`ev_default_fork\*(C'\fR, but acts on an event loop created by 607Like \f(CW\*(C`ev_default_fork\*(C'\fR, but acts on an event loop created by
562\&\f(CW\*(C`ev_loop_new\*(C'\fR. Yes, you have to call this on every allocated event loop 608\&\f(CW\*(C`ev_loop_new\*(C'\fR. Yes, you have to call this on every allocated event loop
563after fork, and how you do this is entirely your own problem. 609after fork, and how you do this is entirely your own problem.
610.IP "unsigned int ev_loop_count (loop)" 4
611.IX Item "unsigned int ev_loop_count (loop)"
612Returns the count of loop iterations for the loop, which is identical to
613the number of times libev did poll for new events. It starts at \f(CW0\fR and
614happily wraps around with enough iterations.
615.Sp
616This value can sometimes be useful as a generation counter of sorts (it
617\&\*(L"ticks\*(R" the number of loop iterations), as it roughly corresponds with
618\&\f(CW\*(C`ev_prepare\*(C'\fR and \f(CW\*(C`ev_check\*(C'\fR calls.
564.IP "unsigned int ev_backend (loop)" 4 619.IP "unsigned int ev_backend (loop)" 4
565.IX Item "unsigned int ev_backend (loop)" 620.IX Item "unsigned int ev_backend (loop)"
566Returns one of the \f(CW\*(C`EVBACKEND_*\*(C'\fR flags indicating the event backend in 621Returns one of the \f(CW\*(C`EVBACKEND_*\*(C'\fR flags indicating the event backend in
567use. 622use.
568.IP "ev_tstamp ev_now (loop)" 4 623.IP "ev_tstamp ev_now (loop)" 4
569.IX Item "ev_tstamp ev_now (loop)" 624.IX Item "ev_tstamp ev_now (loop)"
570Returns the current \*(L"event loop time\*(R", which is the time the event loop 625Returns the current \*(L"event loop time\*(R", which is the time the event loop
571received events and started processing them. This timestamp does not 626received events and started processing them. This timestamp does not
572change as long as callbacks are being processed, and this is also the base 627change as long as callbacks are being processed, and this is also the base
573time used for relative timers. You can treat it as the timestamp of the 628time used for relative timers. You can treat it as the timestamp of the
574event occuring (or more correctly, libev finding out about it). 629event occurring (or more correctly, libev finding out about it).
575.IP "ev_loop (loop, int flags)" 4 630.IP "ev_loop (loop, int flags)" 4
576.IX Item "ev_loop (loop, int flags)" 631.IX Item "ev_loop (loop, int flags)"
577Finally, this is it, the event handler. This function usually is called 632Finally, this is it, the event handler. This function usually is called
578after you initialised all your watchers and you want to start handling 633after you initialised all your watchers and you want to start handling
579events. 634events.
… …
599libev watchers. However, a pair of \f(CW\*(C`ev_prepare\*(C'\fR/\f(CW\*(C`ev_check\*(C'\fR watchers is 654libev watchers. However, a pair of \f(CW\*(C`ev_prepare\*(C'\fR/\f(CW\*(C`ev_check\*(C'\fR watchers is
600usually a better approach for this kind of thing. 655usually a better approach for this kind of thing.
601.Sp 656.Sp
602Here are the gory details of what \f(CW\*(C`ev_loop\*(C'\fR does: 657Here are the gory details of what \f(CW\*(C`ev_loop\*(C'\fR does:
603.Sp 658.Sp
604.Vb 18 659.Vb 19
660\& - Before the first iteration, call any pending watchers.
605\& * If there are no active watchers (reference count is zero), return. 661\& * If there are no active watchers (reference count is zero), return.
606\& - Queue prepare watchers and then call all outstanding watchers. 662\& - Queue all prepare watchers and then call all outstanding watchers.
607\& - If we have been forked, recreate the kernel state. 663\& - If we have been forked, recreate the kernel state.
608\& - Update the kernel state with all outstanding changes. 664\& - Update the kernel state with all outstanding changes.
609\& - Update the "event loop time". 665\& - Update the "event loop time".
610\& - Calculate for how long to block. 666\& - Calculate for how long to block.
611\& - Block the process, waiting for any events. 667\& - Block the process, waiting for any events.
… …
854.IP "bool ev_is_pending (ev_TYPE *watcher)" 4 910.IP "bool ev_is_pending (ev_TYPE *watcher)" 4
855.IX Item "bool ev_is_pending (ev_TYPE *watcher)" 911.IX Item "bool ev_is_pending (ev_TYPE *watcher)"
856Returns a true value iff the watcher is pending, (i.e. it has outstanding 912Returns a true value iff the watcher is pending, (i.e. it has outstanding
857events but its callback has not yet been invoked). As long as a watcher 913events but its callback has not yet been invoked). As long as a watcher
858is pending (but not active) you must not call an init function on it (but 914is pending (but not active) you must not call an init function on it (but
859\&\f(CW\*(C`ev_TYPE_set\*(C'\fR is safe) and you must make sure the watcher is available to 915\&\f(CW\*(C`ev_TYPE_set\*(C'\fR is safe), you must not change its priority, and you must
860libev (e.g. you cnanot \f(CW\*(C`free ()\*(C'\fR it). 916make sure the watcher is available to libev (e.g. you cannot \f(CW\*(C`free ()\*(C'\fR
917it).
861.IP "callback ev_cb (ev_TYPE *watcher)" 4 918.IP "callback ev_cb (ev_TYPE *watcher)" 4
862.IX Item "callback ev_cb (ev_TYPE *watcher)" 919.IX Item "callback ev_cb (ev_TYPE *watcher)"
863Returns the callback currently set on the watcher. 920Returns the callback currently set on the watcher.
864.IP "ev_cb_set (ev_TYPE *watcher, callback)" 4 921.IP "ev_cb_set (ev_TYPE *watcher, callback)" 4
865.IX Item "ev_cb_set (ev_TYPE *watcher, callback)" 922.IX Item "ev_cb_set (ev_TYPE *watcher, callback)"
866Change the callback. You can change the callback at virtually any time 923Change the callback. You can change the callback at virtually any time
867(modulo threads). 924(modulo threads).
925.IP "ev_set_priority (ev_TYPE *watcher, priority)" 4
926.IX Item "ev_set_priority (ev_TYPE *watcher, priority)"
927.PD 0
928.IP "int ev_priority (ev_TYPE *watcher)" 4
929.IX Item "int ev_priority (ev_TYPE *watcher)"
930.PD
931Set and query the priority of the watcher. The priority is a small
932integer between \f(CW\*(C`EV_MAXPRI\*(C'\fR (default: \f(CW2\fR) and \f(CW\*(C`EV_MINPRI\*(C'\fR
933(default: \f(CW\*(C`\-2\*(C'\fR). Pending watchers with higher priority will be invoked
934before watchers with lower priority, but priority will not keep watchers
935from being executed (except for \f(CW\*(C`ev_idle\*(C'\fR watchers).
936.Sp
937This means that priorities are \fIonly\fR used for ordering callback
938invocation after new events have been received. This is useful, for
939example, to reduce latency after idling, or more often, to bind two
940watchers on the same event and make sure one is called first.
941.Sp
942If you need to suppress invocation when higher priority events are pending
943you need to look at \f(CW\*(C`ev_idle\*(C'\fR watchers, which provide this functionality.
944.Sp
945You \fImust not\fR change the priority of a watcher as long as it is active or
946pending.
947.Sp
948The default priority used by watchers when no priority has been set is
949always \f(CW0\fR, which is supposed to not be too high and not be too low :).
950.Sp
951Setting a priority outside the range of \f(CW\*(C`EV_MINPRI\*(C'\fR to \f(CW\*(C`EV_MAXPRI\*(C'\fR is
952fine, as long as you do not mind that the priority value you query might
953or might not have been adjusted to be within valid range.
954.IP "ev_invoke (loop, ev_TYPE *watcher, int revents)" 4
955.IX Item "ev_invoke (loop, ev_TYPE *watcher, int revents)"
956Invoke the \f(CW\*(C`watcher\*(C'\fR with the given \f(CW\*(C`loop\*(C'\fR and \f(CW\*(C`revents\*(C'\fR. Neither
957\&\f(CW\*(C`loop\*(C'\fR nor \f(CW\*(C`revents\*(C'\fR need to be valid as long as the watcher callback
958can deal with that fact.
959.IP "int ev_clear_pending (loop, ev_TYPE *watcher)" 4
960.IX Item "int ev_clear_pending (loop, ev_TYPE *watcher)"
961If the watcher is pending, this function returns clears its pending status
962and returns its \f(CW\*(C`revents\*(C'\fR bitset (as if its callback was invoked). If the
963watcher isn't pending it does nothing and returns \f(CW0\fR.
868.Sh "\s-1ASSOCIATING\s0 \s-1CUSTOM\s0 \s-1DATA\s0 \s-1WITH\s0 A \s-1WATCHER\s0" 964.Sh "\s-1ASSOCIATING\s0 \s-1CUSTOM\s0 \s-1DATA\s0 \s-1WITH\s0 A \s-1WATCHER\s0"
869.IX Subsection "ASSOCIATING CUSTOM DATA WITH A WATCHER" 965.IX Subsection "ASSOCIATING CUSTOM DATA WITH A WATCHER"
870Each watcher has, by default, a member \f(CW\*(C`void *data\*(C'\fR that you can change 966Each watcher has, by default, a member \f(CW\*(C`void *data\*(C'\fR that you can change
871and read at any time, libev will completely ignore it. This can be used 967and read at any time, libev will completely ignore it. This can be used
872to associate arbitrary data with your watcher. If you need more data and 968to associate arbitrary data with your watcher. If you need more data and
… …
983it is best to always use non-blocking I/O: An extra \f(CW\*(C`read\*(C'\fR(2) returning 1079it is best to always use non-blocking I/O: An extra \f(CW\*(C`read\*(C'\fR(2) returning
984\&\f(CW\*(C`EAGAIN\*(C'\fR is far preferable to a program hanging until some data arrives. 1080\&\f(CW\*(C`EAGAIN\*(C'\fR is far preferable to a program hanging until some data arrives.
985.PP 1081.PP
986If you cannot run the fd in non-blocking mode (for example you should not 1082If you cannot run the fd in non-blocking mode (for example you should not
987play around with an Xlib connection), then you have to seperately re-test 1083play around with an Xlib connection), then you have to seperately re-test
988wether a file descriptor is really ready with a known-to-be good interface 1084whether a file descriptor is really ready with a known-to-be good interface
989such as poll (fortunately in our Xlib example, Xlib already does this on 1085such as poll (fortunately in our Xlib example, Xlib already does this on
990its own, so its quite safe to use). 1086its own, so its quite safe to use).
1087.PP
1088\fIThe special problem of disappearing file descriptors\fR
1089.IX Subsection "The special problem of disappearing file descriptors"
1090.PP
1091Some backends (e.g. kqueue, epoll) need to be told about closing a file
1092descriptor (either by calling \f(CW\*(C`close\*(C'\fR explicitly or by any other means,
1093such as \f(CW\*(C`dup\*(C'\fR). The reason is that you register interest in some file
1094descriptor, but when it goes away, the operating system will silently drop
1095this interest. If another file descriptor with the same number then is
1096registered with libev, there is no efficient way to see that this is, in
1097fact, a different file descriptor.
1098.PP
1099To avoid having to explicitly tell libev about such cases, libev follows
1100the following policy: Each time \f(CW\*(C`ev_io_set\*(C'\fR is being called, libev
1101will assume that this is potentially a new file descriptor, otherwise
1102it is assumed that the file descriptor stays the same. That means that
1103you \fIhave\fR to call \f(CW\*(C`ev_io_set\*(C'\fR (or \f(CW\*(C`ev_io_init\*(C'\fR) when you change the
1104descriptor even if the file descriptor number itself did not change.
1105.PP
1106This is how one would do it normally anyway, the important point is that
1107the libev application should not optimise around libev but should leave
1108optimisations to libev.
1109.PP
1110\fIThs special problem of dup'ed file descriptors\fR
1111.IX Subsection "Ths special problem of dup'ed file descriptors"
1112.PP
1113Some backends (e.g. epoll), cannot register events for file descriptors,
1114but only events for the underlying file descriptions. That menas when you
1115have \f(CW\*(C`dup ()\*(C'\fR'ed file descriptors and register events for them, only one
1116file descriptor might actually receive events.
1117.PP
1118There is no workaorund possible except not registering events
1119for potentially \f(CW\*(C`dup ()\*(C'\fR'ed file descriptors or to resort to
1120\&\f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or \f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
1121.PP
1122\fIThe special problem of fork\fR
1123.IX Subsection "The special problem of fork"
1124.PP
1125Some backends (epoll, kqueue) do not support \f(CW\*(C`fork ()\*(C'\fR at all or exhibit
1126useless behaviour. Libev fully supports fork, but needs to be told about
1127it in the child.
1128.PP
1129To support fork in your programs, you either have to call
1130\&\f(CW\*(C`ev_default_fork ()\*(C'\fR or \f(CW\*(C`ev_loop_fork ()\*(C'\fR after a fork in the child,
1131enable \f(CW\*(C`EVFLAG_FORKCHECK\*(C'\fR, or resort to \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or
1132\&\f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
1133.PP
1134\fIWatcher-Specific Functions\fR
1135.IX Subsection "Watcher-Specific Functions"
991.IP "ev_io_init (ev_io *, callback, int fd, int events)" 4 1136.IP "ev_io_init (ev_io *, callback, int fd, int events)" 4
992.IX Item "ev_io_init (ev_io *, callback, int fd, int events)" 1137.IX Item "ev_io_init (ev_io *, callback, int fd, int events)"
993.PD 0 1138.PD 0
994.IP "ev_io_set (ev_io *, int fd, int events)" 4 1139.IP "ev_io_set (ev_io *, int fd, int events)" 4
995.IX Item "ev_io_set (ev_io *, int fd, int events)" 1140.IX Item "ev_io_set (ev_io *, int fd, int events)"
… …
1048.Ve 1193.Ve
1049.PP 1194.PP
1050The callback is guarenteed to be invoked only when its timeout has passed, 1195The callback is guarenteed to be invoked only when its timeout has passed,
1051but if multiple timers become ready during the same loop iteration then 1196but if multiple timers become ready during the same loop iteration then
1052order of execution is undefined. 1197order of execution is undefined.
1198.PP
1199\fIWatcher-Specific Functions and Data Members\fR
1200.IX Subsection "Watcher-Specific Functions and Data Members"
1053.IP "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)" 4 1201.IP "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)" 4
1054.IX Item "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)" 1202.IX Item "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)"
1055.PD 0 1203.PD 0
1056.IP "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)" 4 1204.IP "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)" 4
1057.IX Item "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)" 1205.IX Item "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)"
… …
1069.IP "ev_timer_again (loop)" 4 1217.IP "ev_timer_again (loop)" 4
1070.IX Item "ev_timer_again (loop)" 1218.IX Item "ev_timer_again (loop)"
1071This will act as if the timer timed out and restart it again if it is 1219This will act as if the timer timed out and restart it again if it is
1072repeating. The exact semantics are: 1220repeating. The exact semantics are:
1073.Sp 1221.Sp
1222If the timer is pending, its pending status is cleared.
1223.Sp
1074If the timer is started but nonrepeating, stop it. 1224If the timer is started but nonrepeating, stop it (as if it timed out).
1075.Sp 1225.Sp
1076If the timer is repeating, either start it if necessary (with the repeat 1226If the timer is repeating, either start it if necessary (with the
1077value), or reset the running timer to the repeat value. 1227\&\f(CW\*(C`repeat\*(C'\fR value), or reset the running timer to the \f(CW\*(C`repeat\*(C'\fR value.
1078.Sp 1228.Sp
1079This sounds a bit complicated, but here is a useful and typical 1229This sounds a bit complicated, but here is a useful and typical
1080example: Imagine you have a tcp connection and you want a so-called 1230example: Imagine you have a tcp connection and you want a so-called idle
1081idle timeout, that is, you want to be called when there have been, 1231timeout, that is, you want to be called when there have been, say, 60
1082say, 60 seconds of inactivity on the socket. The easiest way to do 1232seconds of inactivity on the socket. The easiest way to do this is to
1083this is to configure an \f(CW\*(C`ev_timer\*(C'\fR with \f(CW\*(C`after\*(C'\fR=\f(CW\*(C`repeat\*(C'\fR=\f(CW60\fR and calling 1233configure an \f(CW\*(C`ev_timer\*(C'\fR with a \f(CW\*(C`repeat\*(C'\fR value of \f(CW60\fR and then call
1084\&\f(CW\*(C`ev_timer_again\*(C'\fR each time you successfully read or write some data. If 1234\&\f(CW\*(C`ev_timer_again\*(C'\fR each time you successfully read or write some data. If
1085you go into an idle state where you do not expect data to travel on the 1235you go into an idle state where you do not expect data to travel on the
1086socket, you can stop the timer, and again will automatically restart it if 1236socket, you can \f(CW\*(C`ev_timer_stop\*(C'\fR the timer, and \f(CW\*(C`ev_timer_again\*(C'\fR will
1087need be. 1237automatically restart it if need be.
1088.Sp 1238.Sp
1089You can also ignore the \f(CW\*(C`after\*(C'\fR value and \f(CW\*(C`ev_timer_start\*(C'\fR altogether 1239That means you can ignore the \f(CW\*(C`after\*(C'\fR value and \f(CW\*(C`ev_timer_start\*(C'\fR
1090and only ever use the \f(CW\*(C`repeat\*(C'\fR value: 1240altogether and only ever use the \f(CW\*(C`repeat\*(C'\fR value and \f(CW\*(C`ev_timer_again\*(C'\fR:
1091.Sp 1241.Sp
1092.Vb 8 1242.Vb 8
1093\& ev_timer_init (timer, callback, 0., 5.); 1243\& ev_timer_init (timer, callback, 0., 5.);
1094\& ev_timer_again (loop, timer); 1244\& ev_timer_again (loop, timer);
1095\& ... 1245\& ...
… …
1098\& ... 1248\& ...
1099\& timer->again = 10.; 1249\& timer->again = 10.;
1100\& ev_timer_again (loop, timer); 1250\& ev_timer_again (loop, timer);
1101.Ve 1251.Ve
1102.Sp 1252.Sp
1103This is more efficient then stopping/starting the timer eahc time you want 1253This is more slightly efficient then stopping/starting the timer each time
1104to modify its timeout value. 1254you want to modify its timeout value.
1105.IP "ev_tstamp repeat [read\-write]" 4 1255.IP "ev_tstamp repeat [read\-write]" 4
1106.IX Item "ev_tstamp repeat [read-write]" 1256.IX Item "ev_tstamp repeat [read-write]"
1107The current \f(CW\*(C`repeat\*(C'\fR value. Will be used each time the watcher times out 1257The current \f(CW\*(C`repeat\*(C'\fR value. Will be used each time the watcher times out
1108or \f(CW\*(C`ev_timer_again\*(C'\fR is called and determines the next timeout (if any), 1258or \f(CW\*(C`ev_timer_again\*(C'\fR is called and determines the next timeout (if any),
1109which is also when any modifications are taken into account. 1259which is also when any modifications are taken into account.
… …
1157but on wallclock time (absolute time). You can tell a periodic watcher 1307but on wallclock time (absolute time). You can tell a periodic watcher
1158to trigger \*(L"at\*(R" some specific point in time. For example, if you tell a 1308to trigger \*(L"at\*(R" some specific point in time. For example, if you tell a
1159periodic watcher to trigger in 10 seconds (by specifiying e.g. \f(CW\*(C`ev_now () 1309periodic watcher to trigger in 10 seconds (by specifiying e.g. \f(CW\*(C`ev_now ()
1160+ 10.\*(C'\fR) and then reset your system clock to the last year, then it will 1310+ 10.\*(C'\fR) and then reset your system clock to the last year, then it will
1161take a year to trigger the event (unlike an \f(CW\*(C`ev_timer\*(C'\fR, which would trigger 1311take a year to trigger the event (unlike an \f(CW\*(C`ev_timer\*(C'\fR, which would trigger
1162roughly 10 seconds later and of course not if you reset your system time 1312roughly 10 seconds later).
1163again).
1164.PP 1313.PP
1165They can also be used to implement vastly more complex timers, such as 1314They can also be used to implement vastly more complex timers, such as
1166triggering an event on eahc midnight, local time. 1315triggering an event on each midnight, local time or other, complicated,
1316rules.
1167.PP 1317.PP
1168As with timers, the callback is guarenteed to be invoked only when the 1318As with timers, the callback is guarenteed to be invoked only when the
1169time (\f(CW\*(C`at\*(C'\fR) has been passed, but if multiple periodic timers become ready 1319time (\f(CW\*(C`at\*(C'\fR) has been passed, but if multiple periodic timers become ready
1170during the same loop iteration then order of execution is undefined. 1320during the same loop iteration then order of execution is undefined.
1321.PP
1322\fIWatcher-Specific Functions and Data Members\fR
1323.IX Subsection "Watcher-Specific Functions and Data Members"
1171.IP "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)" 4 1324.IP "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)" 4
1172.IX Item "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)" 1325.IX Item "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)"
1173.PD 0 1326.PD 0
1174.IP "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)" 4 1327.IP "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)" 4
1175.IX Item "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)" 1328.IX Item "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)"
1176.PD 1329.PD
1177Lots of arguments, lets sort it out... There are basically three modes of 1330Lots of arguments, lets sort it out... There are basically three modes of
1178operation, and we will explain them from simplest to complex: 1331operation, and we will explain them from simplest to complex:
1179.RS 4 1332.RS 4
1180.IP "* absolute timer (interval = reschedule_cb = 0)" 4 1333.IP "* absolute timer (at = time, interval = reschedule_cb = 0)" 4
1181.IX Item "absolute timer (interval = reschedule_cb = 0)" 1334.IX Item "absolute timer (at = time, interval = reschedule_cb = 0)"
1182In this configuration the watcher triggers an event at the wallclock time 1335In this configuration the watcher triggers an event at the wallclock time
1183\&\f(CW\*(C`at\*(C'\fR and doesn't repeat. It will not adjust when a time jump occurs, 1336\&\f(CW\*(C`at\*(C'\fR and doesn't repeat. It will not adjust when a time jump occurs,
1184that is, if it is to be run at January 1st 2011 then it will run when the 1337that is, if it is to be run at January 1st 2011 then it will run when the
1185system time reaches or surpasses this time. 1338system time reaches or surpasses this time.
1186.IP "* non-repeating interval timer (interval > 0, reschedule_cb = 0)" 4 1339.IP "* non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)" 4
1187.IX Item "non-repeating interval timer (interval > 0, reschedule_cb = 0)" 1340.IX Item "non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)"
1188In this mode the watcher will always be scheduled to time out at the next 1341In this mode the watcher will always be scheduled to time out at the next
1189\&\f(CW\*(C`at + N * interval\*(C'\fR time (for some integer N) and then repeat, regardless 1342\&\f(CW\*(C`at + N * interval\*(C'\fR time (for some integer N, which can also be negative)
1190of any time jumps. 1343and then repeat, regardless of any time jumps.
1191.Sp 1344.Sp
1192This can be used to create timers that do not drift with respect to system 1345This can be used to create timers that do not drift with respect to system
1193time: 1346time:
1194.Sp 1347.Sp
1195.Vb 1 1348.Vb 1
… …
1202by 3600. 1355by 3600.
1203.Sp 1356.Sp
1204Another way to think about it (for the mathematically inclined) is that 1357Another way to think about it (for the mathematically inclined) is that
1205\&\f(CW\*(C`ev_periodic\*(C'\fR will try to run the callback in this mode at the next possible 1358\&\f(CW\*(C`ev_periodic\*(C'\fR will try to run the callback in this mode at the next possible
1206time where \f(CW\*(C`time = at (mod interval)\*(C'\fR, regardless of any time jumps. 1359time where \f(CW\*(C`time = at (mod interval)\*(C'\fR, regardless of any time jumps.
1360.Sp
1361For numerical stability it is preferable that the \f(CW\*(C`at\*(C'\fR value is near
1362\&\f(CW\*(C`ev_now ()\*(C'\fR (the current time), but there is no range requirement for
1363this value.
1207.IP "* manual reschedule mode (reschedule_cb = callback)" 4 1364.IP "* manual reschedule mode (at and interval ignored, reschedule_cb = callback)" 4
1208.IX Item "manual reschedule mode (reschedule_cb = callback)" 1365.IX Item "manual reschedule mode (at and interval ignored, reschedule_cb = callback)"
1209In this mode the values for \f(CW\*(C`interval\*(C'\fR and \f(CW\*(C`at\*(C'\fR are both being 1366In this mode the values for \f(CW\*(C`interval\*(C'\fR and \f(CW\*(C`at\*(C'\fR are both being
1210ignored. Instead, each time the periodic watcher gets scheduled, the 1367ignored. Instead, each time the periodic watcher gets scheduled, the
1211reschedule callback will be called with the watcher as first, and the 1368reschedule callback will be called with the watcher as first, and the
1212current time as second argument. 1369current time as second argument.
1213.Sp 1370.Sp
1214\&\s-1NOTE:\s0 \fIThis callback \s-1MUST\s0 \s-1NOT\s0 stop or destroy any periodic watcher, 1371\&\s-1NOTE:\s0 \fIThis callback \s-1MUST\s0 \s-1NOT\s0 stop or destroy any periodic watcher,
1215ever, or make any event loop modifications\fR. If you need to stop it, 1372ever, or make any event loop modifications\fR. If you need to stop it,
1216return \f(CW\*(C`now + 1e30\*(C'\fR (or so, fudge fudge) and stop it afterwards (e.g. by 1373return \f(CW\*(C`now + 1e30\*(C'\fR (or so, fudge fudge) and stop it afterwards (e.g. by
1217starting a prepare watcher). 1374starting an \f(CW\*(C`ev_prepare\*(C'\fR watcher, which is legal).
1218.Sp 1375.Sp
1219Its prototype is \f(CW\*(C`ev_tstamp (*reschedule_cb)(struct ev_periodic *w, 1376Its prototype is \f(CW\*(C`ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
1220ev_tstamp now)\*(C'\fR, e.g.: 1377ev_tstamp now)\*(C'\fR, e.g.:
1221.Sp 1378.Sp
1222.Vb 4 1379.Vb 4
… …
1246.IX Item "ev_periodic_again (loop, ev_periodic *)" 1403.IX Item "ev_periodic_again (loop, ev_periodic *)"
1247Simply stops and restarts the periodic watcher again. This is only useful 1404Simply stops and restarts the periodic watcher again. This is only useful
1248when you changed some parameters or the reschedule callback would return 1405when you changed some parameters or the reschedule callback would return
1249a different time than the last time it was called (e.g. in a crond like 1406a different time than the last time it was called (e.g. in a crond like
1250program when the crontabs have changed). 1407program when the crontabs have changed).
1408.IP "ev_tstamp offset [read\-write]" 4
1409.IX Item "ev_tstamp offset [read-write]"
1410When repeating, this contains the offset value, otherwise this is the
1411absolute point in time (the \f(CW\*(C`at\*(C'\fR value passed to \f(CW\*(C`ev_periodic_set\*(C'\fR).
1412.Sp
1413Can be modified any time, but changes only take effect when the periodic
1414timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being called.
1251.IP "ev_tstamp interval [read\-write]" 4 1415.IP "ev_tstamp interval [read\-write]" 4
1252.IX Item "ev_tstamp interval [read-write]" 1416.IX Item "ev_tstamp interval [read-write]"
1253The current interval value. Can be modified any time, but changes only 1417The current interval value. Can be modified any time, but changes only
1254take effect when the periodic timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being 1418take effect when the periodic timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being
1255called. 1419called.
1256.IP "ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read\-write]" 4 1420.IP "ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read\-write]" 4
1257.IX Item "ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]" 1421.IX Item "ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]"
1258The current reschedule callback, or \f(CW0\fR, if this functionality is 1422The current reschedule callback, or \f(CW0\fR, if this functionality is
1259switched off. Can be changed any time, but changes only take effect when 1423switched off. Can be changed any time, but changes only take effect when
1260the periodic timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being called. 1424the periodic timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being called.
1425.IP "ev_tstamp at [read\-only]" 4
1426.IX Item "ev_tstamp at [read-only]"
1427When active, contains the absolute time that the watcher is supposed to
1428trigger next.
1261.PP 1429.PP
1262Example: Call a callback every hour, or, more precisely, whenever the 1430Example: Call a callback every hour, or, more precisely, whenever the
1263system clock is divisible by 3600. The callback invocation times have 1431system clock is divisible by 3600. The callback invocation times have
1264potentially a lot of jittering, but good long-term stability. 1432potentially a lot of jittering, but good long-term stability.
1265.PP 1433.PP
… …
1315first watcher gets started will libev actually register a signal watcher 1483first watcher gets started will libev actually register a signal watcher
1316with the kernel (thus it coexists with your own signal handlers as long 1484with the kernel (thus it coexists with your own signal handlers as long
1317as you don't register any with libev). Similarly, when the last signal 1485as you don't register any with libev). Similarly, when the last signal
1318watcher for a signal is stopped libev will reset the signal handler to 1486watcher for a signal is stopped libev will reset the signal handler to
1319\&\s-1SIG_DFL\s0 (regardless of what it was set to before). 1487\&\s-1SIG_DFL\s0 (regardless of what it was set to before).
1488.PP
1489\fIWatcher-Specific Functions and Data Members\fR
1490.IX Subsection "Watcher-Specific Functions and Data Members"
1320.IP "ev_signal_init (ev_signal *, callback, int signum)" 4 1491.IP "ev_signal_init (ev_signal *, callback, int signum)" 4
1321.IX Item "ev_signal_init (ev_signal *, callback, int signum)" 1492.IX Item "ev_signal_init (ev_signal *, callback, int signum)"
1322.PD 0 1493.PD 0
1323.IP "ev_signal_set (ev_signal *, int signum)" 4 1494.IP "ev_signal_set (ev_signal *, int signum)" 4
1324.IX Item "ev_signal_set (ev_signal *, int signum)" 1495.IX Item "ev_signal_set (ev_signal *, int signum)"
… …
1331.ie n .Sh """ev_child"" \- watch out for process status changes" 1502.ie n .Sh """ev_child"" \- watch out for process status changes"
1332.el .Sh "\f(CWev_child\fP \- watch out for process status changes" 1503.el .Sh "\f(CWev_child\fP \- watch out for process status changes"
1333.IX Subsection "ev_child - watch out for process status changes" 1504.IX Subsection "ev_child - watch out for process status changes"
1334Child watchers trigger when your process receives a \s-1SIGCHLD\s0 in response to 1505Child watchers trigger when your process receives a \s-1SIGCHLD\s0 in response to
1335some child status changes (most typically when a child of yours dies). 1506some child status changes (most typically when a child of yours dies).
1507.PP
1508\fIWatcher-Specific Functions and Data Members\fR
1509.IX Subsection "Watcher-Specific Functions and Data Members"
1336.IP "ev_child_init (ev_child *, callback, int pid)" 4 1510.IP "ev_child_init (ev_child *, callback, int pid)" 4
1337.IX Item "ev_child_init (ev_child *, callback, int pid)" 1511.IX Item "ev_child_init (ev_child *, callback, int pid)"
1338.PD 0 1512.PD 0
1339.IP "ev_child_set (ev_child *, int pid)" 4 1513.IP "ev_child_set (ev_child *, int pid)" 4
1340.IX Item "ev_child_set (ev_child *, int pid)" 1514.IX Item "ev_child_set (ev_child *, int pid)"
… …
1382not exist\*(R" is a status change like any other. The condition \*(L"path does 1556not exist\*(R" is a status change like any other. The condition \*(L"path does
1383not exist\*(R" is signified by the \f(CW\*(C`st_nlink\*(C'\fR field being zero (which is 1557not exist\*(R" is signified by the \f(CW\*(C`st_nlink\*(C'\fR field being zero (which is
1384otherwise always forced to be at least one) and all the other fields of 1558otherwise always forced to be at least one) and all the other fields of
1385the stat buffer having unspecified contents. 1559the stat buffer having unspecified contents.
1386.PP 1560.PP
1561The path \fIshould\fR be absolute and \fImust not\fR end in a slash. If it is
1562relative and your working directory changes, the behaviour is undefined.
1563.PP
1387Since there is no standard to do this, the portable implementation simply 1564Since there is no standard to do this, the portable implementation simply
1388calls \f(CW\*(C`stat (2)\*(C'\fR regulalry on the path to see if it changed somehow. You 1565calls \f(CW\*(C`stat (2)\*(C'\fR regularly on the path to see if it changed somehow. You
1389can specify a recommended polling interval for this case. If you specify 1566can specify a recommended polling interval for this case. If you specify
1390a polling interval of \f(CW0\fR (highly recommended!) then a \fIsuitable, 1567a polling interval of \f(CW0\fR (highly recommended!) then a \fIsuitable,
1391unspecified default\fR value will be used (which you can expect to be around 1568unspecified default\fR value will be used (which you can expect to be around
1392five seconds, although this might change dynamically). Libev will also 1569five seconds, although this might change dynamically). Libev will also
1393impose a minimum interval which is currently around \f(CW0.1\fR, but thats 1570impose a minimum interval which is currently around \f(CW0.1\fR, but thats
… …
1395.PP 1572.PP
1396This watcher type is not meant for massive numbers of stat watchers, 1573This watcher type is not meant for massive numbers of stat watchers,
1397as even with OS-supported change notifications, this can be 1574as even with OS-supported change notifications, this can be
1398resource\-intensive. 1575resource\-intensive.
1399.PP 1576.PP
1400At the time of this writing, no specific \s-1OS\s0 backends are implemented, but 1577At the time of this writing, only the Linux inotify interface is
1401if demand increases, at least a kqueue and inotify backend will be added. 1578implemented (implementing kqueue support is left as an exercise for the
1579reader). Inotify will be used to give hints only and should not change the
1580semantics of \f(CW\*(C`ev_stat\*(C'\fR watchers, which means that libev sometimes needs
1581to fall back to regular polling again even with inotify, but changes are
1582usually detected immediately, and if the file exists there will be no
1583polling.
1584.PP
1585\fIWatcher-Specific Functions and Data Members\fR
1586.IX Subsection "Watcher-Specific Functions and Data Members"
1402.IP "ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)" 4 1587.IP "ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)" 4
1403.IX Item "ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)" 1588.IX Item "ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)"
1404.PD 0 1589.PD 0
1405.IP "ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)" 4 1590.IP "ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)" 4
1406.IX Item "ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)" 1591.IX Item "ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)"
… …
1467\& ev_stat_start (loop, &passwd); 1652\& ev_stat_start (loop, &passwd);
1468.Ve 1653.Ve
1469.ie n .Sh """ev_idle"" \- when you've got nothing better to do..." 1654.ie n .Sh """ev_idle"" \- when you've got nothing better to do..."
1470.el .Sh "\f(CWev_idle\fP \- when you've got nothing better to do..." 1655.el .Sh "\f(CWev_idle\fP \- when you've got nothing better to do..."
1471.IX Subsection "ev_idle - when you've got nothing better to do..." 1656.IX Subsection "ev_idle - when you've got nothing better to do..."
1472Idle watchers trigger events when there are no other events are pending 1657Idle watchers trigger events when no other events of the same or higher
1473(prepare, check and other idle watchers do not count). That is, as long 1658priority are pending (prepare, check and other idle watchers do not
1474as your process is busy handling sockets or timeouts (or even signals, 1659count).
1475imagine) it will not be triggered. But when your process is idle all idle 1660.PP
1476watchers are being called again and again, once per event loop iteration \- 1661That is, as long as your process is busy handling sockets or timeouts
1662(or even signals, imagine) of the same or higher priority it will not be
1663triggered. But when your process is idle (or only lower-priority watchers
1664are pending), the idle watchers are being called once per event loop
1477until stopped, that is, or your process receives more events and becomes 1665iteration \- until stopped, that is, or your process receives more events
1478busy. 1666and becomes busy again with higher priority stuff.
1479.PP 1667.PP
1480The most noteworthy effect is that as long as any idle watchers are 1668The most noteworthy effect is that as long as any idle watchers are
1481active, the process will not block when waiting for new events. 1669active, the process will not block when waiting for new events.
1482.PP 1670.PP
1483Apart from keeping your process non-blocking (which is a useful 1671Apart from keeping your process non-blocking (which is a useful
1484effect on its own sometimes), idle watchers are a good place to do 1672effect on its own sometimes), idle watchers are a good place to do
1485\&\*(L"pseudo\-background processing\*(R", or delay processing stuff to after the 1673\&\*(L"pseudo\-background processing\*(R", or delay processing stuff to after the
1486event loop has handled all outstanding events. 1674event loop has handled all outstanding events.
1675.PP
1676\fIWatcher-Specific Functions and Data Members\fR
1677.IX Subsection "Watcher-Specific Functions and Data Members"
1487.IP "ev_idle_init (ev_signal *, callback)" 4 1678.IP "ev_idle_init (ev_signal *, callback)" 4
1488.IX Item "ev_idle_init (ev_signal *, callback)" 1679.IX Item "ev_idle_init (ev_signal *, callback)"
1489Initialises and configures the idle watcher \- it has no parameters of any 1680Initialises and configures the idle watcher \- it has no parameters of any
1490kind. There is a \f(CW\*(C`ev_idle_set\*(C'\fR macro, but using it is utterly pointless, 1681kind. There is a \f(CW\*(C`ev_idle_set\*(C'\fR macro, but using it is utterly pointless,
1491believe me. 1682believe me.
… …
1546are ready to run (it's actually more complicated: it only runs coroutines 1737are ready to run (it's actually more complicated: it only runs coroutines
1547with priority higher than or equal to the event loop and one coroutine 1738with priority higher than or equal to the event loop and one coroutine
1548of lower priority, but only once, using idle watchers to keep the event 1739of lower priority, but only once, using idle watchers to keep the event
1549loop from blocking if lower-priority coroutines are active, thus mapping 1740loop from blocking if lower-priority coroutines are active, thus mapping
1550low-priority coroutines to idle/background tasks). 1741low-priority coroutines to idle/background tasks).
1742.PP
1743It is recommended to give \f(CW\*(C`ev_check\*(C'\fR watchers highest (\f(CW\*(C`EV_MAXPRI\*(C'\fR)
1744priority, to ensure that they are being run before any other watchers
1745after the poll. Also, \f(CW\*(C`ev_check\*(C'\fR watchers (and \f(CW\*(C`ev_prepare\*(C'\fR watchers,
1746too) should not activate (\*(L"feed\*(R") events into libev. While libev fully
1747supports this, they will be called before other \f(CW\*(C`ev_check\*(C'\fR watchers did
1748their job. As \f(CW\*(C`ev_check\*(C'\fR watchers are often used to embed other event
1749loops those other event loops might be in an unusable state until their
1750\&\f(CW\*(C`ev_check\*(C'\fR watcher ran (always remind yourself to coexist peacefully with
1751others).
1752.PP
1753\fIWatcher-Specific Functions and Data Members\fR
1754.IX Subsection "Watcher-Specific Functions and Data Members"
1551.IP "ev_prepare_init (ev_prepare *, callback)" 4 1755.IP "ev_prepare_init (ev_prepare *, callback)" 4
1552.IX Item "ev_prepare_init (ev_prepare *, callback)" 1756.IX Item "ev_prepare_init (ev_prepare *, callback)"
1553.PD 0 1757.PD 0
1554.IP "ev_check_init (ev_check *, callback)" 4 1758.IP "ev_check_init (ev_check *, callback)" 4
1555.IX Item "ev_check_init (ev_check *, callback)" 1759.IX Item "ev_check_init (ev_check *, callback)"
1556.PD 1760.PD
1557Initialises and configures the prepare or check watcher \- they have no 1761Initialises and configures the prepare or check watcher \- they have no
1558parameters of any kind. There are \f(CW\*(C`ev_prepare_set\*(C'\fR and \f(CW\*(C`ev_check_set\*(C'\fR 1762parameters of any kind. There are \f(CW\*(C`ev_prepare_set\*(C'\fR and \f(CW\*(C`ev_check_set\*(C'\fR
1559macros, but using them is utterly, utterly and completely pointless. 1763macros, but using them is utterly, utterly and completely pointless.
1560.PP 1764.PP
1561Example: To include a library such as adns, you would add \s-1IO\s0 watchers 1765There are a number of principal ways to embed other event loops or modules
1562and a timeout watcher in a prepare handler, as required by libadns, and 1766into libev. Here are some ideas on how to include libadns into libev
1767(there is a Perl module named \f(CW\*(C`EV::ADNS\*(C'\fR that does this, which you could
1768use for an actually working example. Another Perl module named \f(CW\*(C`EV::Glib\*(C'\fR
1769embeds a Glib main context into libev, and finally, \f(CW\*(C`Glib::EV\*(C'\fR embeds \s-1EV\s0
1770into the Glib event loop).
1771.PP
1772Method 1: Add \s-1IO\s0 watchers and a timeout watcher in a prepare handler,
1563in a check watcher, destroy them and call into libadns. What follows is 1773and in a check watcher, destroy them and call into libadns. What follows
1564pseudo-code only of course: 1774is pseudo-code only of course. This requires you to either use a low
1775priority for the check watcher or use \f(CW\*(C`ev_clear_pending\*(C'\fR explicitly, as
1776the callbacks for the IO/timeout watchers might not have been called yet.
1565.PP 1777.PP
1566.Vb 2 1778.Vb 2
1567\& static ev_io iow [nfd]; 1779\& static ev_io iow [nfd];
1568\& static ev_timer tw; 1780\& static ev_timer tw;
1569.Ve 1781.Ve
1570.PP 1782.PP
1571.Vb 9 1783.Vb 4
1572\& static void 1784\& static void
1573\& io_cb (ev_loop *loop, ev_io *w, int revents) 1785\& io_cb (ev_loop *loop, ev_io *w, int revents)
1574\& { 1786\& {
1575\& // set the relevant poll flags
1576\& // could also call adns_processreadable etc. here
1577\& struct pollfd *fd = (struct pollfd *)w->data;
1578\& if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1579\& if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1580\& } 1787\& }
1581.Ve 1788.Ve
1582.PP 1789.PP
1583.Vb 7 1790.Vb 8
1584\& // create io watchers for each fd and a timer before blocking 1791\& // create io watchers for each fd and a timer before blocking
1585\& static void 1792\& static void
1586\& adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents) 1793\& adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
1587\& { 1794\& {
1588\& int timeout = 3600000;truct pollfd fds [nfd]; 1795\& int timeout = 3600000;
1796\& struct pollfd fds [nfd];
1589\& // actual code will need to loop here and realloc etc. 1797\& // actual code will need to loop here and realloc etc.
1590\& adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ())); 1798\& adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1591.Ve 1799.Ve
1592.PP 1800.PP
1593.Vb 3 1801.Vb 3
… …
1595\& ev_timer_init (&tw, 0, timeout * 1e-3); 1803\& ev_timer_init (&tw, 0, timeout * 1e-3);
1596\& ev_timer_start (loop, &tw); 1804\& ev_timer_start (loop, &tw);
1597.Ve 1805.Ve
1598.PP 1806.PP
1599.Vb 6 1807.Vb 6
1600\& // create on ev_io per pollfd 1808\& // create one ev_io per pollfd
1601\& for (int i = 0; i < nfd; ++i) 1809\& for (int i = 0; i < nfd; ++i)
1602\& { 1810\& {
1603\& ev_io_init (iow + i, io_cb, fds [i].fd, 1811\& ev_io_init (iow + i, io_cb, fds [i].fd,
1604\& ((fds [i].events & POLLIN ? EV_READ : 0) 1812\& ((fds [i].events & POLLIN ? EV_READ : 0)
1605\& | (fds [i].events & POLLOUT ? EV_WRITE : 0))); 1813\& | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
1606.Ve 1814.Ve
1607.PP 1815.PP
1608.Vb 5 1816.Vb 4
1609\& fds [i].revents = 0; 1817\& fds [i].revents = 0;
1610\& iow [i].data = fds + i;
1611\& ev_io_start (loop, iow + i); 1818\& ev_io_start (loop, iow + i);
1612\& } 1819\& }
1613\& } 1820\& }
1614.Ve 1821.Ve
1615.PP 1822.PP
… …
1619\& adns_check_cb (ev_loop *loop, ev_check *w, int revents) 1826\& adns_check_cb (ev_loop *loop, ev_check *w, int revents)
1620\& { 1827\& {
1621\& ev_timer_stop (loop, &tw); 1828\& ev_timer_stop (loop, &tw);
1622.Ve 1829.Ve
1623.PP 1830.PP
1624.Vb 2 1831.Vb 8
1625\& for (int i = 0; i < nfd; ++i) 1832\& for (int i = 0; i < nfd; ++i)
1833\& {
1834\& // set the relevant poll flags
1835\& // could also call adns_processreadable etc. here
1836\& struct pollfd *fd = fds + i;
1837\& int revents = ev_clear_pending (iow + i);
1838\& if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1839\& if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1840.Ve
1841.PP
1842.Vb 3
1843\& // now stop the watcher
1626\& ev_io_stop (loop, iow + i); 1844\& ev_io_stop (loop, iow + i);
1845\& }
1627.Ve 1846.Ve
1628.PP 1847.PP
1629.Vb 2 1848.Vb 2
1630\& adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop)); 1849\& adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
1850\& }
1851.Ve
1852.PP
1853Method 2: This would be just like method 1, but you run \f(CW\*(C`adns_afterpoll\*(C'\fR
1854in the prepare watcher and would dispose of the check watcher.
1855.PP
1856Method 3: If the module to be embedded supports explicit event
1857notification (adns does), you can also make use of the actual watcher
1858callbacks, and only destroy/create the watchers in the prepare watcher.
1859.PP
1860.Vb 5
1861\& static void
1862\& timer_cb (EV_P_ ev_timer *w, int revents)
1863\& {
1864\& adns_state ads = (adns_state)w->data;
1865\& update_now (EV_A);
1866.Ve
1867.PP
1868.Vb 2
1869\& adns_processtimeouts (ads, &tv_now);
1870\& }
1871.Ve
1872.PP
1873.Vb 5
1874\& static void
1875\& io_cb (EV_P_ ev_io *w, int revents)
1876\& {
1877\& adns_state ads = (adns_state)w->data;
1878\& update_now (EV_A);
1879.Ve
1880.PP
1881.Vb 3
1882\& if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
1883\& if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
1884\& }
1885.Ve
1886.PP
1887.Vb 1
1888\& // do not ever call adns_afterpoll
1889.Ve
1890.PP
1891Method 4: Do not use a prepare or check watcher because the module you
1892want to embed is too inflexible to support it. Instead, youc na override
1893their poll function. The drawback with this solution is that the main
1894loop is now no longer controllable by \s-1EV\s0. The \f(CW\*(C`Glib::EV\*(C'\fR module does
1895this.
1896.PP
1897.Vb 4
1898\& static gint
1899\& event_poll_func (GPollFD *fds, guint nfds, gint timeout)
1900\& {
1901\& int got_events = 0;
1902.Ve
1903.PP
1904.Vb 2
1905\& for (n = 0; n < nfds; ++n)
1906\& // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
1907.Ve
1908.PP
1909.Vb 2
1910\& if (timeout >= 0)
1911\& // create/start timer
1912.Ve
1913.PP
1914.Vb 2
1915\& // poll
1916\& ev_loop (EV_A_ 0);
1917.Ve
1918.PP
1919.Vb 3
1920\& // stop timer again
1921\& if (timeout >= 0)
1922\& ev_timer_stop (EV_A_ &to);
1923.Ve
1924.PP
1925.Vb 3
1926\& // stop io watchers again - their callbacks should have set
1927\& for (n = 0; n < nfds; ++n)
1928\& ev_io_stop (EV_A_ iow [n]);
1929.Ve
1930.PP
1931.Vb 2
1932\& return got_events;
1631\& } 1933\& }
1632.Ve 1934.Ve
1633.ie n .Sh """ev_embed"" \- when one backend isn't enough..." 1935.ie n .Sh """ev_embed"" \- when one backend isn't enough..."
1634.el .Sh "\f(CWev_embed\fP \- when one backend isn't enough..." 1936.el .Sh "\f(CWev_embed\fP \- when one backend isn't enough..."
1635.IX Subsection "ev_embed - when one backend isn't enough..." 1937.IX Subsection "ev_embed - when one backend isn't enough..."
1636This is a rather advanced watcher type that lets you embed one event loop 1938This is a rather advanced watcher type that lets you embed one event loop
1637into another (currently only \f(CW\*(C`ev_io\*(C'\fR events are supported in the embedded 1939into another (currently only \f(CW\*(C`ev_io\*(C'\fR events are supported in the embedded
1638loop, other types of watchers might be handled in a delayed or incorrect 1940loop, other types of watchers might be handled in a delayed or incorrect
1639fashion and must not be used). 1941fashion and must not be used). (See portability notes, below).
1640.PP 1942.PP
1641There are primarily two reasons you would want that: work around bugs and 1943There are primarily two reasons you would want that: work around bugs and
1642prioritise I/O. 1944prioritise I/O.
1643.PP 1945.PP
1644As an example for a bug workaround, the kqueue backend might only support 1946As an example for a bug workaround, the kqueue backend might only support
… …
1704\& ev_embed_start (loop_hi, &embed); 2006\& ev_embed_start (loop_hi, &embed);
1705\& } 2007\& }
1706\& else 2008\& else
1707\& loop_lo = loop_hi; 2009\& loop_lo = loop_hi;
1708.Ve 2010.Ve
2011.Sh "Portability notes"
2012.IX Subsection "Portability notes"
2013Kqueue is nominally embeddable, but this is broken on all BSDs that I
2014tried, in various ways. Usually the embedded event loop will simply never
2015receive events, sometimes it will only trigger a few times, sometimes in a
2016loop. Epoll is also nominally embeddable, but many Linux kernel versions
2017will always eport the epoll fd as ready, even when no events are pending.
2018.PP
2019While libev allows embedding these backends (they are contained in
2020\&\f(CW\*(C`ev_embeddable_backends ()\*(C'\fR), take extreme care that it will actually
2021work.
2022.PP
2023When in doubt, create a dynamic event loop forced to use sockets (this
2024usually works) and possibly another thread and a pipe or so to report to
2025your main event loop.
2026.PP
2027\fIWatcher-Specific Functions and Data Members\fR
2028.IX Subsection "Watcher-Specific Functions and Data Members"
1709.IP "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)" 4 2029.IP "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)" 4
1710.IX Item "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)" 2030.IX Item "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)"
1711.PD 0 2031.PD 0
1712.IP "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)" 4 2032.IP "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)" 4
1713.IX Item "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)" 2033.IX Item "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)"
… …
1720.IP "ev_embed_sweep (loop, ev_embed *)" 4 2040.IP "ev_embed_sweep (loop, ev_embed *)" 4
1721.IX Item "ev_embed_sweep (loop, ev_embed *)" 2041.IX Item "ev_embed_sweep (loop, ev_embed *)"
1722Make a single, non-blocking sweep over the embedded loop. This works 2042Make a single, non-blocking sweep over the embedded loop. This works
1723similarly to \f(CW\*(C`ev_loop (embedded_loop, EVLOOP_NONBLOCK)\*(C'\fR, but in the most 2043similarly to \f(CW\*(C`ev_loop (embedded_loop, EVLOOP_NONBLOCK)\*(C'\fR, but in the most
1724apropriate way for embedded loops. 2044apropriate way for embedded loops.
1725.IP "struct ev_loop *loop [read\-only]" 4 2045.IP "struct ev_loop *other [read\-only]" 4
1726.IX Item "struct ev_loop *loop [read-only]" 2046.IX Item "struct ev_loop *other [read-only]"
1727The embedded event loop. 2047The embedded event loop.
1728.ie n .Sh """ev_fork"" \- the audacity to resume the event loop after a fork" 2048.ie n .Sh """ev_fork"" \- the audacity to resume the event loop after a fork"
1729.el .Sh "\f(CWev_fork\fP \- the audacity to resume the event loop after a fork" 2049.el .Sh "\f(CWev_fork\fP \- the audacity to resume the event loop after a fork"
1730.IX Subsection "ev_fork - the audacity to resume the event loop after a fork" 2050.IX Subsection "ev_fork - the audacity to resume the event loop after a fork"
1731Fork watchers are called when a \f(CW\*(C`fork ()\*(C'\fR was detected (usually because 2051Fork watchers are called when a \f(CW\*(C`fork ()\*(C'\fR was detected (usually because
… …
1733\&\f(CW\*(C`ev_default_fork\*(C'\fR or \f(CW\*(C`ev_loop_fork\*(C'\fR). The invocation is done before the 2053\&\f(CW\*(C`ev_default_fork\*(C'\fR or \f(CW\*(C`ev_loop_fork\*(C'\fR). The invocation is done before the
1734event loop blocks next and before \f(CW\*(C`ev_check\*(C'\fR watchers are being called, 2054event loop blocks next and before \f(CW\*(C`ev_check\*(C'\fR watchers are being called,
1735and only in the child after the fork. If whoever good citizen calling 2055and only in the child after the fork. If whoever good citizen calling
1736\&\f(CW\*(C`ev_default_fork\*(C'\fR cheats and calls it in the wrong process, the fork 2056\&\f(CW\*(C`ev_default_fork\*(C'\fR cheats and calls it in the wrong process, the fork
1737handlers will be invoked, too, of course. 2057handlers will be invoked, too, of course.
2058.PP
2059\fIWatcher-Specific Functions and Data Members\fR
2060.IX Subsection "Watcher-Specific Functions and Data Members"
1738.IP "ev_fork_init (ev_signal *, callback)" 4 2061.IP "ev_fork_init (ev_signal *, callback)" 4
1739.IX Item "ev_fork_init (ev_signal *, callback)" 2062.IX Item "ev_fork_init (ev_signal *, callback)"
1740Initialises and configures the fork watcher \- it has no parameters of any 2063Initialises and configures the fork watcher \- it has no parameters of any
1741kind. There is a \f(CW\*(C`ev_fork_set\*(C'\fR macro, but using it is utterly pointless, 2064kind. There is a \f(CW\*(C`ev_fork_set\*(C'\fR macro, but using it is utterly pointless,
1742believe me. 2065believe me.
… …
1819.PP 2142.PP
1820.Vb 1 2143.Vb 1
1821\& #include <ev++.h> 2144\& #include <ev++.h>
1822.Ve 2145.Ve
1823.PP 2146.PP
1824(it is not installed by default). This automatically includes \fIev.h\fR 2147This automatically includes \fIev.h\fR and puts all of its definitions (many
1825and puts all of its definitions (many of them macros) into the global 2148of them macros) into the global namespace. All \*(C+ specific things are
1826namespace. All \*(C+ specific things are put into the \f(CW\*(C`ev\*(C'\fR namespace. 2149put into the \f(CW\*(C`ev\*(C'\fR namespace. It should support all the same embedding
2150options as \fIev.h\fR, most notably \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR.
1827.PP 2151.PP
1828It should support all the same embedding options as \fIev.h\fR, most notably 2152Care has been taken to keep the overhead low. The only data member the \*(C+
1829\&\f(CW\*(C`EV_MULTIPLICITY\*(C'\fR. 2153classes add (compared to plain C\-style watchers) is the event loop pointer
2154that the watcher is associated with (or no additional members at all if
2155you disable \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR when embedding libev).
2156.PP
2157Currently, functions, and static and non-static member functions can be
2158used as callbacks. Other types should be easy to add as long as they only
2159need one additional pointer for context. If you need support for other
2160types of functors please contact the author (preferably after implementing
2161it).
1830.PP 2162.PP
1831Here is a list of things available in the \f(CW\*(C`ev\*(C'\fR namespace: 2163Here is a list of things available in the \f(CW\*(C`ev\*(C'\fR namespace:
1832.ie n .IP """ev::READ""\fR, \f(CW""ev::WRITE"" etc." 4 2164.ie n .IP """ev::READ""\fR, \f(CW""ev::WRITE"" etc." 4
1833.el .IP "\f(CWev::READ\fR, \f(CWev::WRITE\fR etc." 4 2165.el .IP "\f(CWev::READ\fR, \f(CWev::WRITE\fR etc." 4
1834.IX Item "ev::READ, ev::WRITE etc." 2166.IX Item "ev::READ, ev::WRITE etc."
… …
1846which is called \f(CW\*(C`ev::sig\*(C'\fR to avoid clashes with the \f(CW\*(C`signal\*(C'\fR macro 2178which is called \f(CW\*(C`ev::sig\*(C'\fR to avoid clashes with the \f(CW\*(C`signal\*(C'\fR macro
1847defines by many implementations. 2179defines by many implementations.
1848.Sp 2180.Sp
1849All of those classes have these methods: 2181All of those classes have these methods:
1850.RS 4 2182.RS 4
1851.IP "ev::TYPE::TYPE (object *, object::method *)" 4 2183.IP "ev::TYPE::TYPE ()" 4
1852.IX Item "ev::TYPE::TYPE (object *, object::method *)" 2184.IX Item "ev::TYPE::TYPE ()"
1853.PD 0 2185.PD 0
1854.IP "ev::TYPE::TYPE (object *, object::method *, struct ev_loop *)" 4 2186.IP "ev::TYPE::TYPE (struct ev_loop *)" 4
1855.IX Item "ev::TYPE::TYPE (object *, object::method *, struct ev_loop *)" 2187.IX Item "ev::TYPE::TYPE (struct ev_loop *)"
1856.IP "ev::TYPE::~TYPE" 4 2188.IP "ev::TYPE::~TYPE" 4
1857.IX Item "ev::TYPE::~TYPE" 2189.IX Item "ev::TYPE::~TYPE"
1858.PD 2190.PD
1859The constructor takes a pointer to an object and a method pointer to 2191The constructor (optionally) takes an event loop to associate the watcher
1860the event handler callback to call in this class. The constructor calls 2192with. If it is omitted, it will use \f(CW\*(C`EV_DEFAULT\*(C'\fR.
1861\&\f(CW\*(C`ev_init\*(C'\fR for you, which means you have to call the \f(CW\*(C`set\*(C'\fR method 2193.Sp
1862before starting it. If you do not specify a loop then the constructor 2194The constructor calls \f(CW\*(C`ev_init\*(C'\fR for you, which means you have to call the
1863automatically associates the default loop with this watcher. 2195\&\f(CW\*(C`set\*(C'\fR method before starting it.
2196.Sp
2197It will not set a callback, however: You have to call the templated \f(CW\*(C`set\*(C'\fR
2198method to set a callback before you can start the watcher.
2199.Sp
2200(The reason why you have to use a method is a limitation in \*(C+ which does
2201not allow explicit template arguments for constructors).
1864.Sp 2202.Sp
1865The destructor automatically stops the watcher if it is active. 2203The destructor automatically stops the watcher if it is active.
2204.IP "w\->set<class, &class::method> (object *)" 4
2205.IX Item "w->set<class, &class::method> (object *)"
2206This method sets the callback method to call. The method has to have a
2207signature of \f(CW\*(C`void (*)(ev_TYPE &, int)\*(C'\fR, it receives the watcher as
2208first argument and the \f(CW\*(C`revents\*(C'\fR as second. The object must be given as
2209parameter and is stored in the \f(CW\*(C`data\*(C'\fR member of the watcher.
2210.Sp
2211This method synthesizes efficient thunking code to call your method from
2212the C callback that libev requires. If your compiler can inline your
2213callback (i.e. it is visible to it at the place of the \f(CW\*(C`set\*(C'\fR call and
2214your compiler is good :), then the method will be fully inlined into the
2215thunking function, making it as fast as a direct C callback.
2216.Sp
2217Example: simple class declaration and watcher initialisation
2218.Sp
2219.Vb 4
2220\& struct myclass
2221\& {
2222\& void io_cb (ev::io &w, int revents) { }
2223\& }
2224.Ve
2225.Sp
2226.Vb 3
2227\& myclass obj;
2228\& ev::io iow;
2229\& iow.set <myclass, &myclass::io_cb> (&obj);
2230.Ve
2231.IP "w\->set<function> (void *data = 0)" 4
2232.IX Item "w->set<function> (void *data = 0)"
2233Also sets a callback, but uses a static method or plain function as
2234callback. The optional \f(CW\*(C`data\*(C'\fR argument will be stored in the watcher's
2235\&\f(CW\*(C`data\*(C'\fR member and is free for you to use.
2236.Sp
2237The prototype of the \f(CW\*(C`function\*(C'\fR must be \f(CW\*(C`void (*)(ev::TYPE &w, int)\*(C'\fR.
2238.Sp
2239See the method\-\f(CW\*(C`set\*(C'\fR above for more details.
2240.Sp
2241Example:
2242.Sp
2243.Vb 2
2244\& static void io_cb (ev::io &w, int revents) { }
2245\& iow.set <io_cb> ();
2246.Ve
1866.IP "w\->set (struct ev_loop *)" 4 2247.IP "w\->set (struct ev_loop *)" 4
1867.IX Item "w->set (struct ev_loop *)" 2248.IX Item "w->set (struct ev_loop *)"
1868Associates a different \f(CW\*(C`struct ev_loop\*(C'\fR with this watcher. You can only 2249Associates a different \f(CW\*(C`struct ev_loop\*(C'\fR with this watcher. You can only
1869do this when the watcher is inactive (and not pending either). 2250do this when the watcher is inactive (and not pending either).
1870.IP "w\->set ([args])" 4 2251.IP "w\->set ([args])" 4
1871.IX Item "w->set ([args])" 2252.IX Item "w->set ([args])"
1872Basically the same as \f(CW\*(C`ev_TYPE_set\*(C'\fR, with the same args. Must be 2253Basically the same as \f(CW\*(C`ev_TYPE_set\*(C'\fR, with the same args. Must be
1873called at least once. Unlike the C counterpart, an active watcher gets 2254called at least once. Unlike the C counterpart, an active watcher gets
1874automatically stopped and restarted. 2255automatically stopped and restarted when reconfiguring it with this
2256method.
1875.IP "w\->start ()" 4 2257.IP "w\->start ()" 4
1876.IX Item "w->start ()" 2258.IX Item "w->start ()"
1877Starts the watcher. Note that there is no \f(CW\*(C`loop\*(C'\fR argument as the 2259Starts the watcher. Note that there is no \f(CW\*(C`loop\*(C'\fR argument, as the
1878constructor already takes the loop. 2260constructor already stores the event loop.
1879.IP "w\->stop ()" 4 2261.IP "w\->stop ()" 4
1880.IX Item "w->stop ()" 2262.IX Item "w->stop ()"
1881Stops the watcher if it is active. Again, no \f(CW\*(C`loop\*(C'\fR argument. 2263Stops the watcher if it is active. Again, no \f(CW\*(C`loop\*(C'\fR argument.
1882.ie n .IP "w\->again () ""ev::timer""\fR, \f(CW""ev::periodic"" only" 4 2264.ie n .IP "w\->again () (""ev::timer""\fR, \f(CW""ev::periodic"" only)" 4
1883.el .IP "w\->again () \f(CWev::timer\fR, \f(CWev::periodic\fR only" 4 2265.el .IP "w\->again () (\f(CWev::timer\fR, \f(CWev::periodic\fR only)" 4
1884.IX Item "w->again () ev::timer, ev::periodic only" 2266.IX Item "w->again () (ev::timer, ev::periodic only)"
1885For \f(CW\*(C`ev::timer\*(C'\fR and \f(CW\*(C`ev::periodic\*(C'\fR, this invokes the corresponding 2267For \f(CW\*(C`ev::timer\*(C'\fR and \f(CW\*(C`ev::periodic\*(C'\fR, this invokes the corresponding
1886\&\f(CW\*(C`ev_TYPE_again\*(C'\fR function. 2268\&\f(CW\*(C`ev_TYPE_again\*(C'\fR function.
1887.ie n .IP "w\->sweep () ""ev::embed"" only" 4 2269.ie n .IP "w\->sweep () (""ev::embed"" only)" 4
1888.el .IP "w\->sweep () \f(CWev::embed\fR only" 4 2270.el .IP "w\->sweep () (\f(CWev::embed\fR only)" 4
1889.IX Item "w->sweep () ev::embed only" 2271.IX Item "w->sweep () (ev::embed only)"
1890Invokes \f(CW\*(C`ev_embed_sweep\*(C'\fR. 2272Invokes \f(CW\*(C`ev_embed_sweep\*(C'\fR.
1891.ie n .IP "w\->update () ""ev::stat"" only" 4 2273.ie n .IP "w\->update () (""ev::stat"" only)" 4
1892.el .IP "w\->update () \f(CWev::stat\fR only" 4 2274.el .IP "w\->update () (\f(CWev::stat\fR only)" 4
1893.IX Item "w->update () ev::stat only" 2275.IX Item "w->update () (ev::stat only)"
1894Invokes \f(CW\*(C`ev_stat_stat\*(C'\fR. 2276Invokes \f(CW\*(C`ev_stat_stat\*(C'\fR.
1895.RE 2277.RE
1896.RS 4 2278.RS 4
1897.RE 2279.RE
1898.PP 2280.PP
… …
1909.Vb 2 2291.Vb 2
1910\& myclass (); 2292\& myclass ();
1911\& } 2293\& }
1912.Ve 2294.Ve
1913.PP 2295.PP
1914.Vb 6 2296.Vb 4
1915\& myclass::myclass (int fd) 2297\& myclass::myclass (int fd)
1916\& : io (this, &myclass::io_cb),
1917\& idle (this, &myclass::idle_cb)
1918\& { 2298\& {
2299\& io .set <myclass, &myclass::io_cb > (this);
2300\& idle.set <myclass, &myclass::idle_cb> (this);
2301.Ve
2302.PP
2303.Vb 2
1919\& io.start (fd, ev::READ); 2304\& io.start (fd, ev::READ);
1920\& } 2305\& }
1921.Ve 2306.Ve
1922.SH "MACRO MAGIC" 2307.SH "MACRO MAGIC"
1923.IX Header "MACRO MAGIC" 2308.IX Header "MACRO MAGIC"
1924Libev can be compiled with a variety of options, the most fundemantal is 2309Libev can be compiled with a variety of options, the most fundamantal
1925\&\f(CW\*(C`EV_MULTIPLICITY\*(C'\fR. This option determines wether (most) functions and 2310of which is \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR. This option determines whether (most)
1926callbacks have an initial \f(CW\*(C`struct ev_loop *\*(C'\fR argument. 2311functions and callbacks have an initial \f(CW\*(C`struct ev_loop *\*(C'\fR argument.
1927.PP 2312.PP
1928To make it easier to write programs that cope with either variant, the 2313To make it easier to write programs that cope with either variant, the
1929following macros are defined: 2314following macros are defined:
1930.ie n .IP """EV_A""\fR, \f(CW""EV_A_""" 4 2315.ie n .IP """EV_A""\fR, \f(CW""EV_A_""" 4
1931.el .IP "\f(CWEV_A\fR, \f(CWEV_A_\fR" 4 2316.el .IP "\f(CWEV_A\fR, \f(CWEV_A_\fR" 4
… …
1965.el .IP "\f(CWEV_DEFAULT\fR, \f(CWEV_DEFAULT_\fR" 4 2350.el .IP "\f(CWEV_DEFAULT\fR, \f(CWEV_DEFAULT_\fR" 4
1966.IX Item "EV_DEFAULT, EV_DEFAULT_" 2351.IX Item "EV_DEFAULT, EV_DEFAULT_"
1967Similar to the other two macros, this gives you the value of the default 2352Similar to the other two macros, this gives you the value of the default
1968loop, if multiple loops are supported (\*(L"ev loop default\*(R"). 2353loop, if multiple loops are supported (\*(L"ev loop default\*(R").
1969.PP 2354.PP
1970Example: Declare and initialise a check watcher, working regardless of 2355Example: Declare and initialise a check watcher, utilising the above
1971wether multiple loops are supported or not. 2356macros so it will work regardless of whether multiple loops are supported
2357or not.
1972.PP 2358.PP
1973.Vb 5 2359.Vb 5
1974\& static void 2360\& static void
1975\& check_cb (EV_P_ ev_timer *w, int revents) 2361\& check_cb (EV_P_ ev_timer *w, int revents)
1976\& { 2362\& {
… …
1989Libev can (and often is) directly embedded into host 2375Libev can (and often is) directly embedded into host
1990applications. Examples of applications that embed it include the Deliantra 2376applications. Examples of applications that embed it include the Deliantra
1991Game Server, the \s-1EV\s0 perl module, the \s-1GNU\s0 Virtual Private Ethernet (gvpe) 2377Game Server, the \s-1EV\s0 perl module, the \s-1GNU\s0 Virtual Private Ethernet (gvpe)
1992and rxvt\-unicode. 2378and rxvt\-unicode.
1993.PP 2379.PP
1994The goal is to enable you to just copy the neecssary files into your 2380The goal is to enable you to just copy the necessary files into your
1995source directory without having to change even a single line in them, so 2381source directory without having to change even a single line in them, so
1996you can easily upgrade by simply copying (or having a checked-out copy of 2382you can easily upgrade by simply copying (or having a checked-out copy of
1997libev somewhere in your source tree). 2383libev somewhere in your source tree).
1998.Sh "\s-1FILESETS\s0" 2384.Sh "\s-1FILESETS\s0"
1999.IX Subsection "FILESETS" 2385.IX Subsection "FILESETS"
… …
2039.Vb 1 2425.Vb 1
2040\& ev_win32.c required on win32 platforms only 2426\& ev_win32.c required on win32 platforms only
2041.Ve 2427.Ve
2042.PP 2428.PP
2043.Vb 5 2429.Vb 5
2044\& ev_select.c only when select backend is enabled (which is by default) 2430\& ev_select.c only when select backend is enabled (which is enabled by default)
2045\& ev_poll.c only when poll backend is enabled (disabled by default) 2431\& ev_poll.c only when poll backend is enabled (disabled by default)
2046\& ev_epoll.c only when the epoll backend is enabled (disabled by default) 2432\& ev_epoll.c only when the epoll backend is enabled (disabled by default)
2047\& ev_kqueue.c only when the kqueue backend is enabled (disabled by default) 2433\& ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
2048\& ev_port.c only when the solaris port backend is enabled (disabled by default) 2434\& ev_port.c only when the solaris port backend is enabled (disabled by default)
2049.Ve 2435.Ve
… …
2104.IX Item "EV_USE_MONOTONIC" 2490.IX Item "EV_USE_MONOTONIC"
2105If defined to be \f(CW1\fR, libev will try to detect the availability of the 2491If defined to be \f(CW1\fR, libev will try to detect the availability of the
2106monotonic clock option at both compiletime and runtime. Otherwise no use 2492monotonic clock option at both compiletime and runtime. Otherwise no use
2107of the monotonic clock option will be attempted. If you enable this, you 2493of the monotonic clock option will be attempted. If you enable this, you
2108usually have to link against librt or something similar. Enabling it when 2494usually have to link against librt or something similar. Enabling it when
2109the functionality isn't available is safe, though, althoguh you have 2495the functionality isn't available is safe, though, although you have
2110to make sure you link against any libraries where the \f(CW\*(C`clock_gettime\*(C'\fR 2496to make sure you link against any libraries where the \f(CW\*(C`clock_gettime\*(C'\fR
2111function is hiding in (often \fI\-lrt\fR). 2497function is hiding in (often \fI\-lrt\fR).
2112.IP "\s-1EV_USE_REALTIME\s0" 4 2498.IP "\s-1EV_USE_REALTIME\s0" 4
2113.IX Item "EV_USE_REALTIME" 2499.IX Item "EV_USE_REALTIME"
2114If defined to be \f(CW1\fR, libev will try to detect the availability of the 2500If defined to be \f(CW1\fR, libev will try to detect the availability of the
2115realtime clock option at compiletime (and assume its availability at 2501realtime clock option at compiletime (and assume its availability at
2116runtime if successful). Otherwise no use of the realtime clock option will 2502runtime if successful). Otherwise no use of the realtime clock option will
2117be attempted. This effectively replaces \f(CW\*(C`gettimeofday\*(C'\fR by \f(CW\*(C`clock_get 2503be attempted. This effectively replaces \f(CW\*(C`gettimeofday\*(C'\fR by \f(CW\*(C`clock_get
2118(CLOCK_REALTIME, ...)\*(C'\fR and will not normally affect correctness. See tzhe note about libraries 2504(CLOCK_REALTIME, ...)\*(C'\fR and will not normally affect correctness. See the
2119in the description of \f(CW\*(C`EV_USE_MONOTONIC\*(C'\fR, though. 2505note about libraries in the description of \f(CW\*(C`EV_USE_MONOTONIC\*(C'\fR, though.
2120.IP "\s-1EV_USE_SELECT\s0" 4 2506.IP "\s-1EV_USE_SELECT\s0" 4
2121.IX Item "EV_USE_SELECT" 2507.IX Item "EV_USE_SELECT"
2122If undefined or defined to be \f(CW1\fR, libev will compile in support for the 2508If undefined or defined to be \f(CW1\fR, libev will compile in support for the
2123\&\f(CW\*(C`select\*(C'\fR(2) backend. No attempt at autodetection will be done: if no 2509\&\f(CW\*(C`select\*(C'\fR(2) backend. No attempt at autodetection will be done: if no
2124other method takes over, select will be it. Otherwise the select backend 2510other method takes over, select will be it. Otherwise the select backend
… …
2170otherwise another method will be used as fallback. This is the preferred 2556otherwise another method will be used as fallback. This is the preferred
2171backend for Solaris 10 systems. 2557backend for Solaris 10 systems.
2172.IP "\s-1EV_USE_DEVPOLL\s0" 4 2558.IP "\s-1EV_USE_DEVPOLL\s0" 4
2173.IX Item "EV_USE_DEVPOLL" 2559.IX Item "EV_USE_DEVPOLL"
2174reserved for future expansion, works like the \s-1USE\s0 symbols above. 2560reserved for future expansion, works like the \s-1USE\s0 symbols above.
2561.IP "\s-1EV_USE_INOTIFY\s0" 4
2562.IX Item "EV_USE_INOTIFY"
2563If defined to be \f(CW1\fR, libev will compile in support for the Linux inotify
2564interface to speed up \f(CW\*(C`ev_stat\*(C'\fR watchers. Its actual availability will
2565be detected at runtime.
2175.IP "\s-1EV_H\s0" 4 2566.IP "\s-1EV_H\s0" 4
2176.IX Item "EV_H" 2567.IX Item "EV_H"
2177The name of the \fIev.h\fR header file used to include it. The default if 2568The name of the \fIev.h\fR header file used to include it. The default if
2178undefined is \f(CW\*(C`<ev.h>\*(C'\fR in \fIevent.h\fR and \f(CW"ev.h"\fR in \fIev.c\fR. This 2569undefined is \f(CW\*(C`<ev.h>\*(C'\fR in \fIevent.h\fR and \f(CW"ev.h"\fR in \fIev.c\fR. This
2179can be used to virtually rename the \fIev.h\fR header file in case of conflicts. 2570can be used to virtually rename the \fIev.h\fR header file in case of conflicts.
… …
2197If undefined or defined to \f(CW1\fR, then all event-loop-specific functions 2588If undefined or defined to \f(CW1\fR, then all event-loop-specific functions
2198will have the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument, and you can create 2589will have the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument, and you can create
2199additional independent event loops. Otherwise there will be no support 2590additional independent event loops. Otherwise there will be no support
2200for multiple event loops and there is no first event loop pointer 2591for multiple event loops and there is no first event loop pointer
2201argument. Instead, all functions act on the single default loop. 2592argument. Instead, all functions act on the single default loop.
2593.IP "\s-1EV_MINPRI\s0" 4
2594.IX Item "EV_MINPRI"
2595.PD 0
2596.IP "\s-1EV_MAXPRI\s0" 4
2597.IX Item "EV_MAXPRI"
2598.PD
2599The range of allowed priorities. \f(CW\*(C`EV_MINPRI\*(C'\fR must be smaller or equal to
2600\&\f(CW\*(C`EV_MAXPRI\*(C'\fR, but otherwise there are no non-obvious limitations. You can
2601provide for more priorities by overriding those symbols (usually defined
2602to be \f(CW\*(C`\-2\*(C'\fR and \f(CW2\fR, respectively).
2603.Sp
2604When doing priority-based operations, libev usually has to linearly search
2605all the priorities, so having many of them (hundreds) uses a lot of space
2606and time, so using the defaults of five priorities (\-2 .. +2) is usually
2607fine.
2608.Sp
2609If your embedding app does not need any priorities, defining these both to
2610\&\f(CW0\fR will save some memory and cpu.
2202.IP "\s-1EV_PERIODIC_ENABLE\s0" 4 2611.IP "\s-1EV_PERIODIC_ENABLE\s0" 4
2203.IX Item "EV_PERIODIC_ENABLE" 2612.IX Item "EV_PERIODIC_ENABLE"
2204If undefined or defined to be \f(CW1\fR, then periodic timers are supported. If 2613If undefined or defined to be \f(CW1\fR, then periodic timers are supported. If
2614defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of
2615code.
2616.IP "\s-1EV_IDLE_ENABLE\s0" 4
2617.IX Item "EV_IDLE_ENABLE"
2618If undefined or defined to be \f(CW1\fR, then idle watchers are supported. If
2205defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of 2619defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of
2206code. 2620code.
2207.IP "\s-1EV_EMBED_ENABLE\s0" 4 2621.IP "\s-1EV_EMBED_ENABLE\s0" 4
2208.IX Item "EV_EMBED_ENABLE" 2622.IX Item "EV_EMBED_ENABLE"
2209If undefined or defined to be \f(CW1\fR, then embed watchers are supported. If 2623If undefined or defined to be \f(CW1\fR, then embed watchers are supported. If
… …
2224.IP "\s-1EV_PID_HASHSIZE\s0" 4 2638.IP "\s-1EV_PID_HASHSIZE\s0" 4
2225.IX Item "EV_PID_HASHSIZE" 2639.IX Item "EV_PID_HASHSIZE"
2226\&\f(CW\*(C`ev_child\*(C'\fR watchers use a small hash table to distribute workload by 2640\&\f(CW\*(C`ev_child\*(C'\fR watchers use a small hash table to distribute workload by
2227pid. The default size is \f(CW16\fR (or \f(CW1\fR with \f(CW\*(C`EV_MINIMAL\*(C'\fR), usually more 2641pid. The default size is \f(CW16\fR (or \f(CW1\fR with \f(CW\*(C`EV_MINIMAL\*(C'\fR), usually more
2228than enough. If you need to manage thousands of children you might want to 2642than enough. If you need to manage thousands of children you might want to
2229increase this value. 2643increase this value (\fImust\fR be a power of two).
2644.IP "\s-1EV_INOTIFY_HASHSIZE\s0" 4
2645.IX Item "EV_INOTIFY_HASHSIZE"
2646\&\f(CW\*(C`ev_staz\*(C'\fR watchers use a small hash table to distribute workload by
2647inotify watch id. The default size is \f(CW16\fR (or \f(CW1\fR with \f(CW\*(C`EV_MINIMAL\*(C'\fR),
2648usually more than enough. If you need to manage thousands of \f(CW\*(C`ev_stat\*(C'\fR
2649watchers you might want to increase this value (\fImust\fR be a power of
2650two).
2230.IP "\s-1EV_COMMON\s0" 4 2651.IP "\s-1EV_COMMON\s0" 4
2231.IX Item "EV_COMMON" 2652.IX Item "EV_COMMON"
2232By default, all watchers have a \f(CW\*(C`void *data\*(C'\fR member. By redefining 2653By default, all watchers have a \f(CW\*(C`void *data\*(C'\fR member. By redefining
2233this macro to a something else you can include more and other types of 2654this macro to a something else you can include more and other types of
2234members. You have to define it each time you include one of the files, 2655members. You have to define it each time you include one of the files,
… …
2249.IP "ev_set_cb (ev, cb)" 4 2670.IP "ev_set_cb (ev, cb)" 4
2250.IX Item "ev_set_cb (ev, cb)" 2671.IX Item "ev_set_cb (ev, cb)"
2251.PD 2672.PD
2252Can be used to change the callback member declaration in each watcher, 2673Can be used to change the callback member declaration in each watcher,
2253and the way callbacks are invoked and set. Must expand to a struct member 2674and the way callbacks are invoked and set. Must expand to a struct member
2254definition and a statement, respectively. See the \fIev.v\fR header file for 2675definition and a statement, respectively. See the \fIev.h\fR header file for
2255their default definitions. One possible use for overriding these is to 2676their default definitions. One possible use for overriding these is to
2256avoid the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument in all cases, or to use 2677avoid the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument in all cases, or to use
2257method calls instead of plain function calls in \*(C+. 2678method calls instead of plain function calls in \*(C+.
2679.Sh "\s-1EXPORTED\s0 \s-1API\s0 \s-1SYMBOLS\s0"
2680.IX Subsection "EXPORTED API SYMBOLS"
2681If you need to re-export the \s-1API\s0 (e.g. via a dll) and you need a list of
2682exported symbols, you can use the provided \fISymbol.*\fR files which list
2683all public symbols, one per line:
2684.Sp
2685.Vb 2
2686\& Symbols.ev for libev proper
2687\& Symbols.event for the libevent emulation
2688.Ve
2689.Sp
2690This can also be used to rename all public symbols to avoid clashes with
2691multiple versions of libev linked together (which is obviously bad in
2692itself, but sometimes it is inconvinient to avoid this).
2693.Sp
2694A sed command like this will create wrapper \f(CW\*(C`#define\*(C'\fR's that you need to
2695include before including \fIev.h\fR:
2696.Sp
2697.Vb 1
2698\& <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
2699.Ve
2700.Sp
2701This would create a file \fIwrap.h\fR which essentially looks like this:
2702.Sp
2703.Vb 4
2704\& #define ev_backend myprefix_ev_backend
2705\& #define ev_check_start myprefix_ev_check_start
2706\& #define ev_check_stop myprefix_ev_check_stop
2707\& ...
2708.Ve
2258.Sh "\s-1EXAMPLES\s0" 2709.Sh "\s-1EXAMPLES\s0"
2259.IX Subsection "EXAMPLES" 2710.IX Subsection "EXAMPLES"
2260For a real-world example of a program the includes libev 2711For a real-world example of a program the includes libev
2261verbatim, you can have a look at the \s-1EV\s0 perl module 2712verbatim, you can have a look at the \s-1EV\s0 perl module
2262(<http://software.schmorp.de/pkg/EV.html>). It has the libev files in 2713(<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
… …
2264interface) and \fI\s-1EV\s0.xs\fR (implementation) files. Only the \fI\s-1EV\s0.xs\fR file 2715interface) and \fI\s-1EV\s0.xs\fR (implementation) files. Only the \fI\s-1EV\s0.xs\fR file
2265will be compiled. It is pretty complex because it provides its own header 2716will be compiled. It is pretty complex because it provides its own header
2266file. 2717file.
2267.Sp 2718.Sp
2268The usage in rxvt-unicode is simpler. It has a \fIev_cpp.h\fR header file 2719The usage in rxvt-unicode is simpler. It has a \fIev_cpp.h\fR header file
2269that everybody includes and which overrides some autoconf choices: 2720that everybody includes and which overrides some configure choices:
2270.Sp 2721.Sp
2271.Vb 4 2722.Vb 9
2723\& #define EV_MINIMAL 1
2272\& #define EV_USE_POLL 0 2724\& #define EV_USE_POLL 0
2273\& #define EV_MULTIPLICITY 0 2725\& #define EV_MULTIPLICITY 0
2274\& #define EV_PERIODICS 0 2726\& #define EV_PERIODIC_ENABLE 0
2727\& #define EV_STAT_ENABLE 0
2728\& #define EV_FORK_ENABLE 0
2275\& #define EV_CONFIG_H <config.h> 2729\& #define EV_CONFIG_H <config.h>
2730\& #define EV_MINPRI 0
2731\& #define EV_MAXPRI 0
2276.Ve 2732.Ve
2277.Sp 2733.Sp
2278.Vb 1 2734.Vb 1
2279\& #include "ev++.h" 2735\& #include "ev++.h"
2280.Ve 2736.Ve
… …
2288.SH "COMPLEXITIES" 2744.SH "COMPLEXITIES"
2289.IX Header "COMPLEXITIES" 2745.IX Header "COMPLEXITIES"
2290In this section the complexities of (many of) the algorithms used inside 2746In this section the complexities of (many of) the algorithms used inside
2291libev will be explained. For complexity discussions about backends see the 2747libev will be explained. For complexity discussions about backends see the
2292documentation for \f(CW\*(C`ev_default_init\*(C'\fR. 2748documentation for \f(CW\*(C`ev_default_init\*(C'\fR.
2749.Sp
2750All of the following are about amortised time: If an array needs to be
2751extended, libev needs to realloc and move the whole array, but this
2752happens asymptotically never with higher number of elements, so O(1) might
2753mean it might do a lengthy realloc operation in rare cases, but on average
2754it is much faster and asymptotically approaches constant time.
2293.RS 4 2755.RS 4
2294.IP "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)" 4 2756.IP "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)" 4
2295.IX Item "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)" 2757.IX Item "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)"
2296.PD 0 2758This means that, when you have a watcher that triggers in one hour and
2759there are 100 watchers that would trigger before that then inserting will
2760have to skip those 100 watchers.
2297.IP "Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)" 4 2761.IP "Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)" 4
2298.IX Item "Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)" 2762.IX Item "Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)"
2763That means that for changing a timer costs less than removing/adding them
2764as only the relative motion in the event queue has to be paid for.
2299.IP "Starting io/check/prepare/idle/signal/child watchers: O(1)" 4 2765.IP "Starting io/check/prepare/idle/signal/child watchers: O(1)" 4
2300.IX Item "Starting io/check/prepare/idle/signal/child watchers: O(1)" 2766.IX Item "Starting io/check/prepare/idle/signal/child watchers: O(1)"
2301.IP "Stopping check/prepare/idle watchers: O(1)" 4 2767These just add the watcher into an array or at the head of a list.
2302.IX Item "Stopping check/prepare/idle watchers: O(1)" 2768=item Stopping check/prepare/idle watchers: O(1)
2303.IP "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % 16))" 4 2769.IP "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % \s-1EV_PID_HASHSIZE\s0))" 4
2304.IX Item "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % 16))" 2770.IX Item "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))"
2771These watchers are stored in lists then need to be walked to find the
2772correct watcher to remove. The lists are usually short (you don't usually
2773have many watchers waiting for the same fd or signal).
2305.IP "Finding the next timer per loop iteration: O(1)" 4 2774.IP "Finding the next timer per loop iteration: O(1)" 4
2306.IX Item "Finding the next timer per loop iteration: O(1)" 2775.IX Item "Finding the next timer per loop iteration: O(1)"
2776.PD 0
2307.IP "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)" 4 2777.IP "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)" 4
2308.IX Item "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)" 2778.IX Item "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)"
2779.PD
2780A change means an I/O watcher gets started or stopped, which requires
2781libev to recalculate its status (and possibly tell the kernel).
2309.IP "Activating one watcher: O(1)" 4 2782.IP "Activating one watcher: O(1)" 4
2310.IX Item "Activating one watcher: O(1)" 2783.IX Item "Activating one watcher: O(1)"
2784.PD 0
2785.IP "Priority handling: O(number_of_priorities)" 4
2786.IX Item "Priority handling: O(number_of_priorities)"
2787.PD
2788Priorities are implemented by allocating some space for each
2789priority. When doing priority-based operations, libev usually has to
2790linearly search all the priorities.
2311.RE 2791.RE
2312.RS 4 2792.RS 4
2313.PD
2314.SH "AUTHOR" 2793.SH "AUTHOR"
2315.IX Header "AUTHOR" 2794.IX Header "AUTHOR"
2316Marc Lehmann <libev@schmorp.de>. 2795Marc Lehmann <libev@schmorp.de>.

Diff Legend

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