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

Comparing libev/ev.3 (file contents):
Revision 1.16 by root, Sat Nov 24 10:19:14 2007 UTC vs.
Revision 1.77 by root, Sun Feb 15 01:44:40 2009 UTC

1.\" Automatically generated by Pod::Man v1.37, Pod::Parser v1.35 1.\" Automatically generated by Pod::Man 2.16 (Pod::Simple 3.05)
2.\" 2.\"
3.\" Standard preamble: 3.\" Standard preamble:
4.\" ======================================================================== 4.\" ========================================================================
5.de Sh \" Subsection heading 5.de Sh \" Subsection heading
6.br 6.br
… …
23.ft R 23.ft R
24.fi 24.fi
25.. 25..
26.\" Set up some character translations and predefined strings. \*(-- will 26.\" Set up some character translations and predefined strings. \*(-- will
27.\" give an unbreakable dash, \*(PI will give pi, \*(L" will give a left 27.\" give an unbreakable dash, \*(PI will give pi, \*(L" will give a left
28.\" double quote, and \*(R" will give a right double quote. | will give a 28.\" double quote, and \*(R" will give a right double quote. \*(C+ will
29.\" real vertical bar. \*(C+ will give a nicer C++. Capital omega is used to 29.\" give a nicer C++. Capital omega is used to do unbreakable dashes and
30.\" do unbreakable dashes and therefore won't be available. \*(C` and \*(C' 30.\" therefore won't be available. \*(C` and \*(C' expand to `' in nroff,
31.\" expand to `' in nroff, nothing in troff, for use with C<>. 31.\" nothing in troff, for use with C<>.
32.tr \(*W-|\(bv\*(Tr 32.tr \(*W-
33.ds C+ C\v'-.1v'\h'-1p'\s-2+\h'-1p'+\s0\v'.1v'\h'-1p' 33.ds C+ C\v'-.1v'\h'-1p'\s-2+\h'-1p'+\s0\v'.1v'\h'-1p'
34.ie n \{\ 34.ie n \{\
35. ds -- \(*W- 35. ds -- \(*W-
36. ds PI pi 36. ds PI pi
37. if (\n(.H=4u)&(1m=24u) .ds -- \(*W\h'-12u'\(*W\h'-12u'-\" diablo 10 pitch 37. if (\n(.H=4u)&(1m=24u) .ds -- \(*W\h'-12u'\(*W\h'-12u'-\" diablo 10 pitch
… …
46. ds PI \(*p 46. ds PI \(*p
47. ds L" `` 47. ds L" ``
48. ds R" '' 48. ds R" ''
49'br\} 49'br\}
50.\" 50.\"
51.\" Escape single quotes in literal strings from groff's Unicode transform.
52.ie \n(.g .ds Aq \(aq
53.el .ds Aq '
54.\"
51.\" If the F register is turned on, we'll generate index entries on stderr for 55.\" If the F register is turned on, we'll generate index entries on stderr for
52.\" titles (.TH), headers (.SH), subsections (.Sh), items (.Ip), and index 56.\" titles (.TH), headers (.SH), subsections (.Sh), items (.Ip), and index
53.\" entries marked with X<> in POD. Of course, you'll have to process the 57.\" entries marked with X<> in POD. Of course, you'll have to process the
54.\" output yourself in some meaningful fashion. 58.\" output yourself in some meaningful fashion.
55.if \nF \{\ 59.ie \nF \{\
56. de IX 60. de IX
57. tm Index:\\$1\t\\n%\t"\\$2" 61. tm Index:\\$1\t\\n%\t"\\$2"
58.. 62..
59. nr % 0 63. nr % 0
60. rr F 64. rr F
61.\} 65.\}
62.\" 66.el \{\
63.\" For nroff, turn off justification. Always turn off hyphenation; it makes 67. de IX
64.\" way too many mistakes in technical documents. 68..
65.hy 0 69.\}
66.if n .na
67.\" 70.\"
68.\" Accent mark definitions (@(#)ms.acc 1.5 88/02/08 SMI; from UCB 4.2). 71.\" Accent mark definitions (@(#)ms.acc 1.5 88/02/08 SMI; from UCB 4.2).
69.\" Fear. Run. Save yourself. No user-serviceable parts. 72.\" Fear. Run. Save yourself. No user-serviceable parts.
70. \" fudge factors for nroff and troff 73. \" fudge factors for nroff and troff
71.if n \{\ 74.if n \{\
… …
126. ds Ae AE 129. ds Ae AE
127.\} 130.\}
128.rm #[ #] #H #V #F C 131.rm #[ #] #H #V #F C
129.\" ======================================================================== 132.\" ========================================================================
130.\" 133.\"
131.IX Title ""<STANDARD INPUT>" 1" 134.IX Title "LIBEV 3"
132.TH "<STANDARD INPUT>" 1 "2007-11-24" "perl v5.8.8" "User Contributed Perl Documentation" 135.TH LIBEV 3 "2009-02-06" "libev-3.53" "libev - high performance full featured event loop"
136.\" For nroff, turn off justification. Always turn off hyphenation; it makes
137.\" way too many mistakes in technical documents.
138.if n .ad l
139.nh
133.SH "NAME" 140.SH "NAME"
134libev \- a high performance full\-featured event loop written in C 141libev \- a high performance full\-featured event loop written in C
135.SH "SYNOPSIS" 142.SH "SYNOPSIS"
136.IX Header "SYNOPSIS" 143.IX Header "SYNOPSIS"
137.Vb 1 144.Vb 1
138\& #include <ev.h> 145\& #include <ev.h>
146.Ve
147.Sh "\s-1EXAMPLE\s0 \s-1PROGRAM\s0"
148.IX Subsection "EXAMPLE PROGRAM"
149.Vb 2
150\& // a single header file is required
151\& #include <ev.h>
152\&
153\& #include <stdio.h> // for puts
154\&
155\& // every watcher type has its own typedef\*(Aqd struct
156\& // with the name ev_TYPE
157\& ev_io stdin_watcher;
158\& ev_timer timeout_watcher;
159\&
160\& // all watcher callbacks have a similar signature
161\& // this callback is called when data is readable on stdin
162\& static void
163\& stdin_cb (EV_P_ ev_io *w, int revents)
164\& {
165\& puts ("stdin ready");
166\& // for one\-shot events, one must manually stop the watcher
167\& // with its corresponding stop function.
168\& ev_io_stop (EV_A_ w);
169\&
170\& // this causes all nested ev_loop\*(Aqs to stop iterating
171\& ev_unloop (EV_A_ EVUNLOOP_ALL);
172\& }
173\&
174\& // another callback, this time for a time\-out
175\& static void
176\& timeout_cb (EV_P_ ev_timer *w, int revents)
177\& {
178\& puts ("timeout");
179\& // this causes the innermost ev_loop to stop iterating
180\& ev_unloop (EV_A_ EVUNLOOP_ONE);
181\& }
182\&
183\& int
184\& main (void)
185\& {
186\& // use the default event loop unless you have special needs
187\& struct ev_loop *loop = ev_default_loop (0);
188\&
189\& // initialise an io watcher, then start it
190\& // this one will watch for stdin to become readable
191\& ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
192\& ev_io_start (loop, &stdin_watcher);
193\&
194\& // initialise a timer watcher, then start it
195\& // simple non\-repeating 5.5 second timeout
196\& ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
197\& ev_timer_start (loop, &timeout_watcher);
198\&
199\& // now wait for events to arrive
200\& ev_loop (loop, 0);
201\&
202\& // unloop was called, so exit
203\& return 0;
204\& }
139.Ve 205.Ve
140.SH "DESCRIPTION" 206.SH "DESCRIPTION"
141.IX Header "DESCRIPTION" 207.IX Header "DESCRIPTION"
208The newest version of this document is also available as an html-formatted
209web page you might find easier to navigate when reading it for the first
210time: <http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
211.PP
142Libev is an event loop: you register interest in certain events (such as a 212Libev is an event loop: you register interest in certain events (such as a
143file descriptor being readable or a timeout occuring), and it will manage 213file descriptor being readable or a timeout occurring), and it will manage
144these event sources and provide your program with events. 214these event sources and provide your program with events.
145.PP 215.PP
146To do this, it must take more or less complete control over your process 216To do this, it must take more or less complete control over your process
147(or thread) by executing the \fIevent loop\fR handler, and will then 217(or thread) by executing the \fIevent loop\fR handler, and will then
148communicate events via a callback mechanism. 218communicate events via a callback mechanism.
149.PP 219.PP
150You register interest in certain events by registering so-called \fIevent 220You register interest in certain events by registering so-called \fIevent
151watchers\fR, which are relatively small C structures you initialise with the 221watchers\fR, which are relatively small C structures you initialise with the
152details of the event, and then hand it over to libev by \fIstarting\fR the 222details of the event, and then hand it over to libev by \fIstarting\fR the
153watcher. 223watcher.
154.SH "FEATURES" 224.Sh "\s-1FEATURES\s0"
155.IX Header "FEATURES" 225.IX Subsection "FEATURES"
156Libev supports select, poll, the linux-specific epoll and the bsd-specific 226Libev supports \f(CW\*(C`select\*(C'\fR, \f(CW\*(C`poll\*(C'\fR, the Linux-specific \f(CW\*(C`epoll\*(C'\fR, the
157kqueue mechanisms for file descriptor events, relative timers, absolute 227BSD-specific \f(CW\*(C`kqueue\*(C'\fR and the Solaris-specific event port mechanisms
158timers with customised rescheduling, signal events, process status change 228for file descriptor events (\f(CW\*(C`ev_io\*(C'\fR), the Linux \f(CW\*(C`inotify\*(C'\fR interface
159events (related to \s-1SIGCHLD\s0), and event watchers dealing with the event 229(for \f(CW\*(C`ev_stat\*(C'\fR), relative timers (\f(CW\*(C`ev_timer\*(C'\fR), absolute timers
160loop mechanism itself (idle, prepare and check watchers). It also is quite 230with customised rescheduling (\f(CW\*(C`ev_periodic\*(C'\fR), synchronous signals
161fast (see this benchmark comparing 231(\f(CW\*(C`ev_signal\*(C'\fR), process status change events (\f(CW\*(C`ev_child\*(C'\fR), and event
162it to libevent for example). 232watchers dealing with the event loop mechanism itself (\f(CW\*(C`ev_idle\*(C'\fR,
233\&\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
234file watchers (\f(CW\*(C`ev_stat\*(C'\fR) and even limited support for fork events
235(\f(CW\*(C`ev_fork\*(C'\fR).
236.PP
237It also is quite fast (see this
238benchmark comparing it to libevent
239for example).
163.SH "CONVENTIONS" 240.Sh "\s-1CONVENTIONS\s0"
164.IX Header "CONVENTIONS" 241.IX Subsection "CONVENTIONS"
165Libev is very configurable. In this manual the default configuration 242Libev is very configurable. In this manual the default (and most common)
166will be described, which supports multiple event loops. For more info 243configuration will be described, which supports multiple event loops. For
167about various configuration options please have a look at the file 244more info about various configuration options please have a look at
168\&\fI\s-1README\s0.embed\fR in the libev distribution. If libev was configured without 245\&\fB\s-1EMBED\s0\fR section in this manual. If libev was configured without support
169support for multiple event loops, then all functions taking an initial 246for multiple event loops, then all functions taking an initial argument of
170argument of name \f(CW\*(C`loop\*(C'\fR (which is always of type \f(CW\*(C`struct ev_loop *\*(C'\fR) 247name \f(CW\*(C`loop\*(C'\fR (which is always of type \f(CW\*(C`ev_loop *\*(C'\fR) will not have
171will not have this argument. 248this argument.
172.SH "TIME REPRESENTATION" 249.Sh "\s-1TIME\s0 \s-1REPRESENTATION\s0"
173.IX Header "TIME REPRESENTATION" 250.IX Subsection "TIME REPRESENTATION"
174Libev represents time as a single floating point number, representing the 251Libev represents time as a single floating point number, representing the
175(fractional) number of seconds since the (\s-1POSIX\s0) epoch (somewhere near 252(fractional) number of seconds since the (\s-1POSIX\s0) epoch (somewhere near
176the beginning of 1970, details are complicated, don't ask). This type is 253the beginning of 1970, details are complicated, don't ask). This type is
177called \f(CW\*(C`ev_tstamp\*(C'\fR, which is what you should use too. It usually aliases 254called \f(CW\*(C`ev_tstamp\*(C'\fR, which is what you should use too. It usually aliases
178to the \f(CW\*(C`double\*(C'\fR type in C, and when you need to do any calculations on 255to the \f(CW\*(C`double\*(C'\fR type in C, and when you need to do any calculations on
179it, you should treat it as such. 256it, you should treat it as some floating point value. Unlike the name
257component \f(CW\*(C`stamp\*(C'\fR might indicate, it is also used for time differences
258throughout libev.
259.SH "ERROR HANDLING"
260.IX Header "ERROR HANDLING"
261Libev knows three classes of errors: operating system errors, usage errors
262and internal errors (bugs).
263.PP
264When libev catches an operating system error it cannot handle (for example
265a system call indicating a condition libev cannot fix), it calls the callback
266set via \f(CW\*(C`ev_set_syserr_cb\*(C'\fR, which is supposed to fix the problem or
267abort. The default is to print a diagnostic message and to call \f(CW\*(C`abort
268()\*(C'\fR.
269.PP
270When libev detects a usage error such as a negative timer interval, then
271it will print a diagnostic message and abort (via the \f(CW\*(C`assert\*(C'\fR mechanism,
272so \f(CW\*(C`NDEBUG\*(C'\fR will disable this checking): these are programming errors in
273the libev caller and need to be fixed there.
274.PP
275Libev also has a few internal error-checking \f(CW\*(C`assert\*(C'\fRions, and also has
276extensive consistency checking code. These do not trigger under normal
277circumstances, as they indicate either a bug in libev or worse.
180.SH "GLOBAL FUNCTIONS" 278.SH "GLOBAL FUNCTIONS"
181.IX Header "GLOBAL FUNCTIONS" 279.IX Header "GLOBAL FUNCTIONS"
182These functions can be called anytime, even before initialising the 280These functions can be called anytime, even before initialising the
183library in any way. 281library in any way.
184.IP "ev_tstamp ev_time ()" 4 282.IP "ev_tstamp ev_time ()" 4
185.IX Item "ev_tstamp ev_time ()" 283.IX Item "ev_tstamp ev_time ()"
186Returns the current time as libev would use it. Please note that the 284Returns the current time as libev would use it. Please note that the
187\&\f(CW\*(C`ev_now\*(C'\fR function is usually faster and also often returns the timestamp 285\&\f(CW\*(C`ev_now\*(C'\fR function is usually faster and also often returns the timestamp
188you actually want to know. 286you actually want to know.
287.IP "ev_sleep (ev_tstamp interval)" 4
288.IX Item "ev_sleep (ev_tstamp interval)"
289Sleep for the given interval: The current thread will be blocked until
290either it is interrupted or the given time interval has passed. Basically
291this is a sub-second-resolution \f(CW\*(C`sleep ()\*(C'\fR.
189.IP "int ev_version_major ()" 4 292.IP "int ev_version_major ()" 4
190.IX Item "int ev_version_major ()" 293.IX Item "int ev_version_major ()"
191.PD 0 294.PD 0
192.IP "int ev_version_minor ()" 4 295.IP "int ev_version_minor ()" 4
193.IX Item "int ev_version_minor ()" 296.IX Item "int ev_version_minor ()"
194.PD 297.PD
195You can find out the major and minor version numbers of the library 298You can find out the major and minor \s-1ABI\s0 version numbers of the library
196you linked against by calling the functions \f(CW\*(C`ev_version_major\*(C'\fR and 299you linked against by calling the functions \f(CW\*(C`ev_version_major\*(C'\fR and
197\&\f(CW\*(C`ev_version_minor\*(C'\fR. If you want, you can compare against the global 300\&\f(CW\*(C`ev_version_minor\*(C'\fR. If you want, you can compare against the global
198symbols \f(CW\*(C`EV_VERSION_MAJOR\*(C'\fR and \f(CW\*(C`EV_VERSION_MINOR\*(C'\fR, which specify the 301symbols \f(CW\*(C`EV_VERSION_MAJOR\*(C'\fR and \f(CW\*(C`EV_VERSION_MINOR\*(C'\fR, which specify the
199version of the library your program was compiled against. 302version of the library your program was compiled against.
200.Sp 303.Sp
304These version numbers refer to the \s-1ABI\s0 version of the library, not the
305release version.
306.Sp
201Usually, it's a good idea to terminate if the major versions mismatch, 307Usually, it's a good idea to terminate if the major versions mismatch,
202as this indicates an incompatible change. Minor versions are usually 308as this indicates an incompatible change. Minor versions are usually
203compatible to older versions, so a larger minor version alone is usually 309compatible to older versions, so a larger minor version alone is usually
204not a problem. 310not a problem.
205.Sp 311.Sp
206Example: make sure we haven't accidentally been linked against the wrong 312Example: Make sure we haven't accidentally been linked against the wrong
207version: 313version.
208.Sp 314.Sp
209.Vb 3 315.Vb 3
210\& assert (("libev version mismatch", 316\& assert (("libev version mismatch",
211\& ev_version_major () == EV_VERSION_MAJOR 317\& ev_version_major () == EV_VERSION_MAJOR
212\& && ev_version_minor () >= EV_VERSION_MINOR)); 318\& && ev_version_minor () >= EV_VERSION_MINOR));
213.Ve 319.Ve
214.IP "unsigned int ev_supported_backends ()" 4 320.IP "unsigned int ev_supported_backends ()" 4
215.IX Item "unsigned int ev_supported_backends ()" 321.IX Item "unsigned int ev_supported_backends ()"
216Return the set of all backends (i.e. their corresponding \f(CW\*(C`EV_BACKEND_*\*(C'\fR 322Return the set of all backends (i.e. their corresponding \f(CW\*(C`EV_BACKEND_*\*(C'\fR
217value) compiled into this binary of libev (independent of their 323value) compiled into this binary of libev (independent of their
… …
220.Sp 326.Sp
221Example: make sure we have the epoll method, because yeah this is cool and 327Example: make sure we have the epoll method, because yeah this is cool and
222a must have and can we have a torrent of it please!!!11 328a must have and can we have a torrent of it please!!!11
223.Sp 329.Sp
224.Vb 2 330.Vb 2
225\& assert (("sorry, no epoll, no sex", 331\& assert (("sorry, no epoll, no sex",
226\& ev_supported_backends () & EVBACKEND_EPOLL)); 332\& ev_supported_backends () & EVBACKEND_EPOLL));
227.Ve 333.Ve
228.IP "unsigned int ev_recommended_backends ()" 4 334.IP "unsigned int ev_recommended_backends ()" 4
229.IX Item "unsigned int ev_recommended_backends ()" 335.IX Item "unsigned int ev_recommended_backends ()"
230Return the set of all backends compiled into this binary of libev and also 336Return the set of all backends compiled into this binary of libev and also
231recommended for this platform. This set is often smaller than the one 337recommended for this platform. This set is often smaller than the one
232returned by \f(CW\*(C`ev_supported_backends\*(C'\fR, as for example kqueue is broken on 338returned by \f(CW\*(C`ev_supported_backends\*(C'\fR, as for example kqueue is broken on
233most BSDs and will not be autodetected unless you explicitly request it 339most BSDs and will not be auto-detected unless you explicitly request it
234(assuming you know what you are doing). This is the set of backends that 340(assuming you know what you are doing). This is the set of backends that
235libev will probe for if you specify no backends explicitly. 341libev will probe for if you specify no backends explicitly.
236.IP "unsigned int ev_embeddable_backends ()" 4 342.IP "unsigned int ev_embeddable_backends ()" 4
237.IX Item "unsigned int ev_embeddable_backends ()" 343.IX Item "unsigned int ev_embeddable_backends ()"
238Returns the set of backends that are embeddable in other event loops. This 344Returns the set of backends that are embeddable in other event loops. This
239is the theoretical, all\-platform, value. To find which backends 345is the theoretical, all-platform, value. To find which backends
240might be supported on the current system, you would need to look at 346might be supported on the current system, you would need to look at
241\&\f(CW\*(C`ev_embeddable_backends () & ev_supported_backends ()\*(C'\fR, likewise for 347\&\f(CW\*(C`ev_embeddable_backends () & ev_supported_backends ()\*(C'\fR, likewise for
242recommended ones. 348recommended ones.
243.Sp 349.Sp
244See the description of \f(CW\*(C`ev_embed\*(C'\fR watchers for more info. 350See the description of \f(CW\*(C`ev_embed\*(C'\fR watchers for more info.
245.IP "ev_set_allocator (void *(*cb)(void *ptr, long size))" 4 351.IP "ev_set_allocator (void *(*cb)(void *ptr, long size)) [\s-1NOT\s0 \s-1REENTRANT\s0]" 4
246.IX Item "ev_set_allocator (void *(*cb)(void *ptr, long size))" 352.IX Item "ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]"
247Sets the allocation function to use (the prototype is similar to the 353Sets the allocation function to use (the prototype is similar \- the
248realloc C function, the semantics are identical). It is used to allocate 354semantics are identical to the \f(CW\*(C`realloc\*(C'\fR C89/SuS/POSIX function). It is
249and free memory (no surprises here). If it returns zero when memory 355used to allocate and free memory (no surprises here). If it returns zero
250needs to be allocated, the library might abort or take some potentially 356when memory needs to be allocated (\f(CW\*(C`size != 0\*(C'\fR), the library might abort
251destructive action. The default is your system realloc function. 357or take some potentially destructive action.
358.Sp
359Since some systems (at least OpenBSD and Darwin) fail to implement
360correct \f(CW\*(C`realloc\*(C'\fR semantics, libev will use a wrapper around the system
361\&\f(CW\*(C`realloc\*(C'\fR and \f(CW\*(C`free\*(C'\fR functions by default.
252.Sp 362.Sp
253You could override this function in high-availability programs to, say, 363You could override this function in high-availability programs to, say,
254free some memory if it cannot allocate memory, to use a special allocator, 364free some memory if it cannot allocate memory, to use a special allocator,
255or even to sleep a while and retry until some memory is available. 365or even to sleep a while and retry until some memory is available.
256.Sp 366.Sp
257Example: replace the libev allocator with one that waits a bit and then 367Example: Replace the libev allocator with one that waits a bit and then
258retries: better than mine). 368retries (example requires a standards-compliant \f(CW\*(C`realloc\*(C'\fR).
259.Sp 369.Sp
260.Vb 6 370.Vb 6
261\& static void * 371\& static void *
262\& persistent_realloc (void *ptr, long size) 372\& persistent_realloc (void *ptr, size_t size)
263\& { 373\& {
264\& for (;;) 374\& for (;;)
265\& { 375\& {
266\& void *newptr = realloc (ptr, size); 376\& void *newptr = realloc (ptr, size);
267.Ve 377\&
268.Sp
269.Vb 2
270\& if (newptr) 378\& if (newptr)
271\& return newptr; 379\& return newptr;
272.Ve 380\&
273.Sp
274.Vb 3
275\& sleep (60); 381\& sleep (60);
276\& } 382\& }
277\& } 383\& }
278.Ve 384\&
279.Sp
280.Vb 2
281\& ... 385\& ...
282\& ev_set_allocator (persistent_realloc); 386\& ev_set_allocator (persistent_realloc);
283.Ve 387.Ve
284.IP "ev_set_syserr_cb (void (*cb)(const char *msg));" 4 388.IP "ev_set_syserr_cb (void (*cb)(const char *msg)); [\s-1NOT\s0 \s-1REENTRANT\s0]" 4
285.IX Item "ev_set_syserr_cb (void (*cb)(const char *msg));" 389.IX Item "ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]"
286Set the callback function to call on a retryable syscall error (such 390Set the callback function to call on a retryable system call error (such
287as failed select, poll, epoll_wait). The message is a printable string 391as failed select, poll, epoll_wait). The message is a printable string
288indicating the system call or subsystem causing the problem. If this 392indicating the system call or subsystem causing the problem. If this
289callback is set, then libev will expect it to remedy the sitution, no 393callback is set, then libev will expect it to remedy the situation, no
290matter what, when it returns. That is, libev will generally retry the 394matter what, when it returns. That is, libev will generally retry the
291requested operation, or, if the condition doesn't go away, do bad stuff 395requested operation, or, if the condition doesn't go away, do bad stuff
292(such as abort). 396(such as abort).
293.Sp 397.Sp
294Example: do the same thing as libev does internally: 398Example: This is basically the same thing that libev does internally, too.
295.Sp 399.Sp
296.Vb 6 400.Vb 6
297\& static void 401\& static void
298\& fatal_error (const char *msg) 402\& fatal_error (const char *msg)
299\& { 403\& {
300\& perror (msg); 404\& perror (msg);
301\& abort (); 405\& abort ();
302\& } 406\& }
303.Ve 407\&
304.Sp
305.Vb 2
306\& ... 408\& ...
307\& ev_set_syserr_cb (fatal_error); 409\& ev_set_syserr_cb (fatal_error);
308.Ve 410.Ve
309.SH "FUNCTIONS CONTROLLING THE EVENT LOOP" 411.SH "FUNCTIONS CONTROLLING THE EVENT LOOP"
310.IX Header "FUNCTIONS CONTROLLING THE EVENT LOOP" 412.IX Header "FUNCTIONS CONTROLLING THE EVENT LOOP"
311An event loop is described by a \f(CW\*(C`struct ev_loop *\*(C'\fR. The library knows two 413An event loop is described by a \f(CW\*(C`struct ev_loop *\*(C'\fR (the \f(CW\*(C`struct\*(C'\fR
312types of such loops, the \fIdefault\fR loop, which supports signals and child 414is \fInot\fR optional in this case, as there is also an \f(CW\*(C`ev_loop\*(C'\fR
313events, and dynamically created loops which do not. 415\&\fIfunction\fR).
314.PP 416.PP
315If you use threads, a common model is to run the default event loop 417The library knows two types of such loops, the \fIdefault\fR loop, which
316in your main thread (or in a separate thread) and for each thread you 418supports signals and child events, and dynamically created loops which do
317create, you also create another event loop. Libev itself does no locking 419not.
318whatsoever, so if you mix calls to the same event loop in different
319threads, make sure you lock (this is usually a bad idea, though, even if
320done correctly, because it's hideous and inefficient).
321.IP "struct ev_loop *ev_default_loop (unsigned int flags)" 4 420.IP "struct ev_loop *ev_default_loop (unsigned int flags)" 4
322.IX Item "struct ev_loop *ev_default_loop (unsigned int flags)" 421.IX Item "struct ev_loop *ev_default_loop (unsigned int flags)"
323This will initialise the default event loop if it hasn't been initialised 422This will initialise the default event loop if it hasn't been initialised
324yet and return it. If the default loop could not be initialised, returns 423yet and return it. If the default loop could not be initialised, returns
325false. If it already was initialised it simply returns it (and ignores the 424false. If it already was initialised it simply returns it (and ignores the
326flags. If that is troubling you, check \f(CW\*(C`ev_backend ()\*(C'\fR afterwards). 425flags. If that is troubling you, check \f(CW\*(C`ev_backend ()\*(C'\fR afterwards).
327.Sp 426.Sp
328If you don't know what event loop to use, use the one returned from this 427If you don't know what event loop to use, use the one returned from this
329function. 428function.
429.Sp
430Note that this function is \fInot\fR thread-safe, so if you want to use it
431from multiple threads, you have to lock (note also that this is unlikely,
432as loops cannot be shared easily between threads anyway).
433.Sp
434The default loop is the only loop that can handle \f(CW\*(C`ev_signal\*(C'\fR and
435\&\f(CW\*(C`ev_child\*(C'\fR watchers, and to do this, it always registers a handler
436for \f(CW\*(C`SIGCHLD\*(C'\fR. If this is a problem for your application you can either
437create a dynamic loop with \f(CW\*(C`ev_loop_new\*(C'\fR that doesn't do that, or you
438can simply overwrite the \f(CW\*(C`SIGCHLD\*(C'\fR signal handler \fIafter\fR calling
439\&\f(CW\*(C`ev_default_init\*(C'\fR.
330.Sp 440.Sp
331The flags argument can be used to specify special behaviour or specific 441The flags argument can be used to specify special behaviour or specific
332backends to use, and is usually specified as \f(CW0\fR (or \f(CW\*(C`EVFLAG_AUTO\*(C'\fR). 442backends to use, and is usually specified as \f(CW0\fR (or \f(CW\*(C`EVFLAG_AUTO\*(C'\fR).
333.Sp 443.Sp
334The following flags are supported: 444The following flags are supported:
… …
339The default flags value. Use this if you have no clue (it's the right 449The default flags value. Use this if you have no clue (it's the right
340thing, believe me). 450thing, believe me).
341.ie n .IP """EVFLAG_NOENV""" 4 451.ie n .IP """EVFLAG_NOENV""" 4
342.el .IP "\f(CWEVFLAG_NOENV\fR" 4 452.el .IP "\f(CWEVFLAG_NOENV\fR" 4
343.IX Item "EVFLAG_NOENV" 453.IX Item "EVFLAG_NOENV"
344If this flag bit is ored into the flag value (or the program runs setuid 454If this flag bit is or'ed into the flag value (or the program runs setuid
345or setgid) then libev will \fInot\fR look at the environment variable 455or setgid) then libev will \fInot\fR look at the environment variable
346\&\f(CW\*(C`LIBEV_FLAGS\*(C'\fR. Otherwise (the default), this environment variable will 456\&\f(CW\*(C`LIBEV_FLAGS\*(C'\fR. Otherwise (the default), this environment variable will
347override the flags completely if it is found in the environment. This is 457override the flags completely if it is found in the environment. This is
348useful to try out specific backends to test their performance, or to work 458useful to try out specific backends to test their performance, or to work
349around bugs. 459around bugs.
460.ie n .IP """EVFLAG_FORKCHECK""" 4
461.el .IP "\f(CWEVFLAG_FORKCHECK\fR" 4
462.IX Item "EVFLAG_FORKCHECK"
463Instead of calling \f(CW\*(C`ev_default_fork\*(C'\fR or \f(CW\*(C`ev_loop_fork\*(C'\fR manually after
464a fork, you can also make libev check for a fork in each iteration by
465enabling this flag.
466.Sp
467This works by calling \f(CW\*(C`getpid ()\*(C'\fR on every iteration of the loop,
468and thus this might slow down your event loop if you do a lot of loop
469iterations and little real work, but is usually not noticeable (on my
470GNU/Linux system for example, \f(CW\*(C`getpid\*(C'\fR is actually a simple 5\-insn sequence
471without a system call and thus \fIvery\fR fast, but my GNU/Linux system also has
472\&\f(CW\*(C`pthread_atfork\*(C'\fR which is even faster).
473.Sp
474The big advantage of this flag is that you can forget about fork (and
475forget about forgetting to tell libev about forking) when you use this
476flag.
477.Sp
478This flag setting cannot be overridden or specified in the \f(CW\*(C`LIBEV_FLAGS\*(C'\fR
479environment variable.
350.ie n .IP """EVBACKEND_SELECT"" (value 1, portable select backend)" 4 480.ie n .IP """EVBACKEND_SELECT"" (value 1, portable select backend)" 4
351.el .IP "\f(CWEVBACKEND_SELECT\fR (value 1, portable select backend)" 4 481.el .IP "\f(CWEVBACKEND_SELECT\fR (value 1, portable select backend)" 4
352.IX Item "EVBACKEND_SELECT (value 1, portable select backend)" 482.IX Item "EVBACKEND_SELECT (value 1, portable select backend)"
353This is your standard \fIselect\fR\|(2) backend. Not \fIcompletely\fR standard, as 483This is your standard \fIselect\fR\|(2) backend. Not \fIcompletely\fR standard, as
354libev tries to roll its own fd_set with no limits on the number of fds, 484libev tries to roll its own fd_set with no limits on the number of fds,
355but if that fails, expect a fairly low limit on the number of fds when 485but if that fails, expect a fairly low limit on the number of fds when
356using this backend. It doesn't scale too well (O(highest_fd)), but its usually 486using this backend. It doesn't scale too well (O(highest_fd)), but its
357the fastest backend for a low number of fds. 487usually the fastest backend for a low number of (low-numbered :) fds.
488.Sp
489To get good performance out of this backend you need a high amount of
490parallelism (most of the file descriptors should be busy). If you are
491writing a server, you should \f(CW\*(C`accept ()\*(C'\fR in a loop to accept as many
492connections as possible during one iteration. You might also want to have
493a look at \f(CW\*(C`ev_set_io_collect_interval ()\*(C'\fR to increase the amount of
494readiness notifications you get per iteration.
495.Sp
496This backend maps \f(CW\*(C`EV_READ\*(C'\fR to the \f(CW\*(C`readfds\*(C'\fR set and \f(CW\*(C`EV_WRITE\*(C'\fR to the
497\&\f(CW\*(C`writefds\*(C'\fR set (and to work around Microsoft Windows bugs, also onto the
498\&\f(CW\*(C`exceptfds\*(C'\fR set on that platform).
358.ie n .IP """EVBACKEND_POLL"" (value 2, poll backend, available everywhere except on windows)" 4 499.ie n .IP """EVBACKEND_POLL"" (value 2, poll backend, available everywhere except on windows)" 4
359.el .IP "\f(CWEVBACKEND_POLL\fR (value 2, poll backend, available everywhere except on windows)" 4 500.el .IP "\f(CWEVBACKEND_POLL\fR (value 2, poll backend, available everywhere except on windows)" 4
360.IX Item "EVBACKEND_POLL (value 2, poll backend, available everywhere except on windows)" 501.IX Item "EVBACKEND_POLL (value 2, poll backend, available everywhere except on windows)"
361And this is your standard \fIpoll\fR\|(2) backend. It's more complicated than 502And this is your standard \fIpoll\fR\|(2) backend. It's more complicated
362select, but handles sparse fds better and has no artificial limit on the 503than select, but handles sparse fds better and has no artificial
363number of fds you can use (except it will slow down considerably with a 504limit on the number of fds you can use (except it will slow down
364lot of inactive fds). It scales similarly to select, i.e. O(total_fds). 505considerably with a lot of inactive fds). It scales similarly to select,
506i.e. O(total_fds). See the entry for \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR, above, for
507performance tips.
508.Sp
509This backend maps \f(CW\*(C`EV_READ\*(C'\fR to \f(CW\*(C`POLLIN | POLLERR | POLLHUP\*(C'\fR, and
510\&\f(CW\*(C`EV_WRITE\*(C'\fR to \f(CW\*(C`POLLOUT | POLLERR | POLLHUP\*(C'\fR.
365.ie n .IP """EVBACKEND_EPOLL"" (value 4, Linux)" 4 511.ie n .IP """EVBACKEND_EPOLL"" (value 4, Linux)" 4
366.el .IP "\f(CWEVBACKEND_EPOLL\fR (value 4, Linux)" 4 512.el .IP "\f(CWEVBACKEND_EPOLL\fR (value 4, Linux)" 4
367.IX Item "EVBACKEND_EPOLL (value 4, Linux)" 513.IX Item "EVBACKEND_EPOLL (value 4, Linux)"
368For few fds, this backend is a bit little slower than poll and select, 514For few fds, this backend is a bit little slower than poll and select,
369but it scales phenomenally better. While poll and select usually scale like 515but it scales phenomenally better. While poll and select usually scale
370O(total_fds) where n is the total number of fds (or the highest fd), epoll scales 516like O(total_fds) where n is the total number of fds (or the highest fd),
371either O(1) or O(active_fds). 517epoll scales either O(1) or O(active_fds).
372.Sp 518.Sp
519The epoll mechanism deserves honorable mention as the most misdesigned
520of the more advanced event mechanisms: mere annoyances include silently
521dropping file descriptors, requiring a system call per change per file
522descriptor (and unnecessary guessing of parameters), problems with dup and
523so on. The biggest issue is fork races, however \- if a program forks then
524\&\fIboth\fR parent and child process have to recreate the epoll set, which can
525take considerable time (one syscall per file descriptor) and is of course
526hard to detect.
527.Sp
528Epoll is also notoriously buggy \- embedding epoll fds \fIshould\fR work, but
529of course \fIdoesn't\fR, and epoll just loves to report events for totally
530\&\fIdifferent\fR file descriptors (even already closed ones, so one cannot
531even remove them from the set) than registered in the set (especially
532on \s-1SMP\s0 systems). Libev tries to counter these spurious notifications by
533employing an additional generation counter and comparing that against the
534events to filter out spurious ones, recreating the set when required.
535.Sp
373While stopping and starting an I/O watcher in the same iteration will 536While stopping, setting and starting an I/O watcher in the same iteration
374result in some caching, there is still a syscall per such incident 537will result in some caching, there is still a system call per such
375(because the fd could point to a different file description now), so its 538incident (because the same \fIfile descriptor\fR could point to a different
376best to avoid that. Also, \fIdup()\fRed file descriptors might not work very 539\&\fIfile description\fR now), so its best to avoid that. Also, \f(CW\*(C`dup ()\*(C'\fR'ed
377well if you register events for both fds. 540file descriptors might not work very well if you register events for both
541file descriptors.
378.Sp 542.Sp
379Please note that epoll sometimes generates spurious notifications, so you 543Best performance from this backend is achieved by not unregistering all
380need to use non-blocking I/O or other means to avoid blocking when no data 544watchers for a file descriptor until it has been closed, if possible,
381(or space) is available. 545i.e. keep at least one watcher active per fd at all times. Stopping and
546starting a watcher (without re-setting it) also usually doesn't cause
547extra overhead. A fork can both result in spurious notifications as well
548as in libev having to destroy and recreate the epoll object, which can
549take considerable time and thus should be avoided.
550.Sp
551All this means that, in practice, \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR can be as fast or
552faster than epoll for maybe up to a hundred file descriptors, depending on
553the usage. So sad.
554.Sp
555While nominally embeddable in other event loops, this feature is broken in
556all kernel versions tested so far.
557.Sp
558This backend maps \f(CW\*(C`EV_READ\*(C'\fR and \f(CW\*(C`EV_WRITE\*(C'\fR in the same way as
559\&\f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
382.ie n .IP """EVBACKEND_KQUEUE"" (value 8, most \s-1BSD\s0 clones)" 4 560.ie n .IP """EVBACKEND_KQUEUE"" (value 8, most \s-1BSD\s0 clones)" 4
383.el .IP "\f(CWEVBACKEND_KQUEUE\fR (value 8, most \s-1BSD\s0 clones)" 4 561.el .IP "\f(CWEVBACKEND_KQUEUE\fR (value 8, most \s-1BSD\s0 clones)" 4
384.IX Item "EVBACKEND_KQUEUE (value 8, most BSD clones)" 562.IX Item "EVBACKEND_KQUEUE (value 8, most BSD clones)"
385Kqueue deserves special mention, as at the time of this writing, it 563Kqueue deserves special mention, as at the time of this writing, it
386was broken on all BSDs except NetBSD (usually it doesn't work with 564was broken on all BSDs except NetBSD (usually it doesn't work reliably
387anything but sockets and pipes, except on Darwin, where of course its 565with anything but sockets and pipes, except on Darwin, where of course
388completely useless). For this reason its not being \*(L"autodetected\*(R" 566it's completely useless). Unlike epoll, however, whose brokenness
567is by design, these kqueue bugs can (and eventually will) be fixed
568without \s-1API\s0 changes to existing programs. For this reason it's not being
389unless you explicitly specify it explicitly in the flags (i.e. using 569\&\*(L"auto-detected\*(R" unless you explicitly specify it in the flags (i.e. using
390\&\f(CW\*(C`EVBACKEND_KQUEUE\*(C'\fR). 570\&\f(CW\*(C`EVBACKEND_KQUEUE\*(C'\fR) or libev was compiled on a known-to-be-good (\-enough)
571system like NetBSD.
572.Sp
573You still can embed kqueue into a normal poll or select backend and use it
574only for sockets (after having made sure that sockets work with kqueue on
575the target platform). See \f(CW\*(C`ev_embed\*(C'\fR watchers for more info.
391.Sp 576.Sp
392It scales in the same way as the epoll backend, but the interface to the 577It scales in the same way as the epoll backend, but the interface to the
393kernel is more efficient (which says nothing about its actual speed, of 578kernel is more efficient (which says nothing about its actual speed, of
394course). While starting and stopping an I/O watcher does not cause an 579course). While stopping, setting and starting an I/O watcher does never
395extra syscall as with epoll, it still adds up to four event changes per 580cause an extra system call as with \f(CW\*(C`EVBACKEND_EPOLL\*(C'\fR, it still adds up to
396incident, so its best to avoid that. 581two event changes per incident. Support for \f(CW\*(C`fork ()\*(C'\fR is very bad (but
582sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
583cases
584.Sp
585This backend usually performs well under most conditions.
586.Sp
587While nominally embeddable in other event loops, this doesn't work
588everywhere, so you might need to test for this. And since it is broken
589almost everywhere, you should only use it when you have a lot of sockets
590(for which it usually works), by embedding it into another event loop
591(e.g. \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or \f(CW\*(C`EVBACKEND_POLL\*(C'\fR (but \f(CW\*(C`poll\*(C'\fR is of course
592also broken on \s-1OS\s0 X)) and, did I mention it, using it only for sockets.
593.Sp
594This backend maps \f(CW\*(C`EV_READ\*(C'\fR into an \f(CW\*(C`EVFILT_READ\*(C'\fR kevent with
595\&\f(CW\*(C`NOTE_EOF\*(C'\fR, and \f(CW\*(C`EV_WRITE\*(C'\fR into an \f(CW\*(C`EVFILT_WRITE\*(C'\fR kevent with
596\&\f(CW\*(C`NOTE_EOF\*(C'\fR.
397.ie n .IP """EVBACKEND_DEVPOLL"" (value 16, Solaris 8)" 4 597.ie n .IP """EVBACKEND_DEVPOLL"" (value 16, Solaris 8)" 4
398.el .IP "\f(CWEVBACKEND_DEVPOLL\fR (value 16, Solaris 8)" 4 598.el .IP "\f(CWEVBACKEND_DEVPOLL\fR (value 16, Solaris 8)" 4
399.IX Item "EVBACKEND_DEVPOLL (value 16, Solaris 8)" 599.IX Item "EVBACKEND_DEVPOLL (value 16, Solaris 8)"
400This is not implemented yet (and might never be). 600This is not implemented yet (and might never be, unless you send me an
601implementation). According to reports, \f(CW\*(C`/dev/poll\*(C'\fR only supports sockets
602and is not embeddable, which would limit the usefulness of this backend
603immensely.
401.ie n .IP """EVBACKEND_PORT"" (value 32, Solaris 10)" 4 604.ie n .IP """EVBACKEND_PORT"" (value 32, Solaris 10)" 4
402.el .IP "\f(CWEVBACKEND_PORT\fR (value 32, Solaris 10)" 4 605.el .IP "\f(CWEVBACKEND_PORT\fR (value 32, Solaris 10)" 4
403.IX Item "EVBACKEND_PORT (value 32, Solaris 10)" 606.IX Item "EVBACKEND_PORT (value 32, Solaris 10)"
404This uses the Solaris 10 port mechanism. As with everything on Solaris, 607This uses the Solaris 10 event port mechanism. As with everything on Solaris,
405it's really slow, but it still scales very well (O(active_fds)). 608it's really slow, but it still scales very well (O(active_fds)).
406.Sp 609.Sp
407Please note that solaris ports can result in a lot of spurious 610Please note that Solaris event ports can deliver a lot of spurious
408notifications, so you need to use non-blocking I/O or other means to avoid 611notifications, so you need to use non-blocking I/O or other means to avoid
409blocking when no data (or space) is available. 612blocking when no data (or space) is available.
613.Sp
614While this backend scales well, it requires one system call per active
615file descriptor per loop iteration. For small and medium numbers of file
616descriptors a \*(L"slow\*(R" \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or \f(CW\*(C`EVBACKEND_POLL\*(C'\fR backend
617might perform better.
618.Sp
619On the positive side, with the exception of the spurious readiness
620notifications, this backend actually performed fully to specification
621in all tests and is fully embeddable, which is a rare feat among the
622OS-specific backends (I vastly prefer correctness over speed hacks).
623.Sp
624This backend maps \f(CW\*(C`EV_READ\*(C'\fR and \f(CW\*(C`EV_WRITE\*(C'\fR in the same way as
625\&\f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
410.ie n .IP """EVBACKEND_ALL""" 4 626.ie n .IP """EVBACKEND_ALL""" 4
411.el .IP "\f(CWEVBACKEND_ALL\fR" 4 627.el .IP "\f(CWEVBACKEND_ALL\fR" 4
412.IX Item "EVBACKEND_ALL" 628.IX Item "EVBACKEND_ALL"
413Try all backends (even potentially broken ones that wouldn't be tried 629Try all backends (even potentially broken ones that wouldn't be tried
414with \f(CW\*(C`EVFLAG_AUTO\*(C'\fR). Since this is a mask, you can do stuff such as 630with \f(CW\*(C`EVFLAG_AUTO\*(C'\fR). Since this is a mask, you can do stuff such as
415\&\f(CW\*(C`EVBACKEND_ALL & ~EVBACKEND_KQUEUE\*(C'\fR. 631\&\f(CW\*(C`EVBACKEND_ALL & ~EVBACKEND_KQUEUE\*(C'\fR.
632.Sp
633It is definitely not recommended to use this flag.
416.RE 634.RE
417.RS 4 635.RS 4
418.Sp 636.Sp
419If one or more of these are ored into the flags value, then only these 637If one or more of these are or'ed into the flags value, then only these
420backends will be tried (in the reverse order as given here). If none are 638backends will be tried (in the reverse order as listed here). If none are
421specified, most compiled-in backend will be tried, usually in reverse 639specified, all backends in \f(CW\*(C`ev_recommended_backends ()\*(C'\fR will be tried.
422order of their flag values :)
423.Sp 640.Sp
424The most typical usage is like this: 641Example: This is the most typical usage.
425.Sp 642.Sp
426.Vb 2 643.Vb 2
427\& if (!ev_default_loop (0)) 644\& if (!ev_default_loop (0))
428\& fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?"); 645\& fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
429.Ve 646.Ve
430.Sp 647.Sp
431Restrict libev to the select and poll backends, and do not allow 648Example: Restrict libev to the select and poll backends, and do not allow
432environment settings to be taken into account: 649environment settings to be taken into account:
433.Sp 650.Sp
434.Vb 1 651.Vb 1
435\& ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV); 652\& ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
436.Ve 653.Ve
437.Sp 654.Sp
438Use whatever libev has to offer, but make sure that kqueue is used if 655Example: Use whatever libev has to offer, but make sure that kqueue is
439available (warning, breaks stuff, best use only with your own private 656used if available (warning, breaks stuff, best use only with your own
440event loop and only if you know the \s-1OS\s0 supports your types of fds): 657private event loop and only if you know the \s-1OS\s0 supports your types of
658fds):
441.Sp 659.Sp
442.Vb 1 660.Vb 1
443\& ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE); 661\& ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
444.Ve 662.Ve
445.RE 663.RE
446.IP "struct ev_loop *ev_loop_new (unsigned int flags)" 4 664.IP "struct ev_loop *ev_loop_new (unsigned int flags)" 4
447.IX Item "struct ev_loop *ev_loop_new (unsigned int flags)" 665.IX Item "struct ev_loop *ev_loop_new (unsigned int flags)"
448Similar to \f(CW\*(C`ev_default_loop\*(C'\fR, but always creates a new event loop that is 666Similar to \f(CW\*(C`ev_default_loop\*(C'\fR, but always creates a new event loop that is
449always distinct from the default loop. Unlike the default loop, it cannot 667always distinct from the default loop. Unlike the default loop, it cannot
450handle signal and child watchers, and attempts to do so will be greeted by 668handle signal and child watchers, and attempts to do so will be greeted by
451undefined behaviour (or a failed assertion if assertions are enabled). 669undefined behaviour (or a failed assertion if assertions are enabled).
452.Sp 670.Sp
671Note that this function \fIis\fR thread-safe, and the recommended way to use
672libev with threads is indeed to create one loop per thread, and using the
673default loop in the \*(L"main\*(R" or \*(L"initial\*(R" thread.
674.Sp
453Example: try to create a event loop that uses epoll and nothing else. 675Example: Try to create a event loop that uses epoll and nothing else.
454.Sp 676.Sp
455.Vb 3 677.Vb 3
456\& struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV); 678\& struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
457\& if (!epoller) 679\& if (!epoller)
458\& fatal ("no epoll found here, maybe it hides under your chair"); 680\& fatal ("no epoll found here, maybe it hides under your chair");
459.Ve 681.Ve
460.IP "ev_default_destroy ()" 4 682.IP "ev_default_destroy ()" 4
461.IX Item "ev_default_destroy ()" 683.IX Item "ev_default_destroy ()"
462Destroys the default loop again (frees all memory and kernel state 684Destroys the default loop again (frees all memory and kernel state
463etc.). None of the active event watchers will be stopped in the normal 685etc.). None of the active event watchers will be stopped in the normal
464sense, so e.g. \f(CW\*(C`ev_is_active\*(C'\fR might still return true. It is your 686sense, so e.g. \f(CW\*(C`ev_is_active\*(C'\fR might still return true. It is your
465responsibility to either stop all watchers cleanly yoursef \fIbefore\fR 687responsibility to either stop all watchers cleanly yourself \fIbefore\fR
466calling this function, or cope with the fact afterwards (which is usually 688calling this function, or cope with the fact afterwards (which is usually
467the easiest thing, youc na just ignore the watchers and/or \f(CW\*(C`free ()\*(C'\fR them 689the easiest thing, you can just ignore the watchers and/or \f(CW\*(C`free ()\*(C'\fR them
468for example). 690for example).
691.Sp
692Note that certain global state, such as signal state (and installed signal
693handlers), will not be freed by this function, and related watchers (such
694as signal and child watchers) would need to be stopped manually.
695.Sp
696In general it is not advisable to call this function except in the
697rare occasion where you really need to free e.g. the signal handling
698pipe fds. If you need dynamically allocated loops it is better to use
699\&\f(CW\*(C`ev_loop_new\*(C'\fR and \f(CW\*(C`ev_loop_destroy\*(C'\fR).
469.IP "ev_loop_destroy (loop)" 4 700.IP "ev_loop_destroy (loop)" 4
470.IX Item "ev_loop_destroy (loop)" 701.IX Item "ev_loop_destroy (loop)"
471Like \f(CW\*(C`ev_default_destroy\*(C'\fR, but destroys an event loop created by an 702Like \f(CW\*(C`ev_default_destroy\*(C'\fR, but destroys an event loop created by an
472earlier call to \f(CW\*(C`ev_loop_new\*(C'\fR. 703earlier call to \f(CW\*(C`ev_loop_new\*(C'\fR.
473.IP "ev_default_fork ()" 4 704.IP "ev_default_fork ()" 4
474.IX Item "ev_default_fork ()" 705.IX Item "ev_default_fork ()"
706This function sets a flag that causes subsequent \f(CW\*(C`ev_loop\*(C'\fR iterations
475This function reinitialises the kernel state for backends that have 707to reinitialise the kernel state for backends that have one. Despite the
476one. Despite the name, you can call it anytime, but it makes most sense 708name, you can call it anytime, but it makes most sense after forking, in
477after forking, in either the parent or child process (or both, but that 709the child process (or both child and parent, but that again makes little
478again makes little sense). 710sense). You \fImust\fR call it in the child before using any of the libev
711functions, and it will only take effect at the next \f(CW\*(C`ev_loop\*(C'\fR iteration.
479.Sp 712.Sp
480You \fImust\fR call this function in the child process after forking if and 713On the other hand, you only need to call this function in the child
481only if you want to use the event library in both processes. If you just 714process if and only if you want to use the event library in the child. If
482fork+exec, you don't have to call it. 715you just fork+exec, you don't have to call it at all.
483.Sp 716.Sp
484The function itself is quite fast and it's usually not a problem to call 717The function itself is quite fast and it's usually not a problem to call
485it just in case after a fork. To make this easy, the function will fit in 718it just in case after a fork. To make this easy, the function will fit in
486quite nicely into a call to \f(CW\*(C`pthread_atfork\*(C'\fR: 719quite nicely into a call to \f(CW\*(C`pthread_atfork\*(C'\fR:
487.Sp 720.Sp
488.Vb 1 721.Vb 1
489\& pthread_atfork (0, 0, ev_default_fork); 722\& pthread_atfork (0, 0, ev_default_fork);
490.Ve 723.Ve
491.Sp
492At the moment, \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR and \f(CW\*(C`EVBACKEND_POLL\*(C'\fR are safe to use
493without calling this function, so if you force one of those backends you
494do not need to care.
495.IP "ev_loop_fork (loop)" 4 724.IP "ev_loop_fork (loop)" 4
496.IX Item "ev_loop_fork (loop)" 725.IX Item "ev_loop_fork (loop)"
497Like \f(CW\*(C`ev_default_fork\*(C'\fR, but acts on an event loop created by 726Like \f(CW\*(C`ev_default_fork\*(C'\fR, but acts on an event loop created by
498\&\f(CW\*(C`ev_loop_new\*(C'\fR. Yes, you have to call this on every allocated event loop 727\&\f(CW\*(C`ev_loop_new\*(C'\fR. Yes, you have to call this on every allocated event loop
499after fork, and how you do this is entirely your own problem. 728after fork that you want to re-use in the child, and how you do this is
729entirely your own problem.
730.IP "int ev_is_default_loop (loop)" 4
731.IX Item "int ev_is_default_loop (loop)"
732Returns true when the given loop is, in fact, the default loop, and false
733otherwise.
734.IP "unsigned int ev_loop_count (loop)" 4
735.IX Item "unsigned int ev_loop_count (loop)"
736Returns the count of loop iterations for the loop, which is identical to
737the number of times libev did poll for new events. It starts at \f(CW0\fR and
738happily wraps around with enough iterations.
739.Sp
740This value can sometimes be useful as a generation counter of sorts (it
741\&\*(L"ticks\*(R" the number of loop iterations), as it roughly corresponds with
742\&\f(CW\*(C`ev_prepare\*(C'\fR and \f(CW\*(C`ev_check\*(C'\fR calls.
500.IP "unsigned int ev_backend (loop)" 4 743.IP "unsigned int ev_backend (loop)" 4
501.IX Item "unsigned int ev_backend (loop)" 744.IX Item "unsigned int ev_backend (loop)"
502Returns one of the \f(CW\*(C`EVBACKEND_*\*(C'\fR flags indicating the event backend in 745Returns one of the \f(CW\*(C`EVBACKEND_*\*(C'\fR flags indicating the event backend in
503use. 746use.
504.IP "ev_tstamp ev_now (loop)" 4 747.IP "ev_tstamp ev_now (loop)" 4
505.IX Item "ev_tstamp ev_now (loop)" 748.IX Item "ev_tstamp ev_now (loop)"
506Returns the current \*(L"event loop time\*(R", which is the time the event loop 749Returns the current \*(L"event loop time\*(R", which is the time the event loop
507received events and started processing them. This timestamp does not 750received events and started processing them. This timestamp does not
508change as long as callbacks are being processed, and this is also the base 751change as long as callbacks are being processed, and this is also the base
509time used for relative timers. You can treat it as the timestamp of the 752time used for relative timers. You can treat it as the timestamp of the
510event occuring (or more correctly, libev finding out about it). 753event occurring (or more correctly, libev finding out about it).
754.IP "ev_now_update (loop)" 4
755.IX Item "ev_now_update (loop)"
756Establishes the current time by querying the kernel, updating the time
757returned by \f(CW\*(C`ev_now ()\*(C'\fR in the progress. This is a costly operation and
758is usually done automatically within \f(CW\*(C`ev_loop ()\*(C'\fR.
759.Sp
760This function is rarely useful, but when some event callback runs for a
761very long time without entering the event loop, updating libev's idea of
762the current time is a good idea.
763.Sp
764See also \*(L"The special problem of time updates\*(R" in the \f(CW\*(C`ev_timer\*(C'\fR section.
511.IP "ev_loop (loop, int flags)" 4 765.IP "ev_loop (loop, int flags)" 4
512.IX Item "ev_loop (loop, int flags)" 766.IX Item "ev_loop (loop, int flags)"
513Finally, this is it, the event handler. This function usually is called 767Finally, this is it, the event handler. This function usually is called
514after you initialised all your watchers and you want to start handling 768after you initialised all your watchers and you want to start handling
515events. 769events.
… …
517If the flags argument is specified as \f(CW0\fR, it will not return until 771If the flags argument is specified as \f(CW0\fR, it will not return until
518either no event watchers are active anymore or \f(CW\*(C`ev_unloop\*(C'\fR was called. 772either no event watchers are active anymore or \f(CW\*(C`ev_unloop\*(C'\fR was called.
519.Sp 773.Sp
520Please note that an explicit \f(CW\*(C`ev_unloop\*(C'\fR is usually better than 774Please note that an explicit \f(CW\*(C`ev_unloop\*(C'\fR is usually better than
521relying on all watchers to be stopped when deciding when a program has 775relying on all watchers to be stopped when deciding when a program has
522finished (especially in interactive programs), but having a program that 776finished (especially in interactive programs), but having a program
523automatically loops as long as it has to and no longer by virtue of 777that automatically loops as long as it has to and no longer by virtue
524relying on its watchers stopping correctly is a thing of beauty. 778of relying on its watchers stopping correctly, that is truly a thing of
779beauty.
525.Sp 780.Sp
526A flags value of \f(CW\*(C`EVLOOP_NONBLOCK\*(C'\fR will look for new events, will handle 781A flags value of \f(CW\*(C`EVLOOP_NONBLOCK\*(C'\fR will look for new events, will handle
527those events and any outstanding ones, but will not block your process in 782those events and any already outstanding ones, but will not block your
528case there are no events and will return after one iteration of the loop. 783process in case there are no events and will return after one iteration of
784the loop.
529.Sp 785.Sp
530A flags value of \f(CW\*(C`EVLOOP_ONESHOT\*(C'\fR will look for new events (waiting if 786A flags value of \f(CW\*(C`EVLOOP_ONESHOT\*(C'\fR will look for new events (waiting if
531neccessary) and will handle those and any outstanding ones. It will block 787necessary) and will handle those and any already outstanding ones. It
532your process until at least one new event arrives, and will return after 788will block your process until at least one new event arrives (which could
533one iteration of the loop. This is useful if you are waiting for some 789be an event internal to libev itself, so there is no guarantee that a
534external event in conjunction with something not expressible using other 790user-registered callback will be called), and will return after one
791iteration of the loop.
792.Sp
793This is useful if you are waiting for some external event in conjunction
794with something not expressible using other libev watchers (i.e. "roll your
535libev watchers. However, a pair of \f(CW\*(C`ev_prepare\*(C'\fR/\f(CW\*(C`ev_check\*(C'\fR watchers is 795own \f(CW\*(C`ev_loop\*(C'\fR"). However, a pair of \f(CW\*(C`ev_prepare\*(C'\fR/\f(CW\*(C`ev_check\*(C'\fR watchers is
536usually a better approach for this kind of thing. 796usually a better approach for this kind of thing.
537.Sp 797.Sp
538Here are the gory details of what \f(CW\*(C`ev_loop\*(C'\fR does: 798Here are the gory details of what \f(CW\*(C`ev_loop\*(C'\fR does:
539.Sp 799.Sp
540.Vb 18 800.Vb 10
541\& * If there are no active watchers (reference count is zero), return. 801\& \- Before the first iteration, call any pending watchers.
542\& - Queue prepare watchers and then call all outstanding watchers. 802\& * If EVFLAG_FORKCHECK was used, check for a fork.
803\& \- If a fork was detected (by any means), queue and call all fork watchers.
804\& \- Queue and call all prepare watchers.
543\& - If we have been forked, recreate the kernel state. 805\& \- If we have been forked, detach and recreate the kernel state
806\& as to not disturb the other process.
544\& - Update the kernel state with all outstanding changes. 807\& \- Update the kernel state with all outstanding changes.
545\& - Update the "event loop time". 808\& \- Update the "event loop time" (ev_now ()).
546\& - Calculate for how long to block. 809\& \- Calculate for how long to sleep or block, if at all
810\& (active idle watchers, EVLOOP_NONBLOCK or not having
811\& any active watchers at all will result in not sleeping).
812\& \- Sleep if the I/O and timer collect interval say so.
547\& - Block the process, waiting for any events. 813\& \- Block the process, waiting for any events.
548\& - Queue all outstanding I/O (fd) events. 814\& \- Queue all outstanding I/O (fd) events.
549\& - Update the "event loop time" and do time jump handling. 815\& \- Update the "event loop time" (ev_now ()), and do time jump adjustments.
550\& - Queue all outstanding timers. 816\& \- Queue all expired timers.
551\& - Queue all outstanding periodics. 817\& \- Queue all expired periodics.
552\& - If no events are pending now, queue all idle watchers. 818\& \- Unless any events are pending now, queue all idle watchers.
553\& - Queue all check watchers. 819\& \- Queue all check watchers.
554\& - Call all queued watchers in reverse order (i.e. check watchers first). 820\& \- Call all queued watchers in reverse order (i.e. check watchers first).
555\& Signals and child watchers are implemented as I/O watchers, and will 821\& Signals and child watchers are implemented as I/O watchers, and will
556\& be handled here by queueing them when their watcher gets executed. 822\& be handled here by queueing them when their watcher gets executed.
557\& - If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK 823\& \- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
558\& were used, return, otherwise continue with step *. 824\& were used, or there are no active watchers, return, otherwise
825\& continue with step *.
559.Ve 826.Ve
560.Sp 827.Sp
561Example: queue some jobs and then loop until no events are outsanding 828Example: Queue some jobs and then loop until no events are outstanding
562anymore. 829anymore.
563.Sp 830.Sp
564.Vb 4 831.Vb 4
565\& ... queue jobs here, make sure they register event watchers as long 832\& ... queue jobs here, make sure they register event watchers as long
566\& ... as they still have work to do (even an idle watcher will do..) 833\& ... as they still have work to do (even an idle watcher will do..)
567\& ev_loop (my_loop, 0); 834\& ev_loop (my_loop, 0);
568\& ... jobs done. yeah! 835\& ... jobs done or somebody called unloop. yeah!
569.Ve 836.Ve
570.IP "ev_unloop (loop, how)" 4 837.IP "ev_unloop (loop, how)" 4
571.IX Item "ev_unloop (loop, how)" 838.IX Item "ev_unloop (loop, how)"
572Can be used to make a call to \f(CW\*(C`ev_loop\*(C'\fR return early (but only after it 839Can be used to make a call to \f(CW\*(C`ev_loop\*(C'\fR return early (but only after it
573has processed all outstanding events). The \f(CW\*(C`how\*(C'\fR argument must be either 840has processed all outstanding events). The \f(CW\*(C`how\*(C'\fR argument must be either
574\&\f(CW\*(C`EVUNLOOP_ONE\*(C'\fR, which will make the innermost \f(CW\*(C`ev_loop\*(C'\fR call return, or 841\&\f(CW\*(C`EVUNLOOP_ONE\*(C'\fR, which will make the innermost \f(CW\*(C`ev_loop\*(C'\fR call return, or
575\&\f(CW\*(C`EVUNLOOP_ALL\*(C'\fR, which will make all nested \f(CW\*(C`ev_loop\*(C'\fR calls return. 842\&\f(CW\*(C`EVUNLOOP_ALL\*(C'\fR, which will make all nested \f(CW\*(C`ev_loop\*(C'\fR calls return.
843.Sp
844This \*(L"unloop state\*(R" will be cleared when entering \f(CW\*(C`ev_loop\*(C'\fR again.
845.Sp
846It is safe to call \f(CW\*(C`ev_unloop\*(C'\fR from otuside any \f(CW\*(C`ev_loop\*(C'\fR calls.
576.IP "ev_ref (loop)" 4 847.IP "ev_ref (loop)" 4
577.IX Item "ev_ref (loop)" 848.IX Item "ev_ref (loop)"
578.PD 0 849.PD 0
579.IP "ev_unref (loop)" 4 850.IP "ev_unref (loop)" 4
580.IX Item "ev_unref (loop)" 851.IX Item "ev_unref (loop)"
581.PD 852.PD
582Ref/unref can be used to add or remove a reference count on the event 853Ref/unref can be used to add or remove a reference count on the event
583loop: Every watcher keeps one reference, and as long as the reference 854loop: Every watcher keeps one reference, and as long as the reference
584count is nonzero, \f(CW\*(C`ev_loop\*(C'\fR will not return on its own. If you have 855count is nonzero, \f(CW\*(C`ev_loop\*(C'\fR will not return on its own.
856.Sp
585a watcher you never unregister that should not keep \f(CW\*(C`ev_loop\*(C'\fR from 857If you have a watcher you never unregister that should not keep \f(CW\*(C`ev_loop\*(C'\fR
586returning, \fIev_unref()\fR after starting, and \fIev_ref()\fR before stopping it. For 858from returning, call \fIev_unref()\fR after starting, and \fIev_ref()\fR before
859stopping it.
860.Sp
587example, libev itself uses this for its internal signal pipe: It is not 861As an example, libev itself uses this for its internal signal pipe: It is
588visible to the libev user and should not keep \f(CW\*(C`ev_loop\*(C'\fR from exiting if 862not visible to the libev user and should not keep \f(CW\*(C`ev_loop\*(C'\fR from exiting
589no event watchers registered by it are active. It is also an excellent 863if no event watchers registered by it are active. It is also an excellent
590way to do this for generic recurring timers or from within third-party 864way to do this for generic recurring timers or from within third-party
591libraries. Just remember to \fIunref after start\fR and \fIref before stop\fR. 865libraries. Just remember to \fIunref after start\fR and \fIref before stop\fR
866(but only if the watcher wasn't active before, or was active before,
867respectively).
592.Sp 868.Sp
593Example: create a signal watcher, but keep it from keeping \f(CW\*(C`ev_loop\*(C'\fR 869Example: Create a signal watcher, but keep it from keeping \f(CW\*(C`ev_loop\*(C'\fR
594running when nothing else is active. 870running when nothing else is active.
595.Sp 871.Sp
596.Vb 4 872.Vb 4
597\& struct dv_signal exitsig; 873\& ev_signal exitsig;
598\& ev_signal_init (&exitsig, sig_cb, SIGINT); 874\& ev_signal_init (&exitsig, sig_cb, SIGINT);
599\& ev_signal_start (myloop, &exitsig); 875\& ev_signal_start (loop, &exitsig);
600\& evf_unref (myloop); 876\& evf_unref (loop);
601.Ve 877.Ve
602.Sp 878.Sp
603Example: for some weird reason, unregister the above signal handler again. 879Example: For some weird reason, unregister the above signal handler again.
604.Sp 880.Sp
605.Vb 2 881.Vb 2
606\& ev_ref (myloop); 882\& ev_ref (loop);
607\& ev_signal_stop (myloop, &exitsig); 883\& ev_signal_stop (loop, &exitsig);
608.Ve 884.Ve
885.IP "ev_set_io_collect_interval (loop, ev_tstamp interval)" 4
886.IX Item "ev_set_io_collect_interval (loop, ev_tstamp interval)"
887.PD 0
888.IP "ev_set_timeout_collect_interval (loop, ev_tstamp interval)" 4
889.IX Item "ev_set_timeout_collect_interval (loop, ev_tstamp interval)"
890.PD
891These advanced functions influence the time that libev will spend waiting
892for events. Both time intervals are by default \f(CW0\fR, meaning that libev
893will try to invoke timer/periodic callbacks and I/O callbacks with minimum
894latency.
895.Sp
896Setting these to a higher value (the \f(CW\*(C`interval\*(C'\fR \fImust\fR be >= \f(CW0\fR)
897allows libev to delay invocation of I/O and timer/periodic callbacks
898to increase efficiency of loop iterations (or to increase power-saving
899opportunities).
900.Sp
901The idea is that sometimes your program runs just fast enough to handle
902one (or very few) event(s) per loop iteration. While this makes the
903program responsive, it also wastes a lot of \s-1CPU\s0 time to poll for new
904events, especially with backends like \f(CW\*(C`select ()\*(C'\fR which have a high
905overhead for the actual polling but can deliver many events at once.
906.Sp
907By setting a higher \fIio collect interval\fR you allow libev to spend more
908time collecting I/O events, so you can handle more events per iteration,
909at the cost of increasing latency. Timeouts (both \f(CW\*(C`ev_periodic\*(C'\fR and
910\&\f(CW\*(C`ev_timer\*(C'\fR) will be not affected. Setting this to a non-null value will
911introduce an additional \f(CW\*(C`ev_sleep ()\*(C'\fR call into most loop iterations.
912.Sp
913Likewise, by setting a higher \fItimeout collect interval\fR you allow libev
914to spend more time collecting timeouts, at the expense of increased
915latency/jitter/inexactness (the watcher callback will be called
916later). \f(CW\*(C`ev_io\*(C'\fR watchers will not be affected. Setting this to a non-null
917value will not introduce any overhead in libev.
918.Sp
919Many (busy) programs can usually benefit by setting the I/O collect
920interval to a value near \f(CW0.1\fR or so, which is often enough for
921interactive servers (of course not for games), likewise for timeouts. It
922usually doesn't make much sense to set it to a lower value than \f(CW0.01\fR,
923as this approaches the timing granularity of most systems.
924.Sp
925Setting the \fItimeout collect interval\fR can improve the opportunity for
926saving power, as the program will \*(L"bundle\*(R" timer callback invocations that
927are \*(L"near\*(R" in time together, by delaying some, thus reducing the number of
928times the process sleeps and wakes up again. Another useful technique to
929reduce iterations/wake\-ups is to use \f(CW\*(C`ev_periodic\*(C'\fR watchers and make sure
930they fire on, say, one-second boundaries only.
931.IP "ev_loop_verify (loop)" 4
932.IX Item "ev_loop_verify (loop)"
933This function only does something when \f(CW\*(C`EV_VERIFY\*(C'\fR support has been
934compiled in, which is the default for non-minimal builds. It tries to go
935through all internal structures and checks them for validity. If anything
936is found to be inconsistent, it will print an error message to standard
937error and call \f(CW\*(C`abort ()\*(C'\fR.
938.Sp
939This can be used to catch bugs inside libev itself: under normal
940circumstances, this function will never abort as of course libev keeps its
941data structures consistent.
609.SH "ANATOMY OF A WATCHER" 942.SH "ANATOMY OF A WATCHER"
610.IX Header "ANATOMY OF A WATCHER" 943.IX Header "ANATOMY OF A WATCHER"
944In the following description, uppercase \f(CW\*(C`TYPE\*(C'\fR in names stands for the
945watcher type, e.g. \f(CW\*(C`ev_TYPE_start\*(C'\fR can mean \f(CW\*(C`ev_timer_start\*(C'\fR for timer
946watchers and \f(CW\*(C`ev_io_start\*(C'\fR for I/O watchers.
947.PP
611A watcher is a structure that you create and register to record your 948A watcher is a structure that you create and register to record your
612interest in some event. For instance, if you want to wait for \s-1STDIN\s0 to 949interest in some event. For instance, if you want to wait for \s-1STDIN\s0 to
613become readable, you would create an \f(CW\*(C`ev_io\*(C'\fR watcher for that: 950become readable, you would create an \f(CW\*(C`ev_io\*(C'\fR watcher for that:
614.PP 951.PP
615.Vb 5 952.Vb 5
616\& static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents) 953\& static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
617\& { 954\& {
618\& ev_io_stop (w); 955\& ev_io_stop (w);
619\& ev_unloop (loop, EVUNLOOP_ALL); 956\& ev_unloop (loop, EVUNLOOP_ALL);
620\& } 957\& }
621.Ve 958\&
622.PP
623.Vb 6
624\& struct ev_loop *loop = ev_default_loop (0); 959\& struct ev_loop *loop = ev_default_loop (0);
960\&
625\& struct ev_io stdin_watcher; 961\& ev_io stdin_watcher;
962\&
626\& ev_init (&stdin_watcher, my_cb); 963\& ev_init (&stdin_watcher, my_cb);
627\& ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ); 964\& ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
628\& ev_io_start (loop, &stdin_watcher); 965\& ev_io_start (loop, &stdin_watcher);
966\&
629\& ev_loop (loop, 0); 967\& ev_loop (loop, 0);
630.Ve 968.Ve
631.PP 969.PP
632As you can see, you are responsible for allocating the memory for your 970As you can see, you are responsible for allocating the memory for your
633watcher structures (and it is usually a bad idea to do this on the stack, 971watcher structures (and it is \fIusually\fR a bad idea to do this on the
634although this can sometimes be quite valid). 972stack).
973.PP
974Each watcher has an associated watcher structure (called \f(CW\*(C`struct ev_TYPE\*(C'\fR
975or simply \f(CW\*(C`ev_TYPE\*(C'\fR, as typedefs are provided for all watcher structs).
635.PP 976.PP
636Each watcher structure must be initialised by a call to \f(CW\*(C`ev_init 977Each watcher structure must be initialised by a call to \f(CW\*(C`ev_init
637(watcher *, callback)\*(C'\fR, which expects a callback to be provided. This 978(watcher *, callback)\*(C'\fR, which expects a callback to be provided. This
638callback gets invoked each time the event occurs (or, in the case of io 979callback gets invoked each time the event occurs (or, in the case of I/O
639watchers, each time the event loop detects that the file descriptor given 980watchers, each time the event loop detects that the file descriptor given
640is readable and/or writable). 981is readable and/or writable).
641.PP 982.PP
642Each watcher type has its own \f(CW\*(C`ev_<type>_set (watcher *, ...)\*(C'\fR macro 983Each watcher type further has its own \f(CW\*(C`ev_TYPE_set (watcher *, ...)\*(C'\fR
643with arguments specific to this watcher type. There is also a macro 984macro to configure it, with arguments specific to the watcher type. There
644to combine initialisation and setting in one call: \f(CW\*(C`ev_<type>_init 985is also a macro to combine initialisation and setting in one call: \f(CW\*(C`ev_TYPE_init (watcher *, callback, ...)\*(C'\fR.
645(watcher *, callback, ...)\*(C'\fR.
646.PP 986.PP
647To make the watcher actually watch out for events, you have to start it 987To make the watcher actually watch out for events, you have to start it
648with a watcher-specific start function (\f(CW\*(C`ev_<type>_start (loop, watcher 988with a watcher-specific start function (\f(CW\*(C`ev_TYPE_start (loop, watcher
649*)\*(C'\fR), and you can stop watching for events at any time by calling the 989*)\*(C'\fR), and you can stop watching for events at any time by calling the
650corresponding stop function (\f(CW\*(C`ev_<type>_stop (loop, watcher *)\*(C'\fR. 990corresponding stop function (\f(CW\*(C`ev_TYPE_stop (loop, watcher *)\*(C'\fR.
651.PP 991.PP
652As long as your watcher is active (has been started but not stopped) you 992As long as your watcher is active (has been started but not stopped) you
653must not touch the values stored in it. Most specifically you must never 993must not touch the values stored in it. Most specifically you must never
654reinitialise it or call its \f(CW\*(C`set\*(C'\fR macro. 994reinitialise it or call its \f(CW\*(C`ev_TYPE_set\*(C'\fR macro.
655.PP 995.PP
656Each and every callback receives the event loop pointer as first, the 996Each and every callback receives the event loop pointer as first, the
657registered watcher structure as second, and a bitset of received events as 997registered watcher structure as second, and a bitset of received events as
658third argument. 998third argument.
659.PP 999.PP
… …
684The signal specified in the \f(CW\*(C`ev_signal\*(C'\fR watcher has been received by a thread. 1024The signal specified in the \f(CW\*(C`ev_signal\*(C'\fR watcher has been received by a thread.
685.ie n .IP """EV_CHILD""" 4 1025.ie n .IP """EV_CHILD""" 4
686.el .IP "\f(CWEV_CHILD\fR" 4 1026.el .IP "\f(CWEV_CHILD\fR" 4
687.IX Item "EV_CHILD" 1027.IX Item "EV_CHILD"
688The pid specified in the \f(CW\*(C`ev_child\*(C'\fR watcher has received a status change. 1028The pid specified in the \f(CW\*(C`ev_child\*(C'\fR watcher has received a status change.
1029.ie n .IP """EV_STAT""" 4
1030.el .IP "\f(CWEV_STAT\fR" 4
1031.IX Item "EV_STAT"
1032The path specified in the \f(CW\*(C`ev_stat\*(C'\fR watcher changed its attributes somehow.
689.ie n .IP """EV_IDLE""" 4 1033.ie n .IP """EV_IDLE""" 4
690.el .IP "\f(CWEV_IDLE\fR" 4 1034.el .IP "\f(CWEV_IDLE\fR" 4
691.IX Item "EV_IDLE" 1035.IX Item "EV_IDLE"
692The \f(CW\*(C`ev_idle\*(C'\fR watcher has determined that you have nothing better to do. 1036The \f(CW\*(C`ev_idle\*(C'\fR watcher has determined that you have nothing better to do.
693.ie n .IP """EV_PREPARE""" 4 1037.ie n .IP """EV_PREPARE""" 4
… …
703\&\f(CW\*(C`ev_loop\*(C'\fR has gathered them, but before it invokes any callbacks for any 1047\&\f(CW\*(C`ev_loop\*(C'\fR has gathered them, but before it invokes any callbacks for any
704received events. Callbacks of both watcher types can start and stop as 1048received events. Callbacks of both watcher types can start and stop as
705many watchers as they want, and all of them will be taken into account 1049many watchers as they want, and all of them will be taken into account
706(for example, a \f(CW\*(C`ev_prepare\*(C'\fR watcher might start an idle watcher to keep 1050(for example, a \f(CW\*(C`ev_prepare\*(C'\fR watcher might start an idle watcher to keep
707\&\f(CW\*(C`ev_loop\*(C'\fR from blocking). 1051\&\f(CW\*(C`ev_loop\*(C'\fR from blocking).
1052.ie n .IP """EV_EMBED""" 4
1053.el .IP "\f(CWEV_EMBED\fR" 4
1054.IX Item "EV_EMBED"
1055The embedded event loop specified in the \f(CW\*(C`ev_embed\*(C'\fR watcher needs attention.
1056.ie n .IP """EV_FORK""" 4
1057.el .IP "\f(CWEV_FORK\fR" 4
1058.IX Item "EV_FORK"
1059The event loop has been resumed in the child process after fork (see
1060\&\f(CW\*(C`ev_fork\*(C'\fR).
1061.ie n .IP """EV_ASYNC""" 4
1062.el .IP "\f(CWEV_ASYNC\fR" 4
1063.IX Item "EV_ASYNC"
1064The given async watcher has been asynchronously notified (see \f(CW\*(C`ev_async\*(C'\fR).
708.ie n .IP """EV_ERROR""" 4 1065.ie n .IP """EV_ERROR""" 4
709.el .IP "\f(CWEV_ERROR\fR" 4 1066.el .IP "\f(CWEV_ERROR\fR" 4
710.IX Item "EV_ERROR" 1067.IX Item "EV_ERROR"
711An unspecified error has occured, the watcher has been stopped. This might 1068An unspecified error has occurred, the watcher has been stopped. This might
712happen because the watcher could not be properly started because libev 1069happen because the watcher could not be properly started because libev
713ran out of memory, a file descriptor was found to be closed or any other 1070ran out of memory, a file descriptor was found to be closed or any other
1071problem. Libev considers these application bugs.
1072.Sp
714problem. You best act on it by reporting the problem and somehow coping 1073You best act on it by reporting the problem and somehow coping with the
715with the watcher being stopped. 1074watcher being stopped. Note that well-written programs should not receive
1075an error ever, so when your watcher receives it, this usually indicates a
1076bug in your program.
716.Sp 1077.Sp
717Libev will usually signal a few \*(L"dummy\*(R" events together with an error, 1078Libev will usually signal a few \*(L"dummy\*(R" events together with an error, for
718for example it might indicate that a fd is readable or writable, and if 1079example it might indicate that a fd is readable or writable, and if your
719your callbacks is well-written it can just attempt the operation and cope 1080callbacks is well-written it can just attempt the operation and cope with
720with the error from \fIread()\fR or \fIwrite()\fR. This will not work in multithreaded 1081the error from \fIread()\fR or \fIwrite()\fR. This will not work in multi-threaded
721programs, though, so beware. 1082programs, though, as the fd could already be closed and reused for another
1083thing, so beware.
722.Sh "\s-1SUMMARY\s0 \s-1OF\s0 \s-1GENERIC\s0 \s-1WATCHER\s0 \s-1FUNCTIONS\s0" 1084.Sh "\s-1GENERIC\s0 \s-1WATCHER\s0 \s-1FUNCTIONS\s0"
723.IX Subsection "SUMMARY OF GENERIC WATCHER FUNCTIONS" 1085.IX Subsection "GENERIC WATCHER FUNCTIONS"
724In the following description, \f(CW\*(C`TYPE\*(C'\fR stands for the watcher type,
725e.g. \f(CW\*(C`timer\*(C'\fR for \f(CW\*(C`ev_timer\*(C'\fR watchers and \f(CW\*(C`io\*(C'\fR for \f(CW\*(C`ev_io\*(C'\fR watchers.
726.ie n .IP """ev_init"" (ev_TYPE *watcher, callback)" 4 1086.ie n .IP """ev_init"" (ev_TYPE *watcher, callback)" 4
727.el .IP "\f(CWev_init\fR (ev_TYPE *watcher, callback)" 4 1087.el .IP "\f(CWev_init\fR (ev_TYPE *watcher, callback)" 4
728.IX Item "ev_init (ev_TYPE *watcher, callback)" 1088.IX Item "ev_init (ev_TYPE *watcher, callback)"
729This macro initialises the generic portion of a watcher. The contents 1089This macro initialises the generic portion of a watcher. The contents
730of the watcher object can be arbitrary (so \f(CW\*(C`malloc\*(C'\fR will do). Only 1090of the watcher object can be arbitrary (so \f(CW\*(C`malloc\*(C'\fR will do). Only
… …
734which rolls both calls into one. 1094which rolls both calls into one.
735.Sp 1095.Sp
736You can reinitialise a watcher at any time as long as it has been stopped 1096You can reinitialise a watcher at any time as long as it has been stopped
737(or never started) and there are no pending events outstanding. 1097(or never started) and there are no pending events outstanding.
738.Sp 1098.Sp
739The callbakc is always of type \f(CW\*(C`void (*)(ev_loop *loop, ev_TYPE *watcher, 1099The callback is always of type \f(CW\*(C`void (*)(struct ev_loop *loop, ev_TYPE *watcher,
740int revents)\*(C'\fR. 1100int revents)\*(C'\fR.
1101.Sp
1102Example: Initialise an \f(CW\*(C`ev_io\*(C'\fR watcher in two steps.
1103.Sp
1104.Vb 3
1105\& ev_io w;
1106\& ev_init (&w, my_cb);
1107\& ev_io_set (&w, STDIN_FILENO, EV_READ);
1108.Ve
741.ie n .IP """ev_TYPE_set"" (ev_TYPE *, [args])" 4 1109.ie n .IP """ev_TYPE_set"" (ev_TYPE *, [args])" 4
742.el .IP "\f(CWev_TYPE_set\fR (ev_TYPE *, [args])" 4 1110.el .IP "\f(CWev_TYPE_set\fR (ev_TYPE *, [args])" 4
743.IX Item "ev_TYPE_set (ev_TYPE *, [args])" 1111.IX Item "ev_TYPE_set (ev_TYPE *, [args])"
744This macro initialises the type-specific parts of a watcher. You need to 1112This macro initialises the type-specific parts of a watcher. You need to
745call \f(CW\*(C`ev_init\*(C'\fR at least once before you call this macro, but you can 1113call \f(CW\*(C`ev_init\*(C'\fR at least once before you call this macro, but you can
… …
747macro on a watcher that is active (it can be pending, however, which is a 1115macro on a watcher that is active (it can be pending, however, which is a
748difference to the \f(CW\*(C`ev_init\*(C'\fR macro). 1116difference to the \f(CW\*(C`ev_init\*(C'\fR macro).
749.Sp 1117.Sp
750Although some watcher types do not have type-specific arguments 1118Although some watcher types do not have type-specific arguments
751(e.g. \f(CW\*(C`ev_prepare\*(C'\fR) you still need to call its \f(CW\*(C`set\*(C'\fR macro. 1119(e.g. \f(CW\*(C`ev_prepare\*(C'\fR) you still need to call its \f(CW\*(C`set\*(C'\fR macro.
1120.Sp
1121See \f(CW\*(C`ev_init\*(C'\fR, above, for an example.
752.ie n .IP """ev_TYPE_init"" (ev_TYPE *watcher, callback, [args])" 4 1122.ie n .IP """ev_TYPE_init"" (ev_TYPE *watcher, callback, [args])" 4
753.el .IP "\f(CWev_TYPE_init\fR (ev_TYPE *watcher, callback, [args])" 4 1123.el .IP "\f(CWev_TYPE_init\fR (ev_TYPE *watcher, callback, [args])" 4
754.IX Item "ev_TYPE_init (ev_TYPE *watcher, callback, [args])" 1124.IX Item "ev_TYPE_init (ev_TYPE *watcher, callback, [args])"
755This convinience macro rolls both \f(CW\*(C`ev_init\*(C'\fR and \f(CW\*(C`ev_TYPE_set\*(C'\fR macro 1125This convenience macro rolls both \f(CW\*(C`ev_init\*(C'\fR and \f(CW\*(C`ev_TYPE_set\*(C'\fR macro
756calls into a single call. This is the most convinient method to initialise 1126calls into a single call. This is the most convenient method to initialise
757a watcher. The same limitations apply, of course. 1127a watcher. The same limitations apply, of course.
1128.Sp
1129Example: Initialise and set an \f(CW\*(C`ev_io\*(C'\fR watcher in one step.
1130.Sp
1131.Vb 1
1132\& ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
1133.Ve
758.ie n .IP """ev_TYPE_start"" (loop *, ev_TYPE *watcher)" 4 1134.ie n .IP """ev_TYPE_start"" (loop *, ev_TYPE *watcher)" 4
759.el .IP "\f(CWev_TYPE_start\fR (loop *, ev_TYPE *watcher)" 4 1135.el .IP "\f(CWev_TYPE_start\fR (loop *, ev_TYPE *watcher)" 4
760.IX Item "ev_TYPE_start (loop *, ev_TYPE *watcher)" 1136.IX Item "ev_TYPE_start (loop *, ev_TYPE *watcher)"
761Starts (activates) the given watcher. Only active watchers will receive 1137Starts (activates) the given watcher. Only active watchers will receive
762events. If the watcher is already active nothing will happen. 1138events. If the watcher is already active nothing will happen.
1139.Sp
1140Example: Start the \f(CW\*(C`ev_io\*(C'\fR watcher that is being abused as example in this
1141whole section.
1142.Sp
1143.Vb 1
1144\& ev_io_start (EV_DEFAULT_UC, &w);
1145.Ve
763.ie n .IP """ev_TYPE_stop"" (loop *, ev_TYPE *watcher)" 4 1146.ie n .IP """ev_TYPE_stop"" (loop *, ev_TYPE *watcher)" 4
764.el .IP "\f(CWev_TYPE_stop\fR (loop *, ev_TYPE *watcher)" 4 1147.el .IP "\f(CWev_TYPE_stop\fR (loop *, ev_TYPE *watcher)" 4
765.IX Item "ev_TYPE_stop (loop *, ev_TYPE *watcher)" 1148.IX Item "ev_TYPE_stop (loop *, ev_TYPE *watcher)"
766Stops the given watcher again (if active) and clears the pending 1149Stops the given watcher if active, and clears the pending status (whether
1150the watcher was active or not).
1151.Sp
767status. It is possible that stopped watchers are pending (for example, 1152It is possible that stopped watchers are pending \- for example,
768non-repeating timers are being stopped when they become pending), but 1153non-repeating timers are being stopped when they become pending \- but
769\&\f(CW\*(C`ev_TYPE_stop\*(C'\fR ensures that the watcher is neither active nor pending. If 1154calling \f(CW\*(C`ev_TYPE_stop\*(C'\fR ensures that the watcher is neither active nor
770you want to free or reuse the memory used by the watcher it is therefore a 1155pending. If you want to free or reuse the memory used by the watcher it is
771good idea to always call its \f(CW\*(C`ev_TYPE_stop\*(C'\fR function. 1156therefore a good idea to always call its \f(CW\*(C`ev_TYPE_stop\*(C'\fR function.
772.IP "bool ev_is_active (ev_TYPE *watcher)" 4 1157.IP "bool ev_is_active (ev_TYPE *watcher)" 4
773.IX Item "bool ev_is_active (ev_TYPE *watcher)" 1158.IX Item "bool ev_is_active (ev_TYPE *watcher)"
774Returns a true value iff the watcher is active (i.e. it has been started 1159Returns a true value iff the watcher is active (i.e. it has been started
775and not yet been stopped). As long as a watcher is active you must not modify 1160and not yet been stopped). As long as a watcher is active you must not modify
776it. 1161it.
777.IP "bool ev_is_pending (ev_TYPE *watcher)" 4 1162.IP "bool ev_is_pending (ev_TYPE *watcher)" 4
778.IX Item "bool ev_is_pending (ev_TYPE *watcher)" 1163.IX Item "bool ev_is_pending (ev_TYPE *watcher)"
779Returns a true value iff the watcher is pending, (i.e. it has outstanding 1164Returns a true value iff the watcher is pending, (i.e. it has outstanding
780events but its callback has not yet been invoked). As long as a watcher 1165events but its callback has not yet been invoked). As long as a watcher
781is pending (but not active) you must not call an init function on it (but 1166is pending (but not active) you must not call an init function on it (but
782\&\f(CW\*(C`ev_TYPE_set\*(C'\fR is safe) and you must make sure the watcher is available to 1167\&\f(CW\*(C`ev_TYPE_set\*(C'\fR is safe), you must not change its priority, and you must
783libev (e.g. you cnanot \f(CW\*(C`free ()\*(C'\fR it). 1168make sure the watcher is available to libev (e.g. you cannot \f(CW\*(C`free ()\*(C'\fR
1169it).
784.IP "callback = ev_cb (ev_TYPE *watcher)" 4 1170.IP "callback ev_cb (ev_TYPE *watcher)" 4
785.IX Item "callback = ev_cb (ev_TYPE *watcher)" 1171.IX Item "callback ev_cb (ev_TYPE *watcher)"
786Returns the callback currently set on the watcher. 1172Returns the callback currently set on the watcher.
787.IP "ev_cb_set (ev_TYPE *watcher, callback)" 4 1173.IP "ev_cb_set (ev_TYPE *watcher, callback)" 4
788.IX Item "ev_cb_set (ev_TYPE *watcher, callback)" 1174.IX Item "ev_cb_set (ev_TYPE *watcher, callback)"
789Change the callback. You can change the callback at virtually any time 1175Change the callback. You can change the callback at virtually any time
790(modulo threads). 1176(modulo threads).
1177.IP "ev_set_priority (ev_TYPE *watcher, priority)" 4
1178.IX Item "ev_set_priority (ev_TYPE *watcher, priority)"
1179.PD 0
1180.IP "int ev_priority (ev_TYPE *watcher)" 4
1181.IX Item "int ev_priority (ev_TYPE *watcher)"
1182.PD
1183Set and query the priority of the watcher. The priority is a small
1184integer between \f(CW\*(C`EV_MAXPRI\*(C'\fR (default: \f(CW2\fR) and \f(CW\*(C`EV_MINPRI\*(C'\fR
1185(default: \f(CW\*(C`\-2\*(C'\fR). Pending watchers with higher priority will be invoked
1186before watchers with lower priority, but priority will not keep watchers
1187from being executed (except for \f(CW\*(C`ev_idle\*(C'\fR watchers).
1188.Sp
1189This means that priorities are \fIonly\fR used for ordering callback
1190invocation after new events have been received. This is useful, for
1191example, to reduce latency after idling, or more often, to bind two
1192watchers on the same event and make sure one is called first.
1193.Sp
1194If you need to suppress invocation when higher priority events are pending
1195you need to look at \f(CW\*(C`ev_idle\*(C'\fR watchers, which provide this functionality.
1196.Sp
1197You \fImust not\fR change the priority of a watcher as long as it is active or
1198pending.
1199.Sp
1200The default priority used by watchers when no priority has been set is
1201always \f(CW0\fR, which is supposed to not be too high and not be too low :).
1202.Sp
1203Setting a priority outside the range of \f(CW\*(C`EV_MINPRI\*(C'\fR to \f(CW\*(C`EV_MAXPRI\*(C'\fR is
1204fine, as long as you do not mind that the priority value you query might
1205or might not have been clamped to the valid range.
1206.IP "ev_invoke (loop, ev_TYPE *watcher, int revents)" 4
1207.IX Item "ev_invoke (loop, ev_TYPE *watcher, int revents)"
1208Invoke the \f(CW\*(C`watcher\*(C'\fR with the given \f(CW\*(C`loop\*(C'\fR and \f(CW\*(C`revents\*(C'\fR. Neither
1209\&\f(CW\*(C`loop\*(C'\fR nor \f(CW\*(C`revents\*(C'\fR need to be valid as long as the watcher callback
1210can deal with that fact, as both are simply passed through to the
1211callback.
1212.IP "int ev_clear_pending (loop, ev_TYPE *watcher)" 4
1213.IX Item "int ev_clear_pending (loop, ev_TYPE *watcher)"
1214If the watcher is pending, this function clears its pending status and
1215returns its \f(CW\*(C`revents\*(C'\fR bitset (as if its callback was invoked). If the
1216watcher isn't pending it does nothing and returns \f(CW0\fR.
1217.Sp
1218Sometimes it can be useful to \*(L"poll\*(R" a watcher instead of waiting for its
1219callback to be invoked, which can be accomplished with this function.
791.Sh "\s-1ASSOCIATING\s0 \s-1CUSTOM\s0 \s-1DATA\s0 \s-1WITH\s0 A \s-1WATCHER\s0" 1220.Sh "\s-1ASSOCIATING\s0 \s-1CUSTOM\s0 \s-1DATA\s0 \s-1WITH\s0 A \s-1WATCHER\s0"
792.IX Subsection "ASSOCIATING CUSTOM DATA WITH A WATCHER" 1221.IX Subsection "ASSOCIATING CUSTOM DATA WITH A WATCHER"
793Each watcher has, by default, a member \f(CW\*(C`void *data\*(C'\fR that you can change 1222Each watcher has, by default, a member \f(CW\*(C`void *data\*(C'\fR that you can change
794and read at any time, libev will completely ignore it. This can be used 1223and read at any time: libev will completely ignore it. This can be used
795to associate arbitrary data with your watcher. If you need more data and 1224to associate arbitrary data with your watcher. If you need more data and
796don't want to allocate memory and store a pointer to it in that data 1225don't want to allocate memory and store a pointer to it in that data
797member, you can also \*(L"subclass\*(R" the watcher type and provide your own 1226member, you can also \*(L"subclass\*(R" the watcher type and provide your own
798data: 1227data:
799.PP 1228.PP
800.Vb 7 1229.Vb 7
801\& struct my_io 1230\& struct my_io
802\& { 1231\& {
803\& struct ev_io io; 1232\& ev_io io;
804\& int otherfd; 1233\& int otherfd;
805\& void *somedata; 1234\& void *somedata;
806\& struct whatever *mostinteresting; 1235\& struct whatever *mostinteresting;
807\& } 1236\& };
1237\&
1238\& ...
1239\& struct my_io w;
1240\& ev_io_init (&w.io, my_cb, fd, EV_READ);
808.Ve 1241.Ve
809.PP 1242.PP
810And since your callback will be called with a pointer to the watcher, you 1243And since your callback will be called with a pointer to the watcher, you
811can cast it back to your own type: 1244can cast it back to your own type:
812.PP 1245.PP
813.Vb 5 1246.Vb 5
814\& static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents) 1247\& static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
815\& { 1248\& {
816\& struct my_io *w = (struct my_io *)w_; 1249\& struct my_io *w = (struct my_io *)w_;
817\& ... 1250\& ...
818\& } 1251\& }
819.Ve 1252.Ve
820.PP 1253.PP
821More interesting and less C\-conformant ways of catsing your callback type 1254More interesting and less C\-conformant ways of casting your callback type
822have been omitted.... 1255instead have been omitted.
1256.PP
1257Another common scenario is to use some data structure with multiple
1258embedded watchers:
1259.PP
1260.Vb 6
1261\& struct my_biggy
1262\& {
1263\& int some_data;
1264\& ev_timer t1;
1265\& ev_timer t2;
1266\& }
1267.Ve
1268.PP
1269In this case getting the pointer to \f(CW\*(C`my_biggy\*(C'\fR is a bit more
1270complicated: Either you store the address of your \f(CW\*(C`my_biggy\*(C'\fR struct
1271in the \f(CW\*(C`data\*(C'\fR member of the watcher (for woozies), or you need to use
1272some pointer arithmetic using \f(CW\*(C`offsetof\*(C'\fR inside your watchers (for real
1273programmers):
1274.PP
1275.Vb 1
1276\& #include <stddef.h>
1277\&
1278\& static void
1279\& t1_cb (EV_P_ ev_timer *w, int revents)
1280\& {
1281\& struct my_biggy big = (struct my_biggy *
1282\& (((char *)w) \- offsetof (struct my_biggy, t1));
1283\& }
1284\&
1285\& static void
1286\& t2_cb (EV_P_ ev_timer *w, int revents)
1287\& {
1288\& struct my_biggy big = (struct my_biggy *
1289\& (((char *)w) \- offsetof (struct my_biggy, t2));
1290\& }
1291.Ve
823.SH "WATCHER TYPES" 1292.SH "WATCHER TYPES"
824.IX Header "WATCHER TYPES" 1293.IX Header "WATCHER TYPES"
825This section describes each watcher in detail, but will not repeat 1294This section describes each watcher in detail, but will not repeat
826information given in the last section. 1295information given in the last section. Any initialisation/set macros,
1296functions and members specific to the watcher type are explained.
1297.PP
1298Members are additionally marked with either \fI[read\-only]\fR, meaning that,
1299while the watcher is active, you can look at the member and expect some
1300sensible content, but you must not modify it (you can modify it while the
1301watcher is stopped to your hearts content), or \fI[read\-write]\fR, which
1302means you can expect it to have some sensible content while the watcher
1303is active, but you can also modify it. Modifying it may not do something
1304sensible or take immediate effect (or do anything at all), but libev will
1305not crash or malfunction in any way.
827.ie n .Sh """ev_io"" \- is this file descriptor readable or writable" 1306.ie n .Sh """ev_io"" \- is this file descriptor readable or writable?"
828.el .Sh "\f(CWev_io\fP \- is this file descriptor readable or writable" 1307.el .Sh "\f(CWev_io\fP \- is this file descriptor readable or writable?"
829.IX Subsection "ev_io - is this file descriptor readable or writable" 1308.IX Subsection "ev_io - is this file descriptor readable or writable?"
830I/O watchers check whether a file descriptor is readable or writable 1309I/O watchers check whether a file descriptor is readable or writable
831in each iteration of the event loop (This behaviour is called 1310in each iteration of the event loop, or, more precisely, when reading
832level-triggering because you keep receiving events as long as the 1311would not block the process and writing would at least be able to write
833condition persists. Remember you can stop the watcher if you don't want to 1312some data. This behaviour is called level-triggering because you keep
834act on the event and neither want to receive future events). 1313receiving events as long as the condition persists. Remember you can stop
1314the watcher if you don't want to act on the event and neither want to
1315receive future events.
835.PP 1316.PP
836In general you can register as many read and/or write event watchers per 1317In general you can register as many read and/or write event watchers per
837fd as you want (as long as you don't confuse yourself). Setting all file 1318fd as you want (as long as you don't confuse yourself). Setting all file
838descriptors to non-blocking mode is also usually a good idea (but not 1319descriptors to non-blocking mode is also usually a good idea (but not
839required if you know what you are doing). 1320required if you know what you are doing).
840.PP 1321.PP
841You have to be careful with dup'ed file descriptors, though. Some backends 1322If you cannot use non-blocking mode, then force the use of a
842(the linux epoll backend is a notable example) cannot handle dup'ed file 1323known-to-be-good backend (at the time of this writing, this includes only
843descriptors correctly if you register interest in two or more fds pointing 1324\&\f(CW\*(C`EVBACKEND_SELECT\*(C'\fR and \f(CW\*(C`EVBACKEND_POLL\*(C'\fR).
844to the same underlying file/socket etc. description (that is, they share
845the same underlying \*(L"file open\*(R").
846.PP 1325.PP
847If you must do this, then force the use of a known-to-be-good backend 1326Another thing you have to watch out for is that it is quite easy to
848(at the time of this writing, this includes only \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR and 1327receive \*(L"spurious\*(R" readiness notifications, that is your callback might
1328be called with \f(CW\*(C`EV_READ\*(C'\fR but a subsequent \f(CW\*(C`read\*(C'\fR(2) will actually block
1329because there is no data. Not only are some backends known to create a
1330lot of those (for example Solaris ports), it is very easy to get into
1331this situation even with a relatively standard program structure. Thus
1332it is best to always use non-blocking I/O: An extra \f(CW\*(C`read\*(C'\fR(2) returning
1333\&\f(CW\*(C`EAGAIN\*(C'\fR is far preferable to a program hanging until some data arrives.
1334.PP
1335If you cannot run the fd in non-blocking mode (for example you should
1336not play around with an Xlib connection), then you have to separately
1337re-test whether a file descriptor is really ready with a known-to-be good
1338interface such as poll (fortunately in our Xlib example, Xlib already
1339does this on its own, so its quite safe to use). Some people additionally
1340use \f(CW\*(C`SIGALRM\*(C'\fR and an interval timer, just to be sure you won't block
1341indefinitely.
1342.PP
1343But really, best use non-blocking mode.
1344.PP
1345\fIThe special problem of disappearing file descriptors\fR
1346.IX Subsection "The special problem of disappearing file descriptors"
1347.PP
1348Some backends (e.g. kqueue, epoll) need to be told about closing a file
1349descriptor (either due to calling \f(CW\*(C`close\*(C'\fR explicitly or any other means,
1350such as \f(CW\*(C`dup2\*(C'\fR). The reason is that you register interest in some file
1351descriptor, but when it goes away, the operating system will silently drop
1352this interest. If another file descriptor with the same number then is
1353registered with libev, there is no efficient way to see that this is, in
1354fact, a different file descriptor.
1355.PP
1356To avoid having to explicitly tell libev about such cases, libev follows
1357the following policy: Each time \f(CW\*(C`ev_io_set\*(C'\fR is being called, libev
1358will assume that this is potentially a new file descriptor, otherwise
1359it is assumed that the file descriptor stays the same. That means that
1360you \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
1361descriptor even if the file descriptor number itself did not change.
1362.PP
1363This is how one would do it normally anyway, the important point is that
1364the libev application should not optimise around libev but should leave
1365optimisations to libev.
1366.PP
1367\fIThe special problem of dup'ed file descriptors\fR
1368.IX Subsection "The special problem of dup'ed file descriptors"
1369.PP
1370Some backends (e.g. epoll), cannot register events for file descriptors,
1371but only events for the underlying file descriptions. That means when you
1372have \f(CW\*(C`dup ()\*(C'\fR'ed file descriptors or weirder constellations, and register
1373events for them, only one file descriptor might actually receive events.
1374.PP
1375There is no workaround possible except not registering events
1376for potentially \f(CW\*(C`dup ()\*(C'\fR'ed file descriptors, or to resort to
1377\&\f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or \f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
1378.PP
1379\fIThe special problem of fork\fR
1380.IX Subsection "The special problem of fork"
1381.PP
1382Some backends (epoll, kqueue) do not support \f(CW\*(C`fork ()\*(C'\fR at all or exhibit
1383useless behaviour. Libev fully supports fork, but needs to be told about
1384it in the child.
1385.PP
1386To support fork in your programs, you either have to call
1387\&\f(CW\*(C`ev_default_fork ()\*(C'\fR or \f(CW\*(C`ev_loop_fork ()\*(C'\fR after a fork in the child,
1388enable \f(CW\*(C`EVFLAG_FORKCHECK\*(C'\fR, or resort to \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or
849\&\f(CW\*(C`EVBACKEND_POLL\*(C'\fR). 1389\&\f(CW\*(C`EVBACKEND_POLL\*(C'\fR.
1390.PP
1391\fIThe special problem of \s-1SIGPIPE\s0\fR
1392.IX Subsection "The special problem of SIGPIPE"
1393.PP
1394While not really specific to libev, it is easy to forget about \f(CW\*(C`SIGPIPE\*(C'\fR:
1395when writing to a pipe whose other end has been closed, your program gets
1396sent a \s-1SIGPIPE\s0, which, by default, aborts your program. For most programs
1397this is sensible behaviour, for daemons, this is usually undesirable.
1398.PP
1399So when you encounter spurious, unexplained daemon exits, make sure you
1400ignore \s-1SIGPIPE\s0 (and maybe make sure you log the exit status of your daemon
1401somewhere, as that would have given you a big clue).
1402.PP
1403\fIWatcher-Specific Functions\fR
1404.IX Subsection "Watcher-Specific Functions"
850.IP "ev_io_init (ev_io *, callback, int fd, int events)" 4 1405.IP "ev_io_init (ev_io *, callback, int fd, int events)" 4
851.IX Item "ev_io_init (ev_io *, callback, int fd, int events)" 1406.IX Item "ev_io_init (ev_io *, callback, int fd, int events)"
852.PD 0 1407.PD 0
853.IP "ev_io_set (ev_io *, int fd, int events)" 4 1408.IP "ev_io_set (ev_io *, int fd, int events)" 4
854.IX Item "ev_io_set (ev_io *, int fd, int events)" 1409.IX Item "ev_io_set (ev_io *, int fd, int events)"
855.PD 1410.PD
856Configures an \f(CW\*(C`ev_io\*(C'\fR watcher. The fd is the file descriptor to rceeive 1411Configures an \f(CW\*(C`ev_io\*(C'\fR watcher. The \f(CW\*(C`fd\*(C'\fR is the file descriptor to
857events for and events is either \f(CW\*(C`EV_READ\*(C'\fR, \f(CW\*(C`EV_WRITE\*(C'\fR or \f(CW\*(C`EV_READ | 1412receive events for and \f(CW\*(C`events\*(C'\fR is either \f(CW\*(C`EV_READ\*(C'\fR, \f(CW\*(C`EV_WRITE\*(C'\fR or
858EV_WRITE\*(C'\fR to receive the given events. 1413\&\f(CW\*(C`EV_READ | EV_WRITE\*(C'\fR, to express the desire to receive the given events.
859.Sp 1414.IP "int fd [read\-only]" 4
860Please note that most of the more scalable backend mechanisms (for example 1415.IX Item "int fd [read-only]"
861epoll and solaris ports) can result in spurious readyness notifications 1416The file descriptor being watched.
862for file descriptors, so you practically need to use non-blocking I/O (and 1417.IP "int events [read\-only]" 4
863treat callback invocation as hint only), or retest separately with a safe 1418.IX Item "int events [read-only]"
864interface before doing I/O (XLib can do this), or force the use of either 1419The events being watched.
865\&\f(CW\*(C`EVBACKEND_SELECT\*(C'\fR or \f(CW\*(C`EVBACKEND_POLL\*(C'\fR, which don't suffer from this
866problem. Also note that it is quite easy to have your callback invoked
867when the readyness condition is no longer valid even when employing
868typical ways of handling events, so its a good idea to use non-blocking
869I/O unconditionally.
870.PP 1420.PP
1421\fIExamples\fR
1422.IX Subsection "Examples"
1423.PP
871Example: call \f(CW\*(C`stdin_readable_cb\*(C'\fR when \s-1STDIN_FILENO\s0 has become, well 1424Example: Call \f(CW\*(C`stdin_readable_cb\*(C'\fR when \s-1STDIN_FILENO\s0 has become, well
872readable, but only once. Since it is likely line\-buffered, you could 1425readable, but only once. Since it is likely line-buffered, you could
873attempt to read a whole line in the callback: 1426attempt to read a whole line in the callback.
874.PP 1427.PP
875.Vb 6 1428.Vb 6
876\& static void 1429\& static void
877\& stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents) 1430\& stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
878\& { 1431\& {
879\& ev_io_stop (loop, w); 1432\& ev_io_stop (loop, w);
880\& .. read from stdin here (or from w->fd) and haqndle any I/O errors 1433\& .. read from stdin here (or from w\->fd) and handle any I/O errors
881\& } 1434\& }
882.Ve 1435\&
883.PP
884.Vb 6
885\& ... 1436\& ...
886\& struct ev_loop *loop = ev_default_init (0); 1437\& struct ev_loop *loop = ev_default_init (0);
887\& struct ev_io stdin_readable; 1438\& ev_io stdin_readable;
888\& ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ); 1439\& ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
889\& ev_io_start (loop, &stdin_readable); 1440\& ev_io_start (loop, &stdin_readable);
890\& ev_loop (loop, 0); 1441\& ev_loop (loop, 0);
891.Ve 1442.Ve
892.ie n .Sh """ev_timer"" \- relative and optionally recurring timeouts" 1443.ie n .Sh """ev_timer"" \- relative and optionally repeating timeouts"
893.el .Sh "\f(CWev_timer\fP \- relative and optionally recurring timeouts" 1444.el .Sh "\f(CWev_timer\fP \- relative and optionally repeating timeouts"
894.IX Subsection "ev_timer - relative and optionally recurring timeouts" 1445.IX Subsection "ev_timer - relative and optionally repeating timeouts"
895Timer watchers are simple relative timers that generate an event after a 1446Timer watchers are simple relative timers that generate an event after a
896given time, and optionally repeating in regular intervals after that. 1447given time, and optionally repeating in regular intervals after that.
897.PP 1448.PP
898The timers are based on real time, that is, if you register an event that 1449The timers are based on real time, that is, if you register an event that
899times out after an hour and you reset your system clock to last years 1450times out after an hour and you reset your system clock to January last
900time, it will still time out after (roughly) and hour. \*(L"Roughly\*(R" because 1451year, it will still time out after (roughly) one hour. \*(L"Roughly\*(R" because
901detecting time jumps is hard, and some inaccuracies are unavoidable (the 1452detecting time jumps is hard, and some inaccuracies are unavoidable (the
902monotonic clock option helps a lot here). 1453monotonic clock option helps a lot here).
1454.PP
1455The callback is guaranteed to be invoked only \fIafter\fR its timeout has
1456passed, but if multiple timers become ready during the same loop iteration
1457then order of execution is undefined.
1458.PP
1459\fIBe smart about timeouts\fR
1460.IX Subsection "Be smart about timeouts"
1461.PP
1462Many real-world problems involve some kind of timeout, usually for error
1463recovery. A typical example is an \s-1HTTP\s0 request \- if the other side hangs,
1464you want to raise some error after a while.
1465.PP
1466What follows are some ways to handle this problem, from obvious and
1467inefficient to smart and efficient.
1468.PP
1469In the following, a 60 second activity timeout is assumed \- a timeout that
1470gets reset to 60 seconds each time there is activity (e.g. each time some
1471data or other life sign was received).
1472.IP "1. Use a timer and stop, reinitialise and start it on activity." 4
1473.IX Item "1. Use a timer and stop, reinitialise and start it on activity."
1474This is the most obvious, but not the most simple way: In the beginning,
1475start the watcher:
1476.Sp
1477.Vb 2
1478\& ev_timer_init (timer, callback, 60., 0.);
1479\& ev_timer_start (loop, timer);
1480.Ve
1481.Sp
1482Then, each time there is some activity, \f(CW\*(C`ev_timer_stop\*(C'\fR it, initialise it
1483and start it again:
1484.Sp
1485.Vb 3
1486\& ev_timer_stop (loop, timer);
1487\& ev_timer_set (timer, 60., 0.);
1488\& ev_timer_start (loop, timer);
1489.Ve
1490.Sp
1491This is relatively simple to implement, but means that each time there is
1492some activity, libev will first have to remove the timer from its internal
1493data structure and then add it again. Libev tries to be fast, but it's
1494still not a constant-time operation.
1495.ie n .IP "2. Use a timer and re-start it with ""ev_timer_again"" inactivity." 4
1496.el .IP "2. Use a timer and re-start it with \f(CWev_timer_again\fR inactivity." 4
1497.IX Item "2. Use a timer and re-start it with ev_timer_again inactivity."
1498This is the easiest way, and involves using \f(CW\*(C`ev_timer_again\*(C'\fR instead of
1499\&\f(CW\*(C`ev_timer_start\*(C'\fR.
1500.Sp
1501To implement this, configure an \f(CW\*(C`ev_timer\*(C'\fR with a \f(CW\*(C`repeat\*(C'\fR value
1502of \f(CW60\fR and then call \f(CW\*(C`ev_timer_again\*(C'\fR at start and each time you
1503successfully read or write some data. If you go into an idle state where
1504you do not expect data to travel on the socket, you can \f(CW\*(C`ev_timer_stop\*(C'\fR
1505the timer, and \f(CW\*(C`ev_timer_again\*(C'\fR will automatically restart it if need be.
1506.Sp
1507That means you can ignore both the \f(CW\*(C`ev_timer_start\*(C'\fR function and the
1508\&\f(CW\*(C`after\*(C'\fR argument to \f(CW\*(C`ev_timer_set\*(C'\fR, and only ever use the \f(CW\*(C`repeat\*(C'\fR
1509member and \f(CW\*(C`ev_timer_again\*(C'\fR.
1510.Sp
1511At start:
1512.Sp
1513.Vb 3
1514\& ev_timer_init (timer, callback);
1515\& timer\->repeat = 60.;
1516\& ev_timer_again (loop, timer);
1517.Ve
1518.Sp
1519Each time there is some activity:
1520.Sp
1521.Vb 1
1522\& ev_timer_again (loop, timer);
1523.Ve
1524.Sp
1525It is even possible to change the time-out on the fly, regardless of
1526whether the watcher is active or not:
1527.Sp
1528.Vb 2
1529\& timer\->repeat = 30.;
1530\& ev_timer_again (loop, timer);
1531.Ve
1532.Sp
1533This is slightly more efficient then stopping/starting the timer each time
1534you want to modify its timeout value, as libev does not have to completely
1535remove and re-insert the timer from/into its internal data structure.
1536.Sp
1537It is, however, even simpler than the \*(L"obvious\*(R" way to do it.
1538.IP "3. Let the timer time out, but then re-arm it as required." 4
1539.IX Item "3. Let the timer time out, but then re-arm it as required."
1540This method is more tricky, but usually most efficient: Most timeouts are
1541relatively long compared to the intervals between other activity \- in
1542our example, within 60 seconds, there are usually many I/O events with
1543associated activity resets.
1544.Sp
1545In this case, it would be more efficient to leave the \f(CW\*(C`ev_timer\*(C'\fR alone,
1546but remember the time of last activity, and check for a real timeout only
1547within the callback:
1548.Sp
1549.Vb 1
1550\& ev_tstamp last_activity; // time of last activity
1551\&
1552\& static void
1553\& callback (EV_P_ ev_timer *w, int revents)
1554\& {
1555\& ev_tstamp now = ev_now (EV_A);
1556\& ev_tstamp timeout = last_activity + 60.;
1557\&
1558\& // if last_activity + 60. is older than now, we did time out
1559\& if (timeout < now)
1560\& {
1561\& // timeout occured, take action
1562\& }
1563\& else
1564\& {
1565\& // callback was invoked, but there was some activity, re\-arm
1566\& // the watcher to fire in last_activity + 60, which is
1567\& // guaranteed to be in the future, so "again" is positive:
1568\& w\->repeat = timeout \- now;
1569\& ev_timer_again (EV_A_ w);
1570\& }
1571\& }
1572.Ve
1573.Sp
1574To summarise the callback: first calculate the real timeout (defined
1575as \*(L"60 seconds after the last activity\*(R"), then check if that time has
1576been reached, which means something \fIdid\fR, in fact, time out. Otherwise
1577the callback was invoked too early (\f(CW\*(C`timeout\*(C'\fR is in the future), so
1578re-schedule the timer to fire at that future time, to see if maybe we have
1579a timeout then.
1580.Sp
1581Note how \f(CW\*(C`ev_timer_again\*(C'\fR is used, taking advantage of the
1582\&\f(CW\*(C`ev_timer_again\*(C'\fR optimisation when the timer is already running.
1583.Sp
1584This scheme causes more callback invocations (about one every 60 seconds
1585minus half the average time between activity), but virtually no calls to
1586libev to change the timeout.
1587.Sp
1588To start the timer, simply initialise the watcher and set \f(CW\*(C`last_activity\*(C'\fR
1589to the current time (meaning we just have some activity :), then call the
1590callback, which will \*(L"do the right thing\*(R" and start the timer:
1591.Sp
1592.Vb 3
1593\& ev_timer_init (timer, callback);
1594\& last_activity = ev_now (loop);
1595\& callback (loop, timer, EV_TIMEOUT);
1596.Ve
1597.Sp
1598And when there is some activity, simply store the current time in
1599\&\f(CW\*(C`last_activity\*(C'\fR, no libev calls at all:
1600.Sp
1601.Vb 1
1602\& last_actiivty = ev_now (loop);
1603.Ve
1604.Sp
1605This technique is slightly more complex, but in most cases where the
1606time-out is unlikely to be triggered, much more efficient.
1607.Sp
1608Changing the timeout is trivial as well (if it isn't hard-coded in the
1609callback :) \- just change the timeout and invoke the callback, which will
1610fix things for you.
1611.IP "4. Wee, just use a double-linked list for your timeouts." 4
1612.IX Item "4. Wee, just use a double-linked list for your timeouts."
1613If there is not one request, but many thousands (millions...), all
1614employing some kind of timeout with the same timeout value, then one can
1615do even better:
1616.Sp
1617When starting the timeout, calculate the timeout value and put the timeout
1618at the \fIend\fR of the list.
1619.Sp
1620Then use an \f(CW\*(C`ev_timer\*(C'\fR to fire when the timeout at the \fIbeginning\fR of
1621the list is expected to fire (for example, using the technique #3).
1622.Sp
1623When there is some activity, remove the timer from the list, recalculate
1624the timeout, append it to the end of the list again, and make sure to
1625update the \f(CW\*(C`ev_timer\*(C'\fR if it was taken from the beginning of the list.
1626.Sp
1627This way, one can manage an unlimited number of timeouts in O(1) time for
1628starting, stopping and updating the timers, at the expense of a major
1629complication, and having to use a constant timeout. The constant timeout
1630ensures that the list stays sorted.
1631.PP
1632So which method the best?
1633.PP
1634Method #2 is a simple no-brain-required solution that is adequate in most
1635situations. Method #3 requires a bit more thinking, but handles many cases
1636better, and isn't very complicated either. In most case, choosing either
1637one is fine, with #3 being better in typical situations.
1638.PP
1639Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
1640rather complicated, but extremely efficient, something that really pays
1641off after the first million or so of active timers, i.e. it's usually
1642overkill :)
1643.PP
1644\fIThe special problem of time updates\fR
1645.IX Subsection "The special problem of time updates"
1646.PP
1647Establishing the current time is a costly operation (it usually takes at
1648least two system calls): \s-1EV\s0 therefore updates its idea of the current
1649time only before and after \f(CW\*(C`ev_loop\*(C'\fR collects new events, which causes a
1650growing difference between \f(CW\*(C`ev_now ()\*(C'\fR and \f(CW\*(C`ev_time ()\*(C'\fR when handling
1651lots of events in one iteration.
903.PP 1652.PP
904The relative timeouts are calculated relative to the \f(CW\*(C`ev_now ()\*(C'\fR 1653The relative timeouts are calculated relative to the \f(CW\*(C`ev_now ()\*(C'\fR
905time. This is usually the right thing as this timestamp refers to the time 1654time. This is usually the right thing as this timestamp refers to the time
906of the event triggering whatever timeout you are modifying/starting. If 1655of the event triggering whatever timeout you are modifying/starting. If
907you suspect event processing to be delayed and you \fIneed\fR to base the timeout 1656you suspect event processing to be delayed and you \fIneed\fR to base the
908on the current time, use something like this to adjust for this: 1657timeout on the current time, use something like this to adjust for this:
909.PP 1658.PP
910.Vb 1 1659.Vb 1
911\& ev_timer_set (&timer, after + ev_now () - ev_time (), 0.); 1660\& ev_timer_set (&timer, after + ev_now () \- ev_time (), 0.);
912.Ve 1661.Ve
913.PP 1662.PP
914The callback is guarenteed to be invoked only when its timeout has passed, 1663If the event loop is suspended for a long time, you can also force an
915but if multiple timers become ready during the same loop iteration then 1664update of the time returned by \f(CW\*(C`ev_now ()\*(C'\fR by calling \f(CW\*(C`ev_now_update
916order of execution is undefined. 1665()\*(C'\fR.
1666.PP
1667\fIWatcher-Specific Functions and Data Members\fR
1668.IX Subsection "Watcher-Specific Functions and Data Members"
917.IP "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)" 4 1669.IP "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)" 4
918.IX Item "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)" 1670.IX Item "ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)"
919.PD 0 1671.PD 0
920.IP "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)" 4 1672.IP "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)" 4
921.IX Item "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)" 1673.IX Item "ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)"
922.PD 1674.PD
923Configure the timer to trigger after \f(CW\*(C`after\*(C'\fR seconds. If \f(CW\*(C`repeat\*(C'\fR is 1675Configure the timer to trigger after \f(CW\*(C`after\*(C'\fR seconds. If \f(CW\*(C`repeat\*(C'\fR
924\&\f(CW0.\fR, then it will automatically be stopped. If it is positive, then the 1676is \f(CW0.\fR, then it will automatically be stopped once the timeout is
925timer will automatically be configured to trigger again \f(CW\*(C`repeat\*(C'\fR seconds 1677reached. If it is positive, then the timer will automatically be
926later, again, and again, until stopped manually. 1678configured to trigger again \f(CW\*(C`repeat\*(C'\fR seconds later, again, and again,
1679until stopped manually.
927.Sp 1680.Sp
928The timer itself will do a best-effort at avoiding drift, that is, if you 1681The timer itself will do a best-effort at avoiding drift, that is, if
929configure a timer to trigger every 10 seconds, then it will trigger at 1682you configure a timer to trigger every 10 seconds, then it will normally
930exactly 10 second intervals. If, however, your program cannot keep up with 1683trigger at exactly 10 second intervals. If, however, your program cannot
931the timer (because it takes longer than those 10 seconds to do stuff) the 1684keep up with the timer (because it takes longer than those 10 seconds to
932timer will not fire more than once per event loop iteration. 1685do stuff) the timer will not fire more than once per event loop iteration.
933.IP "ev_timer_again (loop)" 4 1686.IP "ev_timer_again (loop, ev_timer *)" 4
934.IX Item "ev_timer_again (loop)" 1687.IX Item "ev_timer_again (loop, ev_timer *)"
935This will act as if the timer timed out and restart it again if it is 1688This will act as if the timer timed out and restart it again if it is
936repeating. The exact semantics are: 1689repeating. The exact semantics are:
937.Sp 1690.Sp
1691If the timer is pending, its pending status is cleared.
1692.Sp
938If the timer is started but nonrepeating, stop it. 1693If the timer is started but non-repeating, stop it (as if it timed out).
939.Sp 1694.Sp
940If the timer is repeating, either start it if necessary (with the repeat 1695If the timer is repeating, either start it if necessary (with the
941value), or reset the running timer to the repeat value. 1696\&\f(CW\*(C`repeat\*(C'\fR value), or reset the running timer to the \f(CW\*(C`repeat\*(C'\fR value.
942.Sp 1697.Sp
943This sounds a bit complicated, but here is a useful and typical 1698This sounds a bit complicated, see \*(L"Be smart about timeouts\*(R", above, for a
944example: Imagine you have a tcp connection and you want a so-called idle 1699usage example.
945timeout, that is, you want to be called when there have been, say, 60 1700.IP "ev_tstamp repeat [read\-write]" 4
946seconds of inactivity on the socket. The easiest way to do this is to 1701.IX Item "ev_tstamp repeat [read-write]"
947configure an \f(CW\*(C`ev_timer\*(C'\fR with after=repeat=60 and calling ev_timer_again each 1702The current \f(CW\*(C`repeat\*(C'\fR value. Will be used each time the watcher times out
948time you successfully read or write some data. If you go into an idle 1703or \f(CW\*(C`ev_timer_again\*(C'\fR is called, and determines the next timeout (if any),
949state where you do not expect data to travel on the socket, you can stop 1704which is also when any modifications are taken into account.
950the timer, and again will automatically restart it if need be.
951.PP 1705.PP
1706\fIExamples\fR
1707.IX Subsection "Examples"
1708.PP
952Example: create a timer that fires after 60 seconds. 1709Example: Create a timer that fires after 60 seconds.
953.PP 1710.PP
954.Vb 5 1711.Vb 5
955\& static void 1712\& static void
956\& one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents) 1713\& one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
957\& { 1714\& {
958\& .. one minute over, w is actually stopped right here 1715\& .. one minute over, w is actually stopped right here
959\& } 1716\& }
960.Ve 1717\&
961.PP
962.Vb 3
963\& struct ev_timer mytimer; 1718\& ev_timer mytimer;
964\& ev_timer_init (&mytimer, one_minute_cb, 60., 0.); 1719\& ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
965\& ev_timer_start (loop, &mytimer); 1720\& ev_timer_start (loop, &mytimer);
966.Ve 1721.Ve
967.PP 1722.PP
968Example: create a timeout timer that times out after 10 seconds of 1723Example: Create a timeout timer that times out after 10 seconds of
969inactivity. 1724inactivity.
970.PP 1725.PP
971.Vb 5 1726.Vb 5
972\& static void 1727\& static void
973\& timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents) 1728\& timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
974\& { 1729\& {
975\& .. ten seconds without any activity 1730\& .. ten seconds without any activity
976\& } 1731\& }
977.Ve 1732\&
978.PP
979.Vb 4
980\& struct ev_timer mytimer; 1733\& ev_timer mytimer;
981\& ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */ 1734\& ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
982\& ev_timer_again (&mytimer); /* start timer */ 1735\& ev_timer_again (&mytimer); /* start timer */
983\& ev_loop (loop, 0); 1736\& ev_loop (loop, 0);
984.Ve 1737\&
985.PP
986.Vb 3
987\& // and in some piece of code that gets executed on any "activity": 1738\& // and in some piece of code that gets executed on any "activity":
988\& // reset the timeout to start ticking again at 10 seconds 1739\& // reset the timeout to start ticking again at 10 seconds
989\& ev_timer_again (&mytimer); 1740\& ev_timer_again (&mytimer);
990.Ve 1741.Ve
991.ie n .Sh """ev_periodic"" \- to cron or not to cron" 1742.ie n .Sh """ev_periodic"" \- to cron or not to cron?"
992.el .Sh "\f(CWev_periodic\fP \- to cron or not to cron" 1743.el .Sh "\f(CWev_periodic\fP \- to cron or not to cron?"
993.IX Subsection "ev_periodic - to cron or not to cron" 1744.IX Subsection "ev_periodic - to cron or not to cron?"
994Periodic watchers are also timers of a kind, but they are very versatile 1745Periodic watchers are also timers of a kind, but they are very versatile
995(and unfortunately a bit complex). 1746(and unfortunately a bit complex).
996.PP 1747.PP
997Unlike \f(CW\*(C`ev_timer\*(C'\fR's, they are not based on real time (or relative time) 1748Unlike \f(CW\*(C`ev_timer\*(C'\fR's, they are not based on real time (or relative time)
998but on wallclock time (absolute time). You can tell a periodic watcher 1749but on wall clock time (absolute time). You can tell a periodic watcher
999to trigger \*(L"at\*(R" some specific point in time. For example, if you tell a 1750to trigger after some specific point in time. For example, if you tell a
1000periodic watcher to trigger in 10 seconds (by specifiying e.g. \f(CW\*(C`ev_now () 1751periodic watcher to trigger in 10 seconds (by specifying e.g. \f(CW\*(C`ev_now ()
1001+ 10.\*(C'\fR) and then reset your system clock to the last year, then it will 1752+ 10.\*(C'\fR, that is, an absolute time not a delay) and then reset your system
1753clock to January of the previous year, then it will take more than year
1002take a year to trigger the event (unlike an \f(CW\*(C`ev_timer\*(C'\fR, which would trigger 1754to trigger the event (unlike an \f(CW\*(C`ev_timer\*(C'\fR, which would still trigger
1003roughly 10 seconds later and of course not if you reset your system time 1755roughly 10 seconds later as it uses a relative timeout).
1004again).
1005.PP 1756.PP
1006They can also be used to implement vastly more complex timers, such as 1757\&\f(CW\*(C`ev_periodic\*(C'\fRs can also be used to implement vastly more complex timers,
1007triggering an event on eahc midnight, local time. 1758such as triggering an event on each \*(L"midnight, local time\*(R", or other
1759complicated rules.
1008.PP 1760.PP
1009As with timers, the callback is guarenteed to be invoked only when the 1761As with timers, the callback is guaranteed to be invoked only when the
1010time (\f(CW\*(C`at\*(C'\fR) has been passed, but if multiple periodic timers become ready 1762time (\f(CW\*(C`at\*(C'\fR) has passed, but if multiple periodic timers become ready
1011during the same loop iteration then order of execution is undefined. 1763during the same loop iteration, then order of execution is undefined.
1764.PP
1765\fIWatcher-Specific Functions and Data Members\fR
1766.IX Subsection "Watcher-Specific Functions and Data Members"
1012.IP "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)" 4 1767.IP "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)" 4
1013.IX Item "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)" 1768.IX Item "ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)"
1014.PD 0 1769.PD 0
1015.IP "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)" 4 1770.IP "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)" 4
1016.IX Item "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)" 1771.IX Item "ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)"
1017.PD 1772.PD
1018Lots of arguments, lets sort it out... There are basically three modes of 1773Lots of arguments, lets sort it out... There are basically three modes of
1019operation, and we will explain them from simplest to complex: 1774operation, and we will explain them from simplest to most complex:
1020.RS 4 1775.RS 4
1776.IP "\(bu" 4
1021.IP "* absolute timer (interval = reschedule_cb = 0)" 4 1777absolute timer (at = time, interval = reschedule_cb = 0)
1022.IX Item "absolute timer (interval = reschedule_cb = 0)" 1778.Sp
1023In this configuration the watcher triggers an event at the wallclock time 1779In this configuration the watcher triggers an event after the wall clock
1024\&\f(CW\*(C`at\*(C'\fR and doesn't repeat. It will not adjust when a time jump occurs, 1780time \f(CW\*(C`at\*(C'\fR has passed. It will not repeat and will not adjust when a time
1025that is, if it is to be run at January 1st 2011 then it will run when the 1781jump occurs, that is, if it is to be run at January 1st 2011 then it will
1026system time reaches or surpasses this time. 1782only run when the system clock reaches or surpasses this time.
1783.IP "\(bu" 4
1027.IP "* non-repeating interval timer (interval > 0, reschedule_cb = 0)" 4 1784repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1028.IX Item "non-repeating interval timer (interval > 0, reschedule_cb = 0)" 1785.Sp
1029In this mode the watcher will always be scheduled to time out at the next 1786In this mode the watcher will always be scheduled to time out at the next
1030\&\f(CW\*(C`at + N * interval\*(C'\fR time (for some integer N) and then repeat, regardless 1787\&\f(CW\*(C`at + N * interval\*(C'\fR time (for some integer N, which can also be negative)
1031of any time jumps. 1788and then repeat, regardless of any time jumps.
1032.Sp 1789.Sp
1033This can be used to create timers that do not drift with respect to system 1790This can be used to create timers that do not drift with respect to the
1034time: 1791system clock, for example, here is a \f(CW\*(C`ev_periodic\*(C'\fR that triggers each
1792hour, on the hour:
1035.Sp 1793.Sp
1036.Vb 1 1794.Vb 1
1037\& ev_periodic_set (&periodic, 0., 3600., 0); 1795\& ev_periodic_set (&periodic, 0., 3600., 0);
1038.Ve 1796.Ve
1039.Sp 1797.Sp
1040This doesn't mean there will always be 3600 seconds in between triggers, 1798This doesn't mean there will always be 3600 seconds in between triggers,
1041but only that the the callback will be called when the system time shows a 1799but only that the callback will be called when the system time shows a
1042full hour (\s-1UTC\s0), or more correctly, when the system time is evenly divisible 1800full hour (\s-1UTC\s0), or more correctly, when the system time is evenly divisible
1043by 3600. 1801by 3600.
1044.Sp 1802.Sp
1045Another way to think about it (for the mathematically inclined) is that 1803Another way to think about it (for the mathematically inclined) is that
1046\&\f(CW\*(C`ev_periodic\*(C'\fR will try to run the callback in this mode at the next possible 1804\&\f(CW\*(C`ev_periodic\*(C'\fR will try to run the callback in this mode at the next possible
1047time where \f(CW\*(C`time = at (mod interval)\*(C'\fR, regardless of any time jumps. 1805time where \f(CW\*(C`time = at (mod interval)\*(C'\fR, regardless of any time jumps.
1048.IP "* manual reschedule mode (reschedule_cb = callback)" 4 1806.Sp
1049.IX Item "manual reschedule mode (reschedule_cb = callback)" 1807For numerical stability it is preferable that the \f(CW\*(C`at\*(C'\fR value is near
1808\&\f(CW\*(C`ev_now ()\*(C'\fR (the current time), but there is no range requirement for
1809this value, and in fact is often specified as zero.
1810.Sp
1811Note also that there is an upper limit to how often a timer can fire (\s-1CPU\s0
1812speed for example), so if \f(CW\*(C`interval\*(C'\fR is very small then timing stability
1813will of course deteriorate. Libev itself tries to be exact to be about one
1814millisecond (if the \s-1OS\s0 supports it and the machine is fast enough).
1815.IP "\(bu" 4
1816manual reschedule mode (at and interval ignored, reschedule_cb = callback)
1817.Sp
1050In this mode the values for \f(CW\*(C`interval\*(C'\fR and \f(CW\*(C`at\*(C'\fR are both being 1818In this mode the values for \f(CW\*(C`interval\*(C'\fR and \f(CW\*(C`at\*(C'\fR are both being
1051ignored. Instead, each time the periodic watcher gets scheduled, the 1819ignored. Instead, each time the periodic watcher gets scheduled, the
1052reschedule callback will be called with the watcher as first, and the 1820reschedule callback will be called with the watcher as first, and the
1053current time as second argument. 1821current time as second argument.
1054.Sp 1822.Sp
1055\&\s-1NOTE:\s0 \fIThis callback \s-1MUST\s0 \s-1NOT\s0 stop or destroy any periodic watcher, 1823\&\s-1NOTE:\s0 \fIThis callback \s-1MUST\s0 \s-1NOT\s0 stop or destroy any periodic watcher,
1056ever, or make any event loop modifications\fR. If you need to stop it, 1824ever, or make \s-1ANY\s0 event loop modifications whatsoever\fR.
1057return \f(CW\*(C`now + 1e30\*(C'\fR (or so, fudge fudge) and stop it afterwards (e.g. by
1058starting a prepare watcher).
1059.Sp 1825.Sp
1826If you need to stop it, return \f(CW\*(C`now + 1e30\*(C'\fR (or so, fudge fudge) and stop
1827it afterwards (e.g. by starting an \f(CW\*(C`ev_prepare\*(C'\fR watcher, which is the
1828only event loop modification you are allowed to do).
1829.Sp
1060Its prototype is \f(CW\*(C`ev_tstamp (*reschedule_cb)(struct ev_periodic *w, 1830The callback prototype is \f(CW\*(C`ev_tstamp (*reschedule_cb)(ev_periodic
1061ev_tstamp now)\*(C'\fR, e.g.: 1831*w, ev_tstamp now)\*(C'\fR, e.g.:
1062.Sp 1832.Sp
1063.Vb 4 1833.Vb 5
1834\& static ev_tstamp
1064\& static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now) 1835\& my_rescheduler (ev_periodic *w, ev_tstamp now)
1065\& { 1836\& {
1066\& return now + 60.; 1837\& return now + 60.;
1067\& } 1838\& }
1068.Ve 1839.Ve
1069.Sp 1840.Sp
1070It must return the next time to trigger, based on the passed time value 1841It must return the next time to trigger, based on the passed time value
1071(that is, the lowest time value larger than to the second argument). It 1842(that is, the lowest time value larger than to the second argument). It
1072will usually be called just before the callback will be triggered, but 1843will usually be called just before the callback will be triggered, but
1073might be called at other times, too. 1844might be called at other times, too.
1074.Sp 1845.Sp
1075\&\s-1NOTE:\s0 \fIThis callback must always return a time that is later than the 1846\&\s-1NOTE:\s0 \fIThis callback must always return a time that is higher than or
1076passed \f(CI\*(C`now\*(C'\fI value\fR. Not even \f(CW\*(C`now\*(C'\fR itself will do, it \fImust\fR be larger. 1847equal to the passed \f(CI\*(C`now\*(C'\fI value\fR.
1077.Sp 1848.Sp
1078This can be used to create very complex timers, such as a timer that 1849This can be used to create very complex timers, such as a timer that
1079triggers on each midnight, local time. To do this, you would calculate the 1850triggers on \*(L"next midnight, local time\*(R". To do this, you would calculate the
1080next midnight after \f(CW\*(C`now\*(C'\fR and return the timestamp value for this. How 1851next midnight after \f(CW\*(C`now\*(C'\fR and return the timestamp value for this. How
1081you do this is, again, up to you (but it is not trivial, which is the main 1852you do this is, again, up to you (but it is not trivial, which is the main
1082reason I omitted it as an example). 1853reason I omitted it as an example).
1083.RE 1854.RE
1084.RS 4 1855.RS 4
… …
1087.IX Item "ev_periodic_again (loop, ev_periodic *)" 1858.IX Item "ev_periodic_again (loop, ev_periodic *)"
1088Simply stops and restarts the periodic watcher again. This is only useful 1859Simply stops and restarts the periodic watcher again. This is only useful
1089when you changed some parameters or the reschedule callback would return 1860when you changed some parameters or the reschedule callback would return
1090a different time than the last time it was called (e.g. in a crond like 1861a different time than the last time it was called (e.g. in a crond like
1091program when the crontabs have changed). 1862program when the crontabs have changed).
1863.IP "ev_tstamp ev_periodic_at (ev_periodic *)" 4
1864.IX Item "ev_tstamp ev_periodic_at (ev_periodic *)"
1865When active, returns the absolute time that the watcher is supposed to
1866trigger next.
1867.IP "ev_tstamp offset [read\-write]" 4
1868.IX Item "ev_tstamp offset [read-write]"
1869When repeating, this contains the offset value, otherwise this is the
1870absolute point in time (the \f(CW\*(C`at\*(C'\fR value passed to \f(CW\*(C`ev_periodic_set\*(C'\fR).
1871.Sp
1872Can be modified any time, but changes only take effect when the periodic
1873timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being called.
1874.IP "ev_tstamp interval [read\-write]" 4
1875.IX Item "ev_tstamp interval [read-write]"
1876The current interval value. Can be modified any time, but changes only
1877take effect when the periodic timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being
1878called.
1879.IP "ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read\-write]" 4
1880.IX Item "ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]"
1881The current reschedule callback, or \f(CW0\fR, if this functionality is
1882switched off. Can be changed any time, but changes only take effect when
1883the periodic timer fires or \f(CW\*(C`ev_periodic_again\*(C'\fR is being called.
1092.PP 1884.PP
1885\fIExamples\fR
1886.IX Subsection "Examples"
1887.PP
1093Example: call a callback every hour, or, more precisely, whenever the 1888Example: Call a callback every hour, or, more precisely, whenever the
1094system clock is divisible by 3600. The callback invocation times have 1889system time is divisible by 3600. The callback invocation times have
1095potentially a lot of jittering, but good long-term stability. 1890potentially a lot of jitter, but good long-term stability.
1096.PP 1891.PP
1097.Vb 5 1892.Vb 5
1098\& static void 1893\& static void
1099\& clock_cb (struct ev_loop *loop, struct ev_io *w, int revents) 1894\& clock_cb (struct ev_loop *loop, ev_io *w, int revents)
1100\& { 1895\& {
1101\& ... its now a full hour (UTC, or TAI or whatever your clock follows) 1896\& ... its now a full hour (UTC, or TAI or whatever your clock follows)
1102\& } 1897\& }
1103.Ve 1898\&
1104.PP
1105.Vb 3
1106\& struct ev_periodic hourly_tick; 1899\& ev_periodic hourly_tick;
1107\& ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0); 1900\& ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1108\& ev_periodic_start (loop, &hourly_tick); 1901\& ev_periodic_start (loop, &hourly_tick);
1109.Ve 1902.Ve
1110.PP 1903.PP
1111Example: the same as above, but use a reschedule callback to do it: 1904Example: The same as above, but use a reschedule callback to do it:
1112.PP 1905.PP
1113.Vb 1 1906.Vb 1
1114\& #include <math.h> 1907\& #include <math.h>
1115.Ve 1908\&
1116.PP
1117.Vb 5
1118\& static ev_tstamp 1909\& static ev_tstamp
1119\& my_scheduler_cb (struct ev_periodic *w, ev_tstamp now) 1910\& my_scheduler_cb (ev_periodic *w, ev_tstamp now)
1120\& { 1911\& {
1121\& return fmod (now, 3600.) + 3600.; 1912\& return now + (3600. \- fmod (now, 3600.));
1122\& } 1913\& }
1123.Ve 1914\&
1124.PP
1125.Vb 1
1126\& ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb); 1915\& ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1127.Ve 1916.Ve
1128.PP 1917.PP
1129Example: call a callback every hour, starting now: 1918Example: Call a callback every hour, starting now:
1130.PP 1919.PP
1131.Vb 4 1920.Vb 4
1132\& struct ev_periodic hourly_tick; 1921\& ev_periodic hourly_tick;
1133\& ev_periodic_init (&hourly_tick, clock_cb, 1922\& ev_periodic_init (&hourly_tick, clock_cb,
1134\& fmod (ev_now (loop), 3600.), 3600., 0); 1923\& fmod (ev_now (loop), 3600.), 3600., 0);
1135\& ev_periodic_start (loop, &hourly_tick); 1924\& ev_periodic_start (loop, &hourly_tick);
1136.Ve 1925.Ve
1137.ie n .Sh """ev_signal"" \- signal me when a signal gets signalled" 1926.ie n .Sh """ev_signal"" \- signal me when a signal gets signalled!"
1138.el .Sh "\f(CWev_signal\fP \- signal me when a signal gets signalled" 1927.el .Sh "\f(CWev_signal\fP \- signal me when a signal gets signalled!"
1139.IX Subsection "ev_signal - signal me when a signal gets signalled" 1928.IX Subsection "ev_signal - signal me when a signal gets signalled!"
1140Signal watchers will trigger an event when the process receives a specific 1929Signal watchers will trigger an event when the process receives a specific
1141signal one or more times. Even though signals are very asynchronous, libev 1930signal one or more times. Even though signals are very asynchronous, libev
1142will try it's best to deliver signals synchronously, i.e. as part of the 1931will try it's best to deliver signals synchronously, i.e. as part of the
1143normal event processing, like any other event. 1932normal event processing, like any other event.
1144.PP 1933.PP
1934If you want signals asynchronously, just use \f(CW\*(C`sigaction\*(C'\fR as you would
1935do without libev and forget about sharing the signal. You can even use
1936\&\f(CW\*(C`ev_async\*(C'\fR from a signal handler to synchronously wake up an event loop.
1937.PP
1145You can configure as many watchers as you like per signal. Only when the 1938You can configure as many watchers as you like per signal. Only when the
1146first watcher gets started will libev actually register a signal watcher 1939first watcher gets started will libev actually register a signal handler
1147with the kernel (thus it coexists with your own signal handlers as long 1940with the kernel (thus it coexists with your own signal handlers as long as
1148as you don't register any with libev). Similarly, when the last signal 1941you don't register any with libev for the same signal). Similarly, when
1149watcher for a signal is stopped libev will reset the signal handler to 1942the last signal watcher for a signal is stopped, libev will reset the
1150\&\s-1SIG_DFL\s0 (regardless of what it was set to before). 1943signal handler to \s-1SIG_DFL\s0 (regardless of what it was set to before).
1944.PP
1945If possible and supported, libev will install its handlers with
1946\&\f(CW\*(C`SA_RESTART\*(C'\fR behaviour enabled, so system calls should not be unduly
1947interrupted. If you have a problem with system calls getting interrupted by
1948signals you can block all signals in an \f(CW\*(C`ev_check\*(C'\fR watcher and unblock
1949them in an \f(CW\*(C`ev_prepare\*(C'\fR watcher.
1950.PP
1951\fIWatcher-Specific Functions and Data Members\fR
1952.IX Subsection "Watcher-Specific Functions and Data Members"
1151.IP "ev_signal_init (ev_signal *, callback, int signum)" 4 1953.IP "ev_signal_init (ev_signal *, callback, int signum)" 4
1152.IX Item "ev_signal_init (ev_signal *, callback, int signum)" 1954.IX Item "ev_signal_init (ev_signal *, callback, int signum)"
1153.PD 0 1955.PD 0
1154.IP "ev_signal_set (ev_signal *, int signum)" 4 1956.IP "ev_signal_set (ev_signal *, int signum)" 4
1155.IX Item "ev_signal_set (ev_signal *, int signum)" 1957.IX Item "ev_signal_set (ev_signal *, int signum)"
1156.PD 1958.PD
1157Configures the watcher to trigger on the given signal number (usually one 1959Configures the watcher to trigger on the given signal number (usually one
1158of the \f(CW\*(C`SIGxxx\*(C'\fR constants). 1960of the \f(CW\*(C`SIGxxx\*(C'\fR constants).
1961.IP "int signum [read\-only]" 4
1962.IX Item "int signum [read-only]"
1963The signal the watcher watches out for.
1964.PP
1965\fIExamples\fR
1966.IX Subsection "Examples"
1967.PP
1968Example: Try to exit cleanly on \s-1SIGINT\s0.
1969.PP
1970.Vb 5
1971\& static void
1972\& sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
1973\& {
1974\& ev_unloop (loop, EVUNLOOP_ALL);
1975\& }
1976\&
1977\& ev_signal signal_watcher;
1978\& ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1979\& ev_signal_start (loop, &signal_watcher);
1980.Ve
1159.ie n .Sh """ev_child"" \- wait for pid status changes" 1981.ie n .Sh """ev_child"" \- watch out for process status changes"
1160.el .Sh "\f(CWev_child\fP \- wait for pid status changes" 1982.el .Sh "\f(CWev_child\fP \- watch out for process status changes"
1161.IX Subsection "ev_child - wait for pid status changes" 1983.IX Subsection "ev_child - watch out for process status changes"
1162Child watchers trigger when your process receives a \s-1SIGCHLD\s0 in response to 1984Child watchers trigger when your process receives a \s-1SIGCHLD\s0 in response to
1163some child status changes (most typically when a child of yours dies). 1985some child status changes (most typically when a child of yours dies or
1986exits). It is permissible to install a child watcher \fIafter\fR the child
1987has been forked (which implies it might have already exited), as long
1988as the event loop isn't entered (or is continued from a watcher), i.e.,
1989forking and then immediately registering a watcher for the child is fine,
1990but forking and registering a watcher a few event loop iterations later is
1991not.
1992.PP
1993Only the default event loop is capable of handling signals, and therefore
1994you can only register child watchers in the default event loop.
1995.PP
1996\fIProcess Interaction\fR
1997.IX Subsection "Process Interaction"
1998.PP
1999Libev grabs \f(CW\*(C`SIGCHLD\*(C'\fR as soon as the default event loop is
2000initialised. This is necessary to guarantee proper behaviour even if
2001the first child watcher is started after the child exits. The occurrence
2002of \f(CW\*(C`SIGCHLD\*(C'\fR is recorded asynchronously, but child reaping is done
2003synchronously as part of the event loop processing. Libev always reaps all
2004children, even ones not watched.
2005.PP
2006\fIOverriding the Built-In Processing\fR
2007.IX Subsection "Overriding the Built-In Processing"
2008.PP
2009Libev offers no special support for overriding the built-in child
2010processing, but if your application collides with libev's default child
2011handler, you can override it easily by installing your own handler for
2012\&\f(CW\*(C`SIGCHLD\*(C'\fR after initialising the default loop, and making sure the
2013default loop never gets destroyed. You are encouraged, however, to use an
2014event-based approach to child reaping and thus use libev's support for
2015that, so other libev users can use \f(CW\*(C`ev_child\*(C'\fR watchers freely.
2016.PP
2017\fIStopping the Child Watcher\fR
2018.IX Subsection "Stopping the Child Watcher"
2019.PP
2020Currently, the child watcher never gets stopped, even when the
2021child terminates, so normally one needs to stop the watcher in the
2022callback. Future versions of libev might stop the watcher automatically
2023when a child exit is detected.
2024.PP
2025\fIWatcher-Specific Functions and Data Members\fR
2026.IX Subsection "Watcher-Specific Functions and Data Members"
1164.IP "ev_child_init (ev_child *, callback, int pid)" 4 2027.IP "ev_child_init (ev_child *, callback, int pid, int trace)" 4
1165.IX Item "ev_child_init (ev_child *, callback, int pid)" 2028.IX Item "ev_child_init (ev_child *, callback, int pid, int trace)"
1166.PD 0 2029.PD 0
1167.IP "ev_child_set (ev_child *, int pid)" 4 2030.IP "ev_child_set (ev_child *, int pid, int trace)" 4
1168.IX Item "ev_child_set (ev_child *, int pid)" 2031.IX Item "ev_child_set (ev_child *, int pid, int trace)"
1169.PD 2032.PD
1170Configures the watcher to wait for status changes of process \f(CW\*(C`pid\*(C'\fR (or 2033Configures the watcher to wait for status changes of process \f(CW\*(C`pid\*(C'\fR (or
1171\&\fIany\fR process if \f(CW\*(C`pid\*(C'\fR is specified as \f(CW0\fR). The callback can look 2034\&\fIany\fR process if \f(CW\*(C`pid\*(C'\fR is specified as \f(CW0\fR). The callback can look
1172at the \f(CW\*(C`rstatus\*(C'\fR member of the \f(CW\*(C`ev_child\*(C'\fR watcher structure to see 2035at the \f(CW\*(C`rstatus\*(C'\fR member of the \f(CW\*(C`ev_child\*(C'\fR watcher structure to see
1173the status word (use the macros from \f(CW\*(C`sys/wait.h\*(C'\fR and see your systems 2036the status word (use the macros from \f(CW\*(C`sys/wait.h\*(C'\fR and see your systems
1174\&\f(CW\*(C`waitpid\*(C'\fR documentation). The \f(CW\*(C`rpid\*(C'\fR member contains the pid of the 2037\&\f(CW\*(C`waitpid\*(C'\fR documentation). The \f(CW\*(C`rpid\*(C'\fR member contains the pid of the
1175process causing the status change. 2038process causing the status change. \f(CW\*(C`trace\*(C'\fR must be either \f(CW0\fR (only
2039activate the watcher when the process terminates) or \f(CW1\fR (additionally
2040activate the watcher when the process is stopped or continued).
2041.IP "int pid [read\-only]" 4
2042.IX Item "int pid [read-only]"
2043The process id this watcher watches out for, or \f(CW0\fR, meaning any process id.
2044.IP "int rpid [read\-write]" 4
2045.IX Item "int rpid [read-write]"
2046The process id that detected a status change.
2047.IP "int rstatus [read\-write]" 4
2048.IX Item "int rstatus [read-write]"
2049The process exit/trace status caused by \f(CW\*(C`rpid\*(C'\fR (see your systems
2050\&\f(CW\*(C`waitpid\*(C'\fR and \f(CW\*(C`sys/wait.h\*(C'\fR documentation for details).
1176.PP 2051.PP
1177Example: try to exit cleanly on \s-1SIGINT\s0 and \s-1SIGTERM\s0. 2052\fIExamples\fR
2053.IX Subsection "Examples"
1178.PP 2054.PP
2055Example: \f(CW\*(C`fork()\*(C'\fR a new process and install a child handler to wait for
2056its completion.
2057.PP
1179.Vb 5 2058.Vb 1
2059\& ev_child cw;
2060\&
1180\& static void 2061\& static void
1181\& sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents) 2062\& child_cb (EV_P_ ev_child *w, int revents)
1182\& { 2063\& {
1183\& ev_unloop (loop, EVUNLOOP_ALL); 2064\& ev_child_stop (EV_A_ w);
2065\& printf ("process %d exited with status %x\en", w\->rpid, w\->rstatus);
1184\& } 2066\& }
2067\&
2068\& pid_t pid = fork ();
2069\&
2070\& if (pid < 0)
2071\& // error
2072\& else if (pid == 0)
2073\& {
2074\& // the forked child executes here
2075\& exit (1);
2076\& }
2077\& else
2078\& {
2079\& ev_child_init (&cw, child_cb, pid, 0);
2080\& ev_child_start (EV_DEFAULT_ &cw);
2081\& }
1185.Ve 2082.Ve
2083.ie n .Sh """ev_stat"" \- did the file attributes just change?"
2084.el .Sh "\f(CWev_stat\fP \- did the file attributes just change?"
2085.IX Subsection "ev_stat - did the file attributes just change?"
2086This watches a file system path for attribute changes. That is, it calls
2087\&\f(CW\*(C`stat\*(C'\fR on that path in regular intervals (or when the \s-1OS\s0 says it changed)
2088and sees if it changed compared to the last time, invoking the callback if
2089it did.
1186.PP 2090.PP
2091The path does not need to exist: changing from \*(L"path exists\*(R" to \*(L"path does
2092not exist\*(R" is a status change like any other. The condition \*(L"path does not
2093exist\*(R" (or more correctly \*(L"path cannot be stat'ed\*(R") is signified by the
2094\&\f(CW\*(C`st_nlink\*(C'\fR field being zero (which is otherwise always forced to be at
2095least one) and all the other fields of the stat buffer having unspecified
2096contents.
2097.PP
2098The path \fImust not\fR end in a slash or contain special components such as
2099\&\f(CW\*(C`.\*(C'\fR or \f(CW\*(C`..\*(C'\fR. The path \fIshould\fR be absolute: If it is relative and
2100your working directory changes, then the behaviour is undefined.
2101.PP
2102Since there is no portable change notification interface available, the
2103portable implementation simply calls \f(CWstat(2)\fR regularly on the path
2104to see if it changed somehow. You can specify a recommended polling
2105interval for this case. If you specify a polling interval of \f(CW0\fR (highly
2106recommended!) then a \fIsuitable, unspecified default\fR value will be used
2107(which you can expect to be around five seconds, although this might
2108change dynamically). Libev will also impose a minimum interval which is
2109currently around \f(CW0.1\fR, but that's usually overkill.
2110.PP
2111This watcher type is not meant for massive numbers of stat watchers,
2112as even with OS-supported change notifications, this can be
2113resource-intensive.
2114.PP
2115At the time of this writing, the only OS-specific interface implemented
2116is the Linux inotify interface (implementing kqueue support is left as an
2117exercise for the reader. Note, however, that the author sees no way of
2118implementing \f(CW\*(C`ev_stat\*(C'\fR semantics with kqueue, except as a hint).
2119.PP
2120\fI\s-1ABI\s0 Issues (Largefile Support)\fR
2121.IX Subsection "ABI Issues (Largefile Support)"
2122.PP
2123Libev by default (unless the user overrides this) uses the default
2124compilation environment, which means that on systems with large file
2125support disabled by default, you get the 32 bit version of the stat
2126structure. When using the library from programs that change the \s-1ABI\s0 to
2127use 64 bit file offsets the programs will fail. In that case you have to
2128compile libev with the same flags to get binary compatibility. This is
2129obviously the case with any flags that change the \s-1ABI\s0, but the problem is
2130most noticeably displayed with ev_stat and large file support.
2131.PP
2132The solution for this is to lobby your distribution maker to make large
2133file interfaces available by default (as e.g. FreeBSD does) and not
2134optional. Libev cannot simply switch on large file support because it has
2135to exchange stat structures with application programs compiled using the
2136default compilation environment.
2137.PP
2138\fIInotify and Kqueue\fR
2139.IX Subsection "Inotify and Kqueue"
2140.PP
2141When \f(CW\*(C`inotify (7)\*(C'\fR support has been compiled into libev and present at
2142runtime, it will be used to speed up change detection where possible. The
2143inotify descriptor will be created lazily when the first \f(CW\*(C`ev_stat\*(C'\fR
2144watcher is being started.
2145.PP
2146Inotify presence does not change the semantics of \f(CW\*(C`ev_stat\*(C'\fR watchers
2147except that changes might be detected earlier, and in some cases, to avoid
2148making regular \f(CW\*(C`stat\*(C'\fR calls. Even in the presence of inotify support
2149there are many cases where libev has to resort to regular \f(CW\*(C`stat\*(C'\fR polling,
2150but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
2151many bugs), the path exists (i.e. stat succeeds), and the path resides on
2152a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
2153xfs are fully working) libev usually gets away without polling.
2154.PP
2155There is no support for kqueue, as apparently it cannot be used to
2156implement this functionality, due to the requirement of having a file
2157descriptor open on the object at all times, and detecting renames, unlinks
2158etc. is difficult.
2159.PP
2160\fI\f(CI\*(C`stat ()\*(C'\fI is a synchronous operation\fR
2161.IX Subsection "stat () is a synchronous operation"
2162.PP
2163Libev doesn't normally do any kind of I/O itself, and so is not blocking
2164the process. The exception are \f(CW\*(C`ev_stat\*(C'\fR watchers \- those call \f(CW\*(C`stat
2165()\*(C'\fR, which is a synchronous operation.
2166.PP
2167For local paths, this usually doesn't matter: unless the system is very
2168busy or the intervals between stat's are large, a stat call will be fast,
2169as the path data is usually in memory already (except when starting the
2170watcher).
2171.PP
2172For networked file systems, calling \f(CW\*(C`stat ()\*(C'\fR can block an indefinite
2173time due to network issues, and even under good conditions, a stat call
2174often takes multiple milliseconds.
2175.PP
2176Therefore, it is best to avoid using \f(CW\*(C`ev_stat\*(C'\fR watchers on networked
2177paths, although this is fully supported by libev.
2178.PP
2179\fIThe special problem of stat time resolution\fR
2180.IX Subsection "The special problem of stat time resolution"
2181.PP
2182The \f(CW\*(C`stat ()\*(C'\fR system call only supports full-second resolution portably,
2183and even on systems where the resolution is higher, most file systems
2184still only support whole seconds.
2185.PP
2186That means that, if the time is the only thing that changes, you can
2187easily miss updates: on the first update, \f(CW\*(C`ev_stat\*(C'\fR detects a change and
2188calls your callback, which does something. When there is another update
2189within the same second, \f(CW\*(C`ev_stat\*(C'\fR will be unable to detect unless the
2190stat data does change in other ways (e.g. file size).
2191.PP
2192The solution to this is to delay acting on a change for slightly more
2193than a second (or till slightly after the next full second boundary), using
2194a roughly one-second-delay \f(CW\*(C`ev_timer\*(C'\fR (e.g. \f(CW\*(C`ev_timer_set (w, 0., 1.02);
2195ev_timer_again (loop, w)\*(C'\fR).
2196.PP
2197The \f(CW.02\fR offset is added to work around small timing inconsistencies
2198of some operating systems (where the second counter of the current time
2199might be be delayed. One such system is the Linux kernel, where a call to
2200\&\f(CW\*(C`gettimeofday\*(C'\fR might return a timestamp with a full second later than
2201a subsequent \f(CW\*(C`time\*(C'\fR call \- if the equivalent of \f(CW\*(C`time ()\*(C'\fR is used to
2202update file times then there will be a small window where the kernel uses
2203the previous second to update file times but libev might already execute
2204the timer callback).
2205.PP
2206\fIWatcher-Specific Functions and Data Members\fR
2207.IX Subsection "Watcher-Specific Functions and Data Members"
2208.IP "ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)" 4
2209.IX Item "ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)"
2210.PD 0
2211.IP "ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)" 4
2212.IX Item "ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)"
2213.PD
2214Configures the watcher to wait for status changes of the given
2215\&\f(CW\*(C`path\*(C'\fR. The \f(CW\*(C`interval\*(C'\fR is a hint on how quickly a change is expected to
2216be detected and should normally be specified as \f(CW0\fR to let libev choose
2217a suitable value. The memory pointed to by \f(CW\*(C`path\*(C'\fR must point to the same
2218path for as long as the watcher is active.
2219.Sp
2220The callback will receive an \f(CW\*(C`EV_STAT\*(C'\fR event when a change was detected,
2221relative to the attributes at the time the watcher was started (or the
2222last change was detected).
2223.IP "ev_stat_stat (loop, ev_stat *)" 4
2224.IX Item "ev_stat_stat (loop, ev_stat *)"
2225Updates the stat buffer immediately with new values. If you change the
2226watched path in your callback, you could call this function to avoid
2227detecting this change (while introducing a race condition if you are not
2228the only one changing the path). Can also be useful simply to find out the
2229new values.
2230.IP "ev_statdata attr [read\-only]" 4
2231.IX Item "ev_statdata attr [read-only]"
2232The most-recently detected attributes of the file. Although the type is
2233\&\f(CW\*(C`ev_statdata\*(C'\fR, this is usually the (or one of the) \f(CW\*(C`struct stat\*(C'\fR types
2234suitable for your system, but you can only rely on the POSIX-standardised
2235members to be present. If the \f(CW\*(C`st_nlink\*(C'\fR member is \f(CW0\fR, then there was
2236some error while \f(CW\*(C`stat\*(C'\fRing the file.
2237.IP "ev_statdata prev [read\-only]" 4
2238.IX Item "ev_statdata prev [read-only]"
2239The previous attributes of the file. The callback gets invoked whenever
2240\&\f(CW\*(C`prev\*(C'\fR != \f(CW\*(C`attr\*(C'\fR, or, more precisely, one or more of these members
2241differ: \f(CW\*(C`st_dev\*(C'\fR, \f(CW\*(C`st_ino\*(C'\fR, \f(CW\*(C`st_mode\*(C'\fR, \f(CW\*(C`st_nlink\*(C'\fR, \f(CW\*(C`st_uid\*(C'\fR,
2242\&\f(CW\*(C`st_gid\*(C'\fR, \f(CW\*(C`st_rdev\*(C'\fR, \f(CW\*(C`st_size\*(C'\fR, \f(CW\*(C`st_atime\*(C'\fR, \f(CW\*(C`st_mtime\*(C'\fR, \f(CW\*(C`st_ctime\*(C'\fR.
2243.IP "ev_tstamp interval [read\-only]" 4
2244.IX Item "ev_tstamp interval [read-only]"
2245The specified interval.
2246.IP "const char *path [read\-only]" 4
2247.IX Item "const char *path [read-only]"
2248The file system path that is being watched.
2249.PP
2250\fIExamples\fR
2251.IX Subsection "Examples"
2252.PP
2253Example: Watch \f(CW\*(C`/etc/passwd\*(C'\fR for attribute changes.
2254.PP
2255.Vb 10
2256\& static void
2257\& passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
2258\& {
2259\& /* /etc/passwd changed in some way */
2260\& if (w\->attr.st_nlink)
2261\& {
2262\& printf ("passwd current size %ld\en", (long)w\->attr.st_size);
2263\& printf ("passwd current atime %ld\en", (long)w\->attr.st_mtime);
2264\& printf ("passwd current mtime %ld\en", (long)w\->attr.st_mtime);
2265\& }
2266\& else
2267\& /* you shalt not abuse printf for puts */
2268\& puts ("wow, /etc/passwd is not there, expect problems. "
2269\& "if this is windows, they already arrived\en");
2270\& }
2271\&
2272\& ...
2273\& ev_stat passwd;
2274\&
2275\& ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
2276\& ev_stat_start (loop, &passwd);
2277.Ve
2278.PP
2279Example: Like above, but additionally use a one-second delay so we do not
2280miss updates (however, frequent updates will delay processing, too, so
2281one might do the work both on \f(CW\*(C`ev_stat\*(C'\fR callback invocation \fIand\fR on
2282\&\f(CW\*(C`ev_timer\*(C'\fR callback invocation).
2283.PP
1187.Vb 3 2284.Vb 2
1188\& struct ev_signal signal_watcher; 2285\& static ev_stat passwd;
1189\& ev_signal_init (&signal_watcher, sigint_cb, SIGINT); 2286\& static ev_timer timer;
1190\& ev_signal_start (loop, &sigint_cb); 2287\&
2288\& static void
2289\& timer_cb (EV_P_ ev_timer *w, int revents)
2290\& {
2291\& ev_timer_stop (EV_A_ w);
2292\&
2293\& /* now it\*(Aqs one second after the most recent passwd change */
2294\& }
2295\&
2296\& static void
2297\& stat_cb (EV_P_ ev_stat *w, int revents)
2298\& {
2299\& /* reset the one\-second timer */
2300\& ev_timer_again (EV_A_ &timer);
2301\& }
2302\&
2303\& ...
2304\& ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
2305\& ev_stat_start (loop, &passwd);
2306\& ev_timer_init (&timer, timer_cb, 0., 1.02);
1191.Ve 2307.Ve
1192.ie n .Sh """ev_idle"" \- when you've got nothing better to do" 2308.ie n .Sh """ev_idle"" \- when you've got nothing better to do..."
1193.el .Sh "\f(CWev_idle\fP \- when you've got nothing better to do" 2309.el .Sh "\f(CWev_idle\fP \- when you've got nothing better to do..."
1194.IX Subsection "ev_idle - when you've got nothing better to do" 2310.IX Subsection "ev_idle - when you've got nothing better to do..."
1195Idle watchers trigger events when there are no other events are pending 2311Idle watchers trigger events when no other events of the same or higher
1196(prepare, check and other idle watchers do not count). That is, as long 2312priority are pending (prepare, check and other idle watchers do not count
1197as your process is busy handling sockets or timeouts (or even signals, 2313as receiving \*(L"events\*(R").
1198imagine) it will not be triggered. But when your process is idle all idle 2314.PP
1199watchers are being called again and again, once per event loop iteration \- 2315That is, as long as your process is busy handling sockets or timeouts
2316(or even signals, imagine) of the same or higher priority it will not be
2317triggered. But when your process is idle (or only lower-priority watchers
2318are pending), the idle watchers are being called once per event loop
1200until stopped, that is, or your process receives more events and becomes 2319iteration \- until stopped, that is, or your process receives more events
1201busy. 2320and becomes busy again with higher priority stuff.
1202.PP 2321.PP
1203The most noteworthy effect is that as long as any idle watchers are 2322The most noteworthy effect is that as long as any idle watchers are
1204active, the process will not block when waiting for new events. 2323active, the process will not block when waiting for new events.
1205.PP 2324.PP
1206Apart from keeping your process non-blocking (which is a useful 2325Apart from keeping your process non-blocking (which is a useful
1207effect on its own sometimes), idle watchers are a good place to do 2326effect on its own sometimes), idle watchers are a good place to do
1208\&\*(L"pseudo\-background processing\*(R", or delay processing stuff to after the 2327\&\*(L"pseudo-background processing\*(R", or delay processing stuff to after the
1209event loop has handled all outstanding events. 2328event loop has handled all outstanding events.
2329.PP
2330\fIWatcher-Specific Functions and Data Members\fR
2331.IX Subsection "Watcher-Specific Functions and Data Members"
1210.IP "ev_idle_init (ev_signal *, callback)" 4 2332.IP "ev_idle_init (ev_signal *, callback)" 4
1211.IX Item "ev_idle_init (ev_signal *, callback)" 2333.IX Item "ev_idle_init (ev_signal *, callback)"
1212Initialises and configures the idle watcher \- it has no parameters of any 2334Initialises and configures the idle watcher \- it has no parameters of any
1213kind. There is a \f(CW\*(C`ev_idle_set\*(C'\fR macro, but using it is utterly pointless, 2335kind. There is a \f(CW\*(C`ev_idle_set\*(C'\fR macro, but using it is utterly pointless,
1214believe me. 2336believe me.
1215.PP 2337.PP
2338\fIExamples\fR
2339.IX Subsection "Examples"
2340.PP
1216Example: dynamically allocate an \f(CW\*(C`ev_idle\*(C'\fR, start it, and in the 2341Example: Dynamically allocate an \f(CW\*(C`ev_idle\*(C'\fR watcher, start it, and in the
1217callback, free it. Alos, use no error checking, as usual. 2342callback, free it. Also, use no error checking, as usual.
1218.PP 2343.PP
1219.Vb 7 2344.Vb 7
1220\& static void 2345\& static void
1221\& idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents) 2346\& idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
1222\& { 2347\& {
1223\& free (w); 2348\& free (w);
1224\& // now do something you wanted to do when the program has 2349\& // now do something you wanted to do when the program has
1225\& // no longer asnything immediate to do. 2350\& // no longer anything immediate to do.
1226\& } 2351\& }
1227.Ve 2352\&
1228.PP
1229.Vb 3
1230\& struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle)); 2353\& ev_idle *idle_watcher = malloc (sizeof (ev_idle));
1231\& ev_idle_init (idle_watcher, idle_cb); 2354\& ev_idle_init (idle_watcher, idle_cb);
1232\& ev_idle_start (loop, idle_cb); 2355\& ev_idle_start (loop, idle_cb);
1233.Ve 2356.Ve
1234.ie n .Sh """ev_prepare""\fP and \f(CW""ev_check"" \- customise your event loop" 2357.ie n .Sh """ev_prepare""\fP and \f(CW""ev_check"" \- customise your event loop!"
1235.el .Sh "\f(CWev_prepare\fP and \f(CWev_check\fP \- customise your event loop" 2358.el .Sh "\f(CWev_prepare\fP and \f(CWev_check\fP \- customise your event loop!"
1236.IX Subsection "ev_prepare and ev_check - customise your event loop" 2359.IX Subsection "ev_prepare and ev_check - customise your event loop!"
1237Prepare and check watchers are usually (but not always) used in tandem: 2360Prepare and check watchers are usually (but not always) used in pairs:
1238prepare watchers get invoked before the process blocks and check watchers 2361prepare watchers get invoked before the process blocks and check watchers
1239afterwards. 2362afterwards.
1240.PP 2363.PP
2364You \fImust not\fR call \f(CW\*(C`ev_loop\*(C'\fR or similar functions that enter
2365the current event loop from either \f(CW\*(C`ev_prepare\*(C'\fR or \f(CW\*(C`ev_check\*(C'\fR
2366watchers. Other loops than the current one are fine, however. The
2367rationale behind this is that you do not need to check for recursion in
2368those watchers, i.e. the sequence will always be \f(CW\*(C`ev_prepare\*(C'\fR, blocking,
2369\&\f(CW\*(C`ev_check\*(C'\fR so if you have one watcher of each kind they will always be
2370called in pairs bracketing the blocking call.
2371.PP
1241Their main purpose is to integrate other event mechanisms into libev and 2372Their main purpose is to integrate other event mechanisms into libev and
1242their use is somewhat advanced. This could be used, for example, to track 2373their use is somewhat advanced. They could be used, for example, to track
1243variable changes, implement your own watchers, integrate net-snmp or a 2374variable changes, implement your own watchers, integrate net-snmp or a
1244coroutine library and lots more. 2375coroutine library and lots more. They are also occasionally useful if
2376you cache some data and want to flush it before blocking (for example,
2377in X programs you might want to do an \f(CW\*(C`XFlush ()\*(C'\fR in an \f(CW\*(C`ev_prepare\*(C'\fR
2378watcher).
1245.PP 2379.PP
1246This is done by examining in each prepare call which file descriptors need 2380This is done by examining in each prepare call which file descriptors
1247to be watched by the other library, registering \f(CW\*(C`ev_io\*(C'\fR watchers for 2381need to be watched by the other library, registering \f(CW\*(C`ev_io\*(C'\fR watchers
1248them and starting an \f(CW\*(C`ev_timer\*(C'\fR watcher for any timeouts (many libraries 2382for them and starting an \f(CW\*(C`ev_timer\*(C'\fR watcher for any timeouts (many
1249provide just this functionality). Then, in the check watcher you check for 2383libraries provide exactly this functionality). Then, in the check watcher,
1250any events that occured (by checking the pending status of all watchers 2384you check for any events that occurred (by checking the pending status
1251and stopping them) and call back into the library. The I/O and timer 2385of all watchers and stopping them) and call back into the library. The
1252callbacks will never actually be called (but must be valid nevertheless, 2386I/O and timer callbacks will never actually be called (but must be valid
1253because you never know, you know?). 2387nevertheless, because you never know, you know?).
1254.PP 2388.PP
1255As another example, the Perl Coro module uses these hooks to integrate 2389As another example, the Perl Coro module uses these hooks to integrate
1256coroutines into libev programs, by yielding to other active coroutines 2390coroutines into libev programs, by yielding to other active coroutines
1257during each prepare and only letting the process block if no coroutines 2391during each prepare and only letting the process block if no coroutines
1258are ready to run (it's actually more complicated: it only runs coroutines 2392are ready to run (it's actually more complicated: it only runs coroutines
1259with priority higher than or equal to the event loop and one coroutine 2393with priority higher than or equal to the event loop and one coroutine
1260of lower priority, but only once, using idle watchers to keep the event 2394of lower priority, but only once, using idle watchers to keep the event
1261loop from blocking if lower-priority coroutines are active, thus mapping 2395loop from blocking if lower-priority coroutines are active, thus mapping
1262low-priority coroutines to idle/background tasks). 2396low-priority coroutines to idle/background tasks).
2397.PP
2398It is recommended to give \f(CW\*(C`ev_check\*(C'\fR watchers highest (\f(CW\*(C`EV_MAXPRI\*(C'\fR)
2399priority, to ensure that they are being run before any other watchers
2400after the poll (this doesn't matter for \f(CW\*(C`ev_prepare\*(C'\fR watchers).
2401.PP
2402Also, \f(CW\*(C`ev_check\*(C'\fR watchers (and \f(CW\*(C`ev_prepare\*(C'\fR watchers, too) should not
2403activate (\*(L"feed\*(R") events into libev. While libev fully supports this, they
2404might get executed before other \f(CW\*(C`ev_check\*(C'\fR watchers did their job. As
2405\&\f(CW\*(C`ev_check\*(C'\fR watchers are often used to embed other (non-libev) event
2406loops those other event loops might be in an unusable state until their
2407\&\f(CW\*(C`ev_check\*(C'\fR watcher ran (always remind yourself to coexist peacefully with
2408others).
2409.PP
2410\fIWatcher-Specific Functions and Data Members\fR
2411.IX Subsection "Watcher-Specific Functions and Data Members"
1263.IP "ev_prepare_init (ev_prepare *, callback)" 4 2412.IP "ev_prepare_init (ev_prepare *, callback)" 4
1264.IX Item "ev_prepare_init (ev_prepare *, callback)" 2413.IX Item "ev_prepare_init (ev_prepare *, callback)"
1265.PD 0 2414.PD 0
1266.IP "ev_check_init (ev_check *, callback)" 4 2415.IP "ev_check_init (ev_check *, callback)" 4
1267.IX Item "ev_check_init (ev_check *, callback)" 2416.IX Item "ev_check_init (ev_check *, callback)"
1268.PD 2417.PD
1269Initialises and configures the prepare or check watcher \- they have no 2418Initialises and configures the prepare or check watcher \- they have no
1270parameters of any kind. There are \f(CW\*(C`ev_prepare_set\*(C'\fR and \f(CW\*(C`ev_check_set\*(C'\fR 2419parameters of any kind. There are \f(CW\*(C`ev_prepare_set\*(C'\fR and \f(CW\*(C`ev_check_set\*(C'\fR
1271macros, but using them is utterly, utterly and completely pointless. 2420macros, but using them is utterly, utterly, utterly and completely
2421pointless.
1272.PP 2422.PP
1273Example: *TODO*. 2423\fIExamples\fR
2424.IX Subsection "Examples"
2425.PP
2426There are a number of principal ways to embed other event loops or modules
2427into libev. Here are some ideas on how to include libadns into libev
2428(there is a Perl module named \f(CW\*(C`EV::ADNS\*(C'\fR that does this, which you could
2429use as a working example. Another Perl module named \f(CW\*(C`EV::Glib\*(C'\fR embeds a
2430Glib main context into libev, and finally, \f(CW\*(C`Glib::EV\*(C'\fR embeds \s-1EV\s0 into the
2431Glib event loop).
2432.PP
2433Method 1: Add \s-1IO\s0 watchers and a timeout watcher in a prepare handler,
2434and in a check watcher, destroy them and call into libadns. What follows
2435is pseudo-code only of course. This requires you to either use a low
2436priority for the check watcher or use \f(CW\*(C`ev_clear_pending\*(C'\fR explicitly, as
2437the callbacks for the IO/timeout watchers might not have been called yet.
2438.PP
2439.Vb 2
2440\& static ev_io iow [nfd];
2441\& static ev_timer tw;
2442\&
2443\& static void
2444\& io_cb (struct ev_loop *loop, ev_io *w, int revents)
2445\& {
2446\& }
2447\&
2448\& // create io watchers for each fd and a timer before blocking
2449\& static void
2450\& adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
2451\& {
2452\& int timeout = 3600000;
2453\& struct pollfd fds [nfd];
2454\& // actual code will need to loop here and realloc etc.
2455\& adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
2456\&
2457\& /* the callback is illegal, but won\*(Aqt be called as we stop during check */
2458\& ev_timer_init (&tw, 0, timeout * 1e\-3);
2459\& ev_timer_start (loop, &tw);
2460\&
2461\& // create one ev_io per pollfd
2462\& for (int i = 0; i < nfd; ++i)
2463\& {
2464\& ev_io_init (iow + i, io_cb, fds [i].fd,
2465\& ((fds [i].events & POLLIN ? EV_READ : 0)
2466\& | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
2467\&
2468\& fds [i].revents = 0;
2469\& ev_io_start (loop, iow + i);
2470\& }
2471\& }
2472\&
2473\& // stop all watchers after blocking
2474\& static void
2475\& adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
2476\& {
2477\& ev_timer_stop (loop, &tw);
2478\&
2479\& for (int i = 0; i < nfd; ++i)
2480\& {
2481\& // set the relevant poll flags
2482\& // could also call adns_processreadable etc. here
2483\& struct pollfd *fd = fds + i;
2484\& int revents = ev_clear_pending (iow + i);
2485\& if (revents & EV_READ ) fd\->revents |= fd\->events & POLLIN;
2486\& if (revents & EV_WRITE) fd\->revents |= fd\->events & POLLOUT;
2487\&
2488\& // now stop the watcher
2489\& ev_io_stop (loop, iow + i);
2490\& }
2491\&
2492\& adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
2493\& }
2494.Ve
2495.PP
2496Method 2: This would be just like method 1, but you run \f(CW\*(C`adns_afterpoll\*(C'\fR
2497in the prepare watcher and would dispose of the check watcher.
2498.PP
2499Method 3: If the module to be embedded supports explicit event
2500notification (libadns does), you can also make use of the actual watcher
2501callbacks, and only destroy/create the watchers in the prepare watcher.
2502.PP
2503.Vb 5
2504\& static void
2505\& timer_cb (EV_P_ ev_timer *w, int revents)
2506\& {
2507\& adns_state ads = (adns_state)w\->data;
2508\& update_now (EV_A);
2509\&
2510\& adns_processtimeouts (ads, &tv_now);
2511\& }
2512\&
2513\& static void
2514\& io_cb (EV_P_ ev_io *w, int revents)
2515\& {
2516\& adns_state ads = (adns_state)w\->data;
2517\& update_now (EV_A);
2518\&
2519\& if (revents & EV_READ ) adns_processreadable (ads, w\->fd, &tv_now);
2520\& if (revents & EV_WRITE) adns_processwriteable (ads, w\->fd, &tv_now);
2521\& }
2522\&
2523\& // do not ever call adns_afterpoll
2524.Ve
2525.PP
2526Method 4: Do not use a prepare or check watcher because the module you
2527want to embed is not flexible enough to support it. Instead, you can
2528override their poll function. The drawback with this solution is that the
2529main loop is now no longer controllable by \s-1EV\s0. The \f(CW\*(C`Glib::EV\*(C'\fR module uses
2530this approach, effectively embedding \s-1EV\s0 as a client into the horrible
2531libglib event loop.
2532.PP
2533.Vb 4
2534\& static gint
2535\& event_poll_func (GPollFD *fds, guint nfds, gint timeout)
2536\& {
2537\& int got_events = 0;
2538\&
2539\& for (n = 0; n < nfds; ++n)
2540\& // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
2541\&
2542\& if (timeout >= 0)
2543\& // create/start timer
2544\&
2545\& // poll
2546\& ev_loop (EV_A_ 0);
2547\&
2548\& // stop timer again
2549\& if (timeout >= 0)
2550\& ev_timer_stop (EV_A_ &to);
2551\&
2552\& // stop io watchers again \- their callbacks should have set
2553\& for (n = 0; n < nfds; ++n)
2554\& ev_io_stop (EV_A_ iow [n]);
2555\&
2556\& return got_events;
2557\& }
2558.Ve
1274.ie n .Sh """ev_embed"" \- when one backend isn't enough" 2559.ie n .Sh """ev_embed"" \- when one backend isn't enough..."
1275.el .Sh "\f(CWev_embed\fP \- when one backend isn't enough" 2560.el .Sh "\f(CWev_embed\fP \- when one backend isn't enough..."
1276.IX Subsection "ev_embed - when one backend isn't enough" 2561.IX Subsection "ev_embed - when one backend isn't enough..."
1277This is a rather advanced watcher type that lets you embed one event loop 2562This is a rather advanced watcher type that lets you embed one event loop
1278into another (currently only \f(CW\*(C`ev_io\*(C'\fR events are supported in the embedded 2563into another (currently only \f(CW\*(C`ev_io\*(C'\fR events are supported in the embedded
1279loop, other types of watchers might be handled in a delayed or incorrect 2564loop, other types of watchers might be handled in a delayed or incorrect
1280fashion and must not be used). 2565fashion and must not be used).
1281.PP 2566.PP
… …
1283prioritise I/O. 2568prioritise I/O.
1284.PP 2569.PP
1285As an example for a bug workaround, the kqueue backend might only support 2570As an example for a bug workaround, the kqueue backend might only support
1286sockets on some platform, so it is unusable as generic backend, but you 2571sockets on some platform, so it is unusable as generic backend, but you
1287still want to make use of it because you have many sockets and it scales 2572still want to make use of it because you have many sockets and it scales
1288so nicely. In this case, you would create a kqueue-based loop and embed it 2573so nicely. In this case, you would create a kqueue-based loop and embed
1289into your default loop (which might use e.g. poll). Overall operation will 2574it into your default loop (which might use e.g. poll). Overall operation
1290be a bit slower because first libev has to poll and then call kevent, but 2575will be a bit slower because first libev has to call \f(CW\*(C`poll\*(C'\fR and then
1291at least you can use both at what they are best. 2576\&\f(CW\*(C`kevent\*(C'\fR, but at least you can use both mechanisms for what they are
2577best: \f(CW\*(C`kqueue\*(C'\fR for scalable sockets and \f(CW\*(C`poll\*(C'\fR if you want it to work :)
1292.PP 2578.PP
1293As for prioritising I/O: rarely you have the case where some fds have 2579As for prioritising I/O: under rare circumstances you have the case where
1294to be watched and handled very quickly (with low latency), and even 2580some fds have to be watched and handled very quickly (with low latency),
1295priorities and idle watchers might have too much overhead. In this case 2581and even priorities and idle watchers might have too much overhead. In
1296you would put all the high priority stuff in one loop and all the rest in 2582this case you would put all the high priority stuff in one loop and all
1297a second one, and embed the second one in the first. 2583the rest in a second one, and embed the second one in the first.
1298.PP 2584.PP
1299As long as the watcher is active, the callback will be invoked every time 2585As long as the watcher is active, the callback will be invoked every
1300there might be events pending in the embedded loop. The callback must then 2586time there might be events pending in the embedded loop. The callback
1301call \f(CW\*(C`ev_embed_sweep (mainloop, watcher)\*(C'\fR to make a single sweep and invoke 2587must then call \f(CW\*(C`ev_embed_sweep (mainloop, watcher)\*(C'\fR to make a single
1302their callbacks (you could also start an idle watcher to give the embedded 2588sweep and invoke their callbacks (the callback doesn't need to invoke the
1303loop strictly lower priority for example). You can also set the callback 2589\&\f(CW\*(C`ev_embed_sweep\*(C'\fR function directly, it could also start an idle watcher
1304to \f(CW0\fR, in which case the embed watcher will automatically execute the 2590to give the embedded loop strictly lower priority for example).
1305embedded loop sweep.
1306.PP 2591.PP
1307As long as the watcher is started it will automatically handle events. The 2592You can also set the callback to \f(CW0\fR, in which case the embed watcher
1308callback will be invoked whenever some events have been handled. You can 2593will automatically execute the embedded loop sweep whenever necessary.
1309set the callback to \f(CW0\fR to avoid having to specify one if you are not
1310interested in that.
1311.PP 2594.PP
1312Also, there have not currently been made special provisions for forking: 2595Fork detection will be handled transparently while the \f(CW\*(C`ev_embed\*(C'\fR watcher
1313when you fork, you not only have to call \f(CW\*(C`ev_loop_fork\*(C'\fR on both loops, 2596is active, i.e., the embedded loop will automatically be forked when the
1314but you will also have to stop and restart any \f(CW\*(C`ev_embed\*(C'\fR watchers 2597embedding loop forks. In other cases, the user is responsible for calling
1315yourself. 2598\&\f(CW\*(C`ev_loop_fork\*(C'\fR on the embedded loop.
1316.PP 2599.PP
1317Unfortunately, not all backends are embeddable, only the ones returned by 2600Unfortunately, not all backends are embeddable: only the ones returned by
1318\&\f(CW\*(C`ev_embeddable_backends\*(C'\fR are, which, unfortunately, does not include any 2601\&\f(CW\*(C`ev_embeddable_backends\*(C'\fR are, which, unfortunately, does not include any
1319portable one. 2602portable one.
1320.PP 2603.PP
1321So when you want to use this feature you will always have to be prepared 2604So when you want to use this feature you will always have to be prepared
1322that you cannot get an embeddable loop. The recommended way to get around 2605that you cannot get an embeddable loop. The recommended way to get around
1323this is to have a separate variables for your embeddable loop, try to 2606this is to have a separate variables for your embeddable loop, try to
1324create it, and if that fails, use the normal loop for everything: 2607create it, and if that fails, use the normal loop for everything.
1325.PP 2608.PP
1326.Vb 3 2609\fI\f(CI\*(C`ev_embed\*(C'\fI and fork\fR
1327\& struct ev_loop *loop_hi = ev_default_init (0); 2610.IX Subsection "ev_embed and fork"
1328\& struct ev_loop *loop_lo = 0;
1329\& struct ev_embed embed;
1330.Ve
1331.PP 2611.PP
1332.Vb 5 2612While the \f(CW\*(C`ev_embed\*(C'\fR watcher is running, forks in the embedding loop will
1333\& // see if there is a chance of getting one that works 2613automatically be applied to the embedded loop as well, so no special
1334\& // (remember that a flags value of 0 means autodetection) 2614fork handling is required in that case. When the watcher is not running,
1335\& loop_lo = ev_embeddable_backends () & ev_recommended_backends () 2615however, it is still the task of the libev user to call \f(CW\*(C`ev_loop_fork ()\*(C'\fR
1336\& ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ()) 2616as applicable.
1337\& : 0;
1338.Ve
1339.PP 2617.PP
1340.Vb 8 2618\fIWatcher-Specific Functions and Data Members\fR
1341\& // if we got one, then embed it, otherwise default to loop_hi 2619.IX Subsection "Watcher-Specific Functions and Data Members"
1342\& if (loop_lo)
1343\& {
1344\& ev_embed_init (&embed, 0, loop_lo);
1345\& ev_embed_start (loop_hi, &embed);
1346\& }
1347\& else
1348\& loop_lo = loop_hi;
1349.Ve
1350.IP "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)" 4 2620.IP "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)" 4
1351.IX Item "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)" 2621.IX Item "ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)"
1352.PD 0 2622.PD 0
1353.IP "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)" 4 2623.IP "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)" 4
1354.IX Item "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)" 2624.IX Item "ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)"
1355.PD 2625.PD
1356Configures the watcher to embed the given loop, which must be 2626Configures the watcher to embed the given loop, which must be
1357embeddable. If the callback is \f(CW0\fR, then \f(CW\*(C`ev_embed_sweep\*(C'\fR will be 2627embeddable. If the callback is \f(CW0\fR, then \f(CW\*(C`ev_embed_sweep\*(C'\fR will be
1358invoked automatically, otherwise it is the responsibility of the callback 2628invoked automatically, otherwise it is the responsibility of the callback
1359to invoke it (it will continue to be called until the sweep has been done, 2629to invoke it (it will continue to be called until the sweep has been done,
1360if you do not want thta, you need to temporarily stop the embed watcher). 2630if you do not want that, you need to temporarily stop the embed watcher).
1361.IP "ev_embed_sweep (loop, ev_embed *)" 4 2631.IP "ev_embed_sweep (loop, ev_embed *)" 4
1362.IX Item "ev_embed_sweep (loop, ev_embed *)" 2632.IX Item "ev_embed_sweep (loop, ev_embed *)"
1363Make a single, non-blocking sweep over the embedded loop. This works 2633Make a single, non-blocking sweep over the embedded loop. This works
1364similarly to \f(CW\*(C`ev_loop (embedded_loop, EVLOOP_NONBLOCK)\*(C'\fR, but in the most 2634similarly to \f(CW\*(C`ev_loop (embedded_loop, EVLOOP_NONBLOCK)\*(C'\fR, but in the most
1365apropriate way for embedded loops. 2635appropriate way for embedded loops.
2636.IP "struct ev_loop *other [read\-only]" 4
2637.IX Item "struct ev_loop *other [read-only]"
2638The embedded event loop.
2639.PP
2640\fIExamples\fR
2641.IX Subsection "Examples"
2642.PP
2643Example: Try to get an embeddable event loop and embed it into the default
2644event loop. If that is not possible, use the default loop. The default
2645loop is stored in \f(CW\*(C`loop_hi\*(C'\fR, while the embeddable loop is stored in
2646\&\f(CW\*(C`loop_lo\*(C'\fR (which is \f(CW\*(C`loop_hi\*(C'\fR in the case no embeddable loop can be
2647used).
2648.PP
2649.Vb 3
2650\& struct ev_loop *loop_hi = ev_default_init (0);
2651\& struct ev_loop *loop_lo = 0;
2652\& ev_embed embed;
2653\&
2654\& // see if there is a chance of getting one that works
2655\& // (remember that a flags value of 0 means autodetection)
2656\& loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
2657\& ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
2658\& : 0;
2659\&
2660\& // if we got one, then embed it, otherwise default to loop_hi
2661\& if (loop_lo)
2662\& {
2663\& ev_embed_init (&embed, 0, loop_lo);
2664\& ev_embed_start (loop_hi, &embed);
2665\& }
2666\& else
2667\& loop_lo = loop_hi;
2668.Ve
2669.PP
2670Example: Check if kqueue is available but not recommended and create
2671a kqueue backend for use with sockets (which usually work with any
2672kqueue implementation). Store the kqueue/socket\-only event loop in
2673\&\f(CW\*(C`loop_socket\*(C'\fR. (One might optionally use \f(CW\*(C`EVFLAG_NOENV\*(C'\fR, too).
2674.PP
2675.Vb 3
2676\& struct ev_loop *loop = ev_default_init (0);
2677\& struct ev_loop *loop_socket = 0;
2678\& ev_embed embed;
2679\&
2680\& if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2681\& if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2682\& {
2683\& ev_embed_init (&embed, 0, loop_socket);
2684\& ev_embed_start (loop, &embed);
2685\& }
2686\&
2687\& if (!loop_socket)
2688\& loop_socket = loop;
2689\&
2690\& // now use loop_socket for all sockets, and loop for everything else
2691.Ve
2692.ie n .Sh """ev_fork"" \- the audacity to resume the event loop after a fork"
2693.el .Sh "\f(CWev_fork\fP \- the audacity to resume the event loop after a fork"
2694.IX Subsection "ev_fork - the audacity to resume the event loop after a fork"
2695Fork watchers are called when a \f(CW\*(C`fork ()\*(C'\fR was detected (usually because
2696whoever is a good citizen cared to tell libev about it by calling
2697\&\f(CW\*(C`ev_default_fork\*(C'\fR or \f(CW\*(C`ev_loop_fork\*(C'\fR). The invocation is done before the
2698event loop blocks next and before \f(CW\*(C`ev_check\*(C'\fR watchers are being called,
2699and only in the child after the fork. If whoever good citizen calling
2700\&\f(CW\*(C`ev_default_fork\*(C'\fR cheats and calls it in the wrong process, the fork
2701handlers will be invoked, too, of course.
2702.PP
2703\fIWatcher-Specific Functions and Data Members\fR
2704.IX Subsection "Watcher-Specific Functions and Data Members"
2705.IP "ev_fork_init (ev_signal *, callback)" 4
2706.IX Item "ev_fork_init (ev_signal *, callback)"
2707Initialises and configures the fork watcher \- it has no parameters of any
2708kind. There is a \f(CW\*(C`ev_fork_set\*(C'\fR macro, but using it is utterly pointless,
2709believe me.
2710.ie n .Sh """ev_async"" \- how to wake up another event loop"
2711.el .Sh "\f(CWev_async\fP \- how to wake up another event loop"
2712.IX Subsection "ev_async - how to wake up another event loop"
2713In general, you cannot use an \f(CW\*(C`ev_loop\*(C'\fR from multiple threads or other
2714asynchronous sources such as signal handlers (as opposed to multiple event
2715loops \- those are of course safe to use in different threads).
2716.PP
2717Sometimes, however, you need to wake up another event loop you do not
2718control, for example because it belongs to another thread. This is what
2719\&\f(CW\*(C`ev_async\*(C'\fR watchers do: as long as the \f(CW\*(C`ev_async\*(C'\fR watcher is active, you
2720can signal it by calling \f(CW\*(C`ev_async_send\*(C'\fR, which is thread\- and signal
2721safe.
2722.PP
2723This functionality is very similar to \f(CW\*(C`ev_signal\*(C'\fR watchers, as signals,
2724too, are asynchronous in nature, and signals, too, will be compressed
2725(i.e. the number of callback invocations may be less than the number of
2726\&\f(CW\*(C`ev_async_sent\*(C'\fR calls).
2727.PP
2728Unlike \f(CW\*(C`ev_signal\*(C'\fR watchers, \f(CW\*(C`ev_async\*(C'\fR works with any event loop, not
2729just the default loop.
2730.PP
2731\fIQueueing\fR
2732.IX Subsection "Queueing"
2733.PP
2734\&\f(CW\*(C`ev_async\*(C'\fR does not support queueing of data in any way. The reason
2735is that the author does not know of a simple (or any) algorithm for a
2736multiple-writer-single-reader queue that works in all cases and doesn't
2737need elaborate support such as pthreads.
2738.PP
2739That means that if you want to queue data, you have to provide your own
2740queue. But at least I can tell you how to implement locking around your
2741queue:
2742.IP "queueing from a signal handler context" 4
2743.IX Item "queueing from a signal handler context"
2744To implement race-free queueing, you simply add to the queue in the signal
2745handler but you block the signal handler in the watcher callback. Here is
2746an example that does that for some fictitious \s-1SIGUSR1\s0 handler:
2747.Sp
2748.Vb 1
2749\& static ev_async mysig;
2750\&
2751\& static void
2752\& sigusr1_handler (void)
2753\& {
2754\& sometype data;
2755\&
2756\& // no locking etc.
2757\& queue_put (data);
2758\& ev_async_send (EV_DEFAULT_ &mysig);
2759\& }
2760\&
2761\& static void
2762\& mysig_cb (EV_P_ ev_async *w, int revents)
2763\& {
2764\& sometype data;
2765\& sigset_t block, prev;
2766\&
2767\& sigemptyset (&block);
2768\& sigaddset (&block, SIGUSR1);
2769\& sigprocmask (SIG_BLOCK, &block, &prev);
2770\&
2771\& while (queue_get (&data))
2772\& process (data);
2773\&
2774\& if (sigismember (&prev, SIGUSR1)
2775\& sigprocmask (SIG_UNBLOCK, &block, 0);
2776\& }
2777.Ve
2778.Sp
2779(Note: pthreads in theory requires you to use \f(CW\*(C`pthread_setmask\*(C'\fR
2780instead of \f(CW\*(C`sigprocmask\*(C'\fR when you use threads, but libev doesn't do it
2781either...).
2782.IP "queueing from a thread context" 4
2783.IX Item "queueing from a thread context"
2784The strategy for threads is different, as you cannot (easily) block
2785threads but you can easily preempt them, so to queue safely you need to
2786employ a traditional mutex lock, such as in this pthread example:
2787.Sp
2788.Vb 2
2789\& static ev_async mysig;
2790\& static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
2791\&
2792\& static void
2793\& otherthread (void)
2794\& {
2795\& // only need to lock the actual queueing operation
2796\& pthread_mutex_lock (&mymutex);
2797\& queue_put (data);
2798\& pthread_mutex_unlock (&mymutex);
2799\&
2800\& ev_async_send (EV_DEFAULT_ &mysig);
2801\& }
2802\&
2803\& static void
2804\& mysig_cb (EV_P_ ev_async *w, int revents)
2805\& {
2806\& pthread_mutex_lock (&mymutex);
2807\&
2808\& while (queue_get (&data))
2809\& process (data);
2810\&
2811\& pthread_mutex_unlock (&mymutex);
2812\& }
2813.Ve
2814.PP
2815\fIWatcher-Specific Functions and Data Members\fR
2816.IX Subsection "Watcher-Specific Functions and Data Members"
2817.IP "ev_async_init (ev_async *, callback)" 4
2818.IX Item "ev_async_init (ev_async *, callback)"
2819Initialises and configures the async watcher \- it has no parameters of any
2820kind. There is a \f(CW\*(C`ev_async_set\*(C'\fR macro, but using it is utterly pointless,
2821trust me.
2822.IP "ev_async_send (loop, ev_async *)" 4
2823.IX Item "ev_async_send (loop, ev_async *)"
2824Sends/signals/activates the given \f(CW\*(C`ev_async\*(C'\fR watcher, that is, feeds
2825an \f(CW\*(C`EV_ASYNC\*(C'\fR event on the watcher into the event loop. Unlike
2826\&\f(CW\*(C`ev_feed_event\*(C'\fR, this call is safe to do from other threads, signal or
2827similar contexts (see the discussion of \f(CW\*(C`EV_ATOMIC_T\*(C'\fR in the embedding
2828section below on what exactly this means).
2829.Sp
2830This call incurs the overhead of a system call only once per loop iteration,
2831so while the overhead might be noticeable, it doesn't apply to repeated
2832calls to \f(CW\*(C`ev_async_send\*(C'\fR.
2833.IP "bool = ev_async_pending (ev_async *)" 4
2834.IX Item "bool = ev_async_pending (ev_async *)"
2835Returns a non-zero value when \f(CW\*(C`ev_async_send\*(C'\fR has been called on the
2836watcher but the event has not yet been processed (or even noted) by the
2837event loop.
2838.Sp
2839\&\f(CW\*(C`ev_async_send\*(C'\fR sets a flag in the watcher and wakes up the loop. When
2840the loop iterates next and checks for the watcher to have become active,
2841it will reset the flag again. \f(CW\*(C`ev_async_pending\*(C'\fR can be used to very
2842quickly check whether invoking the loop might be a good idea.
2843.Sp
2844Not that this does \fInot\fR check whether the watcher itself is pending, only
2845whether it has been requested to make this watcher pending.
1366.SH "OTHER FUNCTIONS" 2846.SH "OTHER FUNCTIONS"
1367.IX Header "OTHER FUNCTIONS" 2847.IX Header "OTHER FUNCTIONS"
1368There are some other functions of possible interest. Described. Here. Now. 2848There are some other functions of possible interest. Described. Here. Now.
1369.IP "ev_once (loop, int fd, int events, ev_tstamp timeout, callback)" 4 2849.IP "ev_once (loop, int fd, int events, ev_tstamp timeout, callback)" 4
1370.IX Item "ev_once (loop, int fd, int events, ev_tstamp timeout, callback)" 2850.IX Item "ev_once (loop, int fd, int events, ev_tstamp timeout, callback)"
1371This function combines a simple timer and an I/O watcher, calls your 2851This function combines a simple timer and an I/O watcher, calls your
1372callback on whichever event happens first and automatically stop both 2852callback on whichever event happens first and automatically stops both
1373watchers. This is useful if you want to wait for a single event on an fd 2853watchers. This is useful if you want to wait for a single event on an fd
1374or timeout without having to allocate/configure/start/stop/free one or 2854or timeout without having to allocate/configure/start/stop/free one or
1375more watchers yourself. 2855more watchers yourself.
1376.Sp 2856.Sp
1377If \f(CW\*(C`fd\*(C'\fR is less than 0, then no I/O watcher will be started and events 2857If \f(CW\*(C`fd\*(C'\fR is less than 0, then no I/O watcher will be started and the
1378is being ignored. Otherwise, an \f(CW\*(C`ev_io\*(C'\fR watcher for the given \f(CW\*(C`fd\*(C'\fR and 2858\&\f(CW\*(C`events\*(C'\fR argument is being ignored. Otherwise, an \f(CW\*(C`ev_io\*(C'\fR watcher for
1379\&\f(CW\*(C`events\*(C'\fR set will be craeted and started. 2859the given \f(CW\*(C`fd\*(C'\fR and \f(CW\*(C`events\*(C'\fR set will be created and started.
1380.Sp 2860.Sp
1381If \f(CW\*(C`timeout\*(C'\fR is less than 0, then no timeout watcher will be 2861If \f(CW\*(C`timeout\*(C'\fR is less than 0, then no timeout watcher will be
1382started. Otherwise an \f(CW\*(C`ev_timer\*(C'\fR watcher with after = \f(CW\*(C`timeout\*(C'\fR (and 2862started. Otherwise an \f(CW\*(C`ev_timer\*(C'\fR watcher with after = \f(CW\*(C`timeout\*(C'\fR (and
1383repeat = 0) will be started. While \f(CW0\fR is a valid timeout, it is of 2863repeat = 0) will be started. \f(CW0\fR is a valid timeout.
1384dubious value.
1385.Sp 2864.Sp
1386The callback has the type \f(CW\*(C`void (*cb)(int revents, void *arg)\*(C'\fR and gets 2865The callback has the type \f(CW\*(C`void (*cb)(int revents, void *arg)\*(C'\fR and gets
1387passed an \f(CW\*(C`revents\*(C'\fR set like normal event callbacks (a combination of 2866passed an \f(CW\*(C`revents\*(C'\fR set like normal event callbacks (a combination of
1388\&\f(CW\*(C`EV_ERROR\*(C'\fR, \f(CW\*(C`EV_READ\*(C'\fR, \f(CW\*(C`EV_WRITE\*(C'\fR or \f(CW\*(C`EV_TIMEOUT\*(C'\fR) and the \f(CW\*(C`arg\*(C'\fR 2867\&\f(CW\*(C`EV_ERROR\*(C'\fR, \f(CW\*(C`EV_READ\*(C'\fR, \f(CW\*(C`EV_WRITE\*(C'\fR or \f(CW\*(C`EV_TIMEOUT\*(C'\fR) and the \f(CW\*(C`arg\*(C'\fR
1389value passed to \f(CW\*(C`ev_once\*(C'\fR: 2868value passed to \f(CW\*(C`ev_once\*(C'\fR. Note that it is possible to receive \fIboth\fR
2869a timeout and an io event at the same time \- you probably should give io
2870events precedence.
2871.Sp
2872Example: wait up to ten seconds for data to appear on \s-1STDIN_FILENO\s0.
1390.Sp 2873.Sp
1391.Vb 7 2874.Vb 7
1392\& static void stdin_ready (int revents, void *arg) 2875\& static void stdin_ready (int revents, void *arg)
1393\& { 2876\& {
1394\& if (revents & EV_TIMEOUT)
1395\& /* doh, nothing entered */;
1396\& else if (revents & EV_READ) 2877\& if (revents & EV_READ)
1397\& /* stdin might have data for us, joy! */; 2878\& /* stdin might have data for us, joy! */;
2879\& else if (revents & EV_TIMEOUT)
2880\& /* doh, nothing entered */;
1398\& } 2881\& }
1399.Ve 2882\&
1400.Sp
1401.Vb 1
1402\& ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0); 2883\& ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
1403.Ve 2884.Ve
1404.IP "ev_feed_event (ev_loop *, watcher *, int revents)" 4 2885.IP "ev_feed_event (struct ev_loop *, watcher *, int revents)" 4
1405.IX Item "ev_feed_event (ev_loop *, watcher *, int revents)" 2886.IX Item "ev_feed_event (struct ev_loop *, watcher *, int revents)"
1406Feeds the given event set into the event loop, as if the specified event 2887Feeds the given event set into the event loop, as if the specified event
1407had happened for the specified watcher (which must be a pointer to an 2888had happened for the specified watcher (which must be a pointer to an
1408initialised but not necessarily started event watcher). 2889initialised but not necessarily started event watcher).
1409.IP "ev_feed_fd_event (ev_loop *, int fd, int revents)" 4 2890.IP "ev_feed_fd_event (struct ev_loop *, int fd, int revents)" 4
1410.IX Item "ev_feed_fd_event (ev_loop *, int fd, int revents)" 2891.IX Item "ev_feed_fd_event (struct ev_loop *, int fd, int revents)"
1411Feed an event on the given fd, as if a file descriptor backend detected 2892Feed an event on the given fd, as if a file descriptor backend detected
1412the given events it. 2893the given events it.
1413.IP "ev_feed_signal_event (ev_loop *loop, int signum)" 4 2894.IP "ev_feed_signal_event (struct ev_loop *loop, int signum)" 4
1414.IX Item "ev_feed_signal_event (ev_loop *loop, int signum)" 2895.IX Item "ev_feed_signal_event (struct ev_loop *loop, int signum)"
1415Feed an event as if the given signal occured (\f(CW\*(C`loop\*(C'\fR must be the default 2896Feed an event as if the given signal occurred (\f(CW\*(C`loop\*(C'\fR must be the default
1416loop!). 2897loop!).
1417.SH "LIBEVENT EMULATION" 2898.SH "LIBEVENT EMULATION"
1418.IX Header "LIBEVENT EMULATION" 2899.IX Header "LIBEVENT EMULATION"
1419Libev offers a compatibility emulation layer for libevent. It cannot 2900Libev offers a compatibility emulation layer for libevent. It cannot
1420emulate the internals of libevent, so here are some usage hints: 2901emulate the internals of libevent, so here are some usage hints:
2902.IP "\(bu" 4
1421.IP "* Use it by including <event.h>, as usual." 4 2903Use it by including <event.h>, as usual.
1422.IX Item "Use it by including <event.h>, as usual." 2904.IP "\(bu" 4
1423.PD 0 2905The following members are fully supported: ev_base, ev_callback,
1424.IP "* The following members are fully supported: ev_base, ev_callback, ev_arg, ev_fd, ev_res, ev_events." 4 2906ev_arg, ev_fd, ev_res, ev_events.
1425.IX Item "The following members are fully supported: ev_base, ev_callback, ev_arg, ev_fd, ev_res, ev_events." 2907.IP "\(bu" 4
1426.IP "* Avoid using ev_flags and the EVLIST_*\-macros, while it is maintained by libev, it does not work exactly the same way as in libevent (consider it a private \s-1API\s0)." 4 2908Avoid using ev_flags and the EVLIST_*\-macros, while it is
1427.IX Item "Avoid using ev_flags and the EVLIST_*-macros, while it is maintained by libev, it does not work exactly the same way as in libevent (consider it a private API)." 2909maintained by libev, it does not work exactly the same way as in libevent (consider
1428.IP "* Priorities are not currently supported. Initialising priorities will fail and all watchers will have the same priority, even though there is an ev_pri field." 4 2910it a private \s-1API\s0).
1429.IX Item "Priorities are not currently supported. Initialising priorities will fail and all watchers will have the same priority, even though there is an ev_pri field." 2911.IP "\(bu" 4
2912Priorities are not currently supported. Initialising priorities
2913will fail and all watchers will have the same priority, even though there
2914is an ev_pri field.
2915.IP "\(bu" 4
2916In libevent, the last base created gets the signals, in libev, the
2917first base created (== the default loop) gets the signals.
2918.IP "\(bu" 4
1430.IP "* Other members are not supported." 4 2919Other members are not supported.
1431.IX Item "Other members are not supported." 2920.IP "\(bu" 4
1432.IP "* The libev emulation is \fInot\fR \s-1ABI\s0 compatible to libevent, you need to use the libev header file and library." 4 2921The libev emulation is \fInot\fR \s-1ABI\s0 compatible to libevent, you need
1433.IX Item "The libev emulation is not ABI compatible to libevent, you need to use the libev header file and library." 2922to use the libev header file and library.
1434.PD
1435.SH "\*(C+ SUPPORT" 2923.SH "\*(C+ SUPPORT"
1436.IX Header " SUPPORT" 2924.IX Header " SUPPORT"
1437Libev comes with some simplistic wrapper classes for \*(C+ that mainly allow 2925Libev comes with some simplistic wrapper classes for \*(C+ that mainly allow
1438you to use some convinience methods to start/stop watchers and also change 2926you to use some convenience methods to start/stop watchers and also change
1439the callback model to a model using method callbacks on objects. 2927the callback model to a model using method callbacks on objects.
1440.PP 2928.PP
1441To use it, 2929To use it,
1442.PP 2930.PP
1443.Vb 1 2931.Vb 1
1444\& #include <ev++.h> 2932\& #include <ev++.h>
1445.Ve 2933.Ve
1446.PP 2934.PP
1447(it is not installed by default). This automatically includes \fIev.h\fR 2935This automatically includes \fIev.h\fR and puts all of its definitions (many
1448and puts all of its definitions (many of them macros) into the global 2936of them macros) into the global namespace. All \*(C+ specific things are
1449namespace. All \*(C+ specific things are put into the \f(CW\*(C`ev\*(C'\fR namespace. 2937put into the \f(CW\*(C`ev\*(C'\fR namespace. It should support all the same embedding
2938options as \fIev.h\fR, most notably \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR.
1450.PP 2939.PP
1451It should support all the same embedding options as \fIev.h\fR, most notably 2940Care has been taken to keep the overhead low. The only data member the \*(C+
1452\&\f(CW\*(C`EV_MULTIPLICITY\*(C'\fR. 2941classes add (compared to plain C\-style watchers) is the event loop pointer
2942that the watcher is associated with (or no additional members at all if
2943you disable \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR when embedding libev).
2944.PP
2945Currently, functions, and static and non-static member functions can be
2946used as callbacks. Other types should be easy to add as long as they only
2947need one additional pointer for context. If you need support for other
2948types of functors please contact the author (preferably after implementing
2949it).
1453.PP 2950.PP
1454Here is a list of things available in the \f(CW\*(C`ev\*(C'\fR namespace: 2951Here is a list of things available in the \f(CW\*(C`ev\*(C'\fR namespace:
1455.ie n .IP """ev::READ""\fR, \f(CW""ev::WRITE"" etc." 4 2952.ie n .IP """ev::READ""\fR, \f(CW""ev::WRITE"" etc." 4
1456.el .IP "\f(CWev::READ\fR, \f(CWev::WRITE\fR etc." 4 2953.el .IP "\f(CWev::READ\fR, \f(CWev::WRITE\fR etc." 4
1457.IX Item "ev::READ, ev::WRITE etc." 2954.IX Item "ev::READ, ev::WRITE etc."
… …
1469which is called \f(CW\*(C`ev::sig\*(C'\fR to avoid clashes with the \f(CW\*(C`signal\*(C'\fR macro 2966which is called \f(CW\*(C`ev::sig\*(C'\fR to avoid clashes with the \f(CW\*(C`signal\*(C'\fR macro
1470defines by many implementations. 2967defines by many implementations.
1471.Sp 2968.Sp
1472All of those classes have these methods: 2969All of those classes have these methods:
1473.RS 4 2970.RS 4
1474.IP "ev::TYPE::TYPE (object *, object::method *)" 4 2971.IP "ev::TYPE::TYPE ()" 4
1475.IX Item "ev::TYPE::TYPE (object *, object::method *)" 2972.IX Item "ev::TYPE::TYPE ()"
1476.PD 0 2973.PD 0
1477.IP "ev::TYPE::TYPE (object *, object::method *, struct ev_loop *)" 4 2974.IP "ev::TYPE::TYPE (struct ev_loop *)" 4
1478.IX Item "ev::TYPE::TYPE (object *, object::method *, struct ev_loop *)" 2975.IX Item "ev::TYPE::TYPE (struct ev_loop *)"
1479.IP "ev::TYPE::~TYPE" 4 2976.IP "ev::TYPE::~TYPE" 4
1480.IX Item "ev::TYPE::~TYPE" 2977.IX Item "ev::TYPE::~TYPE"
1481.PD 2978.PD
1482The constructor takes a pointer to an object and a method pointer to 2979The constructor (optionally) takes an event loop to associate the watcher
1483the event handler callback to call in this class. The constructor calls 2980with. If it is omitted, it will use \f(CW\*(C`EV_DEFAULT\*(C'\fR.
1484\&\f(CW\*(C`ev_init\*(C'\fR for you, which means you have to call the \f(CW\*(C`set\*(C'\fR method 2981.Sp
1485before starting it. If you do not specify a loop then the constructor 2982The constructor calls \f(CW\*(C`ev_init\*(C'\fR for you, which means you have to call the
1486automatically associates the default loop with this watcher. 2983\&\f(CW\*(C`set\*(C'\fR method before starting it.
2984.Sp
2985It will not set a callback, however: You have to call the templated \f(CW\*(C`set\*(C'\fR
2986method to set a callback before you can start the watcher.
2987.Sp
2988(The reason why you have to use a method is a limitation in \*(C+ which does
2989not allow explicit template arguments for constructors).
1487.Sp 2990.Sp
1488The destructor automatically stops the watcher if it is active. 2991The destructor automatically stops the watcher if it is active.
2992.IP "w\->set<class, &class::method> (object *)" 4
2993.IX Item "w->set<class, &class::method> (object *)"
2994This method sets the callback method to call. The method has to have a
2995signature of \f(CW\*(C`void (*)(ev_TYPE &, int)\*(C'\fR, it receives the watcher as
2996first argument and the \f(CW\*(C`revents\*(C'\fR as second. The object must be given as
2997parameter and is stored in the \f(CW\*(C`data\*(C'\fR member of the watcher.
2998.Sp
2999This method synthesizes efficient thunking code to call your method from
3000the C callback that libev requires. If your compiler can inline your
3001callback (i.e. it is visible to it at the place of the \f(CW\*(C`set\*(C'\fR call and
3002your compiler is good :), then the method will be fully inlined into the
3003thunking function, making it as fast as a direct C callback.
3004.Sp
3005Example: simple class declaration and watcher initialisation
3006.Sp
3007.Vb 4
3008\& struct myclass
3009\& {
3010\& void io_cb (ev::io &w, int revents) { }
3011\& }
3012\&
3013\& myclass obj;
3014\& ev::io iow;
3015\& iow.set <myclass, &myclass::io_cb> (&obj);
3016.Ve
3017.IP "w\->set (object *)" 4
3018.IX Item "w->set (object *)"
3019This is an \fBexperimental\fR feature that might go away in a future version.
3020.Sp
3021This is a variation of a method callback \- leaving out the method to call
3022will default the method to \f(CW\*(C`operator ()\*(C'\fR, which makes it possible to use
3023functor objects without having to manually specify the \f(CW\*(C`operator ()\*(C'\fR all
3024the time. Incidentally, you can then also leave out the template argument
3025list.
3026.Sp
3027The \f(CW\*(C`operator ()\*(C'\fR method prototype must be \f(CW\*(C`void operator ()(watcher &w,
3028int revents)\*(C'\fR.
3029.Sp
3030See the method\-\f(CW\*(C`set\*(C'\fR above for more details.
3031.Sp
3032Example: use a functor object as callback.
3033.Sp
3034.Vb 7
3035\& struct myfunctor
3036\& {
3037\& void operator() (ev::io &w, int revents)
3038\& {
3039\& ...
3040\& }
3041\& }
3042\&
3043\& myfunctor f;
3044\&
3045\& ev::io w;
3046\& w.set (&f);
3047.Ve
3048.IP "w\->set<function> (void *data = 0)" 4
3049.IX Item "w->set<function> (void *data = 0)"
3050Also sets a callback, but uses a static method or plain function as
3051callback. The optional \f(CW\*(C`data\*(C'\fR argument will be stored in the watcher's
3052\&\f(CW\*(C`data\*(C'\fR member and is free for you to use.
3053.Sp
3054The prototype of the \f(CW\*(C`function\*(C'\fR must be \f(CW\*(C`void (*)(ev::TYPE &w, int)\*(C'\fR.
3055.Sp
3056See the method\-\f(CW\*(C`set\*(C'\fR above for more details.
3057.Sp
3058Example: Use a plain function as callback.
3059.Sp
3060.Vb 2
3061\& static void io_cb (ev::io &w, int revents) { }
3062\& iow.set <io_cb> ();
3063.Ve
1489.IP "w\->set (struct ev_loop *)" 4 3064.IP "w\->set (struct ev_loop *)" 4
1490.IX Item "w->set (struct ev_loop *)" 3065.IX Item "w->set (struct ev_loop *)"
1491Associates a different \f(CW\*(C`struct ev_loop\*(C'\fR with this watcher. You can only 3066Associates a different \f(CW\*(C`struct ev_loop\*(C'\fR with this watcher. You can only
1492do this when the watcher is inactive (and not pending either). 3067do this when the watcher is inactive (and not pending either).
1493.IP "w\->set ([args])" 4 3068.IP "w\->set ([arguments])" 4
1494.IX Item "w->set ([args])" 3069.IX Item "w->set ([arguments])"
1495Basically the same as \f(CW\*(C`ev_TYPE_set\*(C'\fR, with the same args. Must be 3070Basically the same as \f(CW\*(C`ev_TYPE_set\*(C'\fR, with the same arguments. Must be
1496called at least once. Unlike the C counterpart, an active watcher gets 3071called at least once. Unlike the C counterpart, an active watcher gets
1497automatically stopped and restarted. 3072automatically stopped and restarted when reconfiguring it with this
3073method.
1498.IP "w\->start ()" 4 3074.IP "w\->start ()" 4
1499.IX Item "w->start ()" 3075.IX Item "w->start ()"
1500Starts the watcher. Note that there is no \f(CW\*(C`loop\*(C'\fR argument as the 3076Starts the watcher. Note that there is no \f(CW\*(C`loop\*(C'\fR argument, as the
1501constructor already takes the loop. 3077constructor already stores the event loop.
1502.IP "w\->stop ()" 4 3078.IP "w\->stop ()" 4
1503.IX Item "w->stop ()" 3079.IX Item "w->stop ()"
1504Stops the watcher if it is active. Again, no \f(CW\*(C`loop\*(C'\fR argument. 3080Stops the watcher if it is active. Again, no \f(CW\*(C`loop\*(C'\fR argument.
1505.ie n .IP "w\->again () ""ev::timer""\fR, \f(CW""ev::periodic"" only" 4 3081.ie n .IP "w\->again () (""ev::timer""\fR, \f(CW""ev::periodic"" only)" 4
1506.el .IP "w\->again () \f(CWev::timer\fR, \f(CWev::periodic\fR only" 4 3082.el .IP "w\->again () (\f(CWev::timer\fR, \f(CWev::periodic\fR only)" 4
1507.IX Item "w->again () ev::timer, ev::periodic only" 3083.IX Item "w->again () (ev::timer, ev::periodic only)"
1508For \f(CW\*(C`ev::timer\*(C'\fR and \f(CW\*(C`ev::periodic\*(C'\fR, this invokes the corresponding 3084For \f(CW\*(C`ev::timer\*(C'\fR and \f(CW\*(C`ev::periodic\*(C'\fR, this invokes the corresponding
1509\&\f(CW\*(C`ev_TYPE_again\*(C'\fR function. 3085\&\f(CW\*(C`ev_TYPE_again\*(C'\fR function.
1510.ie n .IP "w\->sweep () ""ev::embed"" only" 4 3086.ie n .IP "w\->sweep () (""ev::embed"" only)" 4
1511.el .IP "w\->sweep () \f(CWev::embed\fR only" 4 3087.el .IP "w\->sweep () (\f(CWev::embed\fR only)" 4
1512.IX Item "w->sweep () ev::embed only" 3088.IX Item "w->sweep () (ev::embed only)"
1513Invokes \f(CW\*(C`ev_embed_sweep\*(C'\fR. 3089Invokes \f(CW\*(C`ev_embed_sweep\*(C'\fR.
3090.ie n .IP "w\->update () (""ev::stat"" only)" 4
3091.el .IP "w\->update () (\f(CWev::stat\fR only)" 4
3092.IX Item "w->update () (ev::stat only)"
3093Invokes \f(CW\*(C`ev_stat_stat\*(C'\fR.
1514.RE 3094.RE
1515.RS 4 3095.RS 4
1516.RE 3096.RE
1517.PP 3097.PP
1518Example: Define a class with an \s-1IO\s0 and idle watcher, start one of them in 3098Example: Define a class with an \s-1IO\s0 and idle watcher, start one of them in
1519the constructor. 3099the constructor.
1520.PP 3100.PP
1521.Vb 4 3101.Vb 4
1522\& class myclass 3102\& class myclass
1523\& { 3103\& {
1524\& ev_io io; void io_cb (ev::io &w, int revents); 3104\& ev::io io ; void io_cb (ev::io &w, int revents);
1525\& ev_idle idle void idle_cb (ev::idle &w, int revents); 3105\& ev::idle idle; void idle_cb (ev::idle &w, int revents);
3106\&
3107\& myclass (int fd)
3108\& {
3109\& io .set <myclass, &myclass::io_cb > (this);
3110\& idle.set <myclass, &myclass::idle_cb> (this);
3111\&
3112\& io.start (fd, ev::READ);
3113\& }
3114\& };
1526.Ve 3115.Ve
3116.SH "OTHER LANGUAGE BINDINGS"
3117.IX Header "OTHER LANGUAGE BINDINGS"
3118Libev does not offer other language bindings itself, but bindings for a
3119number of languages exist in the form of third-party packages. If you know
3120any interesting language binding in addition to the ones listed here, drop
3121me a note.
3122.IP "Perl" 4
3123.IX Item "Perl"
3124The \s-1EV\s0 module implements the full libev \s-1API\s0 and is actually used to test
3125libev. \s-1EV\s0 is developed together with libev. Apart from the \s-1EV\s0 core module,
3126there are additional modules that implement libev-compatible interfaces
3127to \f(CW\*(C`libadns\*(C'\fR (\f(CW\*(C`EV::ADNS\*(C'\fR, but \f(CW\*(C`AnyEvent::DNS\*(C'\fR is preferred nowadays),
3128\&\f(CW\*(C`Net::SNMP\*(C'\fR (\f(CW\*(C`Net::SNMP::EV\*(C'\fR) and the \f(CW\*(C`libglib\*(C'\fR event core (\f(CW\*(C`Glib::EV\*(C'\fR
3129and \f(CW\*(C`EV::Glib\*(C'\fR).
3130.Sp
3131It can be found and installed via \s-1CPAN\s0, its homepage is at
3132<http://software.schmorp.de/pkg/EV>.
3133.IP "Python" 4
3134.IX Item "Python"
3135Python bindings can be found at <http://code.google.com/p/pyev/>. It
3136seems to be quite complete and well-documented. Note, however, that the
3137patch they require for libev is outright dangerous as it breaks the \s-1ABI\s0
3138for everybody else, and therefore, should never be applied in an installed
3139libev (if python requires an incompatible \s-1ABI\s0 then it needs to embed
3140libev).
3141.IP "Ruby" 4
3142.IX Item "Ruby"
3143Tony Arcieri has written a ruby extension that offers access to a subset
3144of the libev \s-1API\s0 and adds file handle abstractions, asynchronous \s-1DNS\s0 and
3145more on top of it. It can be found via gem servers. Its homepage is at
3146<http://rev.rubyforge.org/>.
3147.Sp
3148Roger Pack reports that using the link order \f(CW\*(C`\-lws2_32 \-lmsvcrt\-ruby\-190\*(C'\fR
3149makes rev work even on mingw.
3150.IP "D" 4
3151.IX Item "D"
3152Leandro Lucarella has written a D language binding (\fIev.d\fR) for libev, to
3153be found at <http://proj.llucax.com.ar/wiki/evd>.
3154.IP "Ocaml" 4
3155.IX Item "Ocaml"
3156Erkki Seppala has written Ocaml bindings for libev, to be found at
3157<http://modeemi.cs.tut.fi/~flux/software/ocaml\-ev/>.
3158.SH "MACRO MAGIC"
3159.IX Header "MACRO MAGIC"
3160Libev can be compiled with a variety of options, the most fundamental
3161of which is \f(CW\*(C`EV_MULTIPLICITY\*(C'\fR. This option determines whether (most)
3162functions and callbacks have an initial \f(CW\*(C`struct ev_loop *\*(C'\fR argument.
1527.PP 3163.PP
3164To make it easier to write programs that cope with either variant, the
3165following macros are defined:
3166.ie n .IP """EV_A""\fR, \f(CW""EV_A_""" 4
3167.el .IP "\f(CWEV_A\fR, \f(CWEV_A_\fR" 4
3168.IX Item "EV_A, EV_A_"
3169This provides the loop \fIargument\fR for functions, if one is required (\*(L"ev
3170loop argument\*(R"). The \f(CW\*(C`EV_A\*(C'\fR form is used when this is the sole argument,
3171\&\f(CW\*(C`EV_A_\*(C'\fR is used when other arguments are following. Example:
3172.Sp
3173.Vb 3
3174\& ev_unref (EV_A);
3175\& ev_timer_add (EV_A_ watcher);
3176\& ev_loop (EV_A_ 0);
3177.Ve
3178.Sp
3179It assumes the variable \f(CW\*(C`loop\*(C'\fR of type \f(CW\*(C`struct ev_loop *\*(C'\fR is in scope,
3180which is often provided by the following macro.
3181.ie n .IP """EV_P""\fR, \f(CW""EV_P_""" 4
3182.el .IP "\f(CWEV_P\fR, \f(CWEV_P_\fR" 4
3183.IX Item "EV_P, EV_P_"
3184This provides the loop \fIparameter\fR for functions, if one is required (\*(L"ev
3185loop parameter\*(R"). The \f(CW\*(C`EV_P\*(C'\fR form is used when this is the sole parameter,
3186\&\f(CW\*(C`EV_P_\*(C'\fR is used when other parameters are following. Example:
3187.Sp
1528.Vb 2 3188.Vb 2
1529\& myclass (); 3189\& // this is how ev_unref is being declared
1530\& } 3190\& static void ev_unref (EV_P);
3191\&
3192\& // this is how you can declare your typical callback
3193\& static void cb (EV_P_ ev_timer *w, int revents)
1531.Ve 3194.Ve
3195.Sp
3196It declares a parameter \f(CW\*(C`loop\*(C'\fR of type \f(CW\*(C`struct ev_loop *\*(C'\fR, quite
3197suitable for use with \f(CW\*(C`EV_A\*(C'\fR.
3198.ie n .IP """EV_DEFAULT""\fR, \f(CW""EV_DEFAULT_""" 4
3199.el .IP "\f(CWEV_DEFAULT\fR, \f(CWEV_DEFAULT_\fR" 4
3200.IX Item "EV_DEFAULT, EV_DEFAULT_"
3201Similar to the other two macros, this gives you the value of the default
3202loop, if multiple loops are supported (\*(L"ev loop default\*(R").
3203.ie n .IP """EV_DEFAULT_UC""\fR, \f(CW""EV_DEFAULT_UC_""" 4
3204.el .IP "\f(CWEV_DEFAULT_UC\fR, \f(CWEV_DEFAULT_UC_\fR" 4
3205.IX Item "EV_DEFAULT_UC, EV_DEFAULT_UC_"
3206Usage identical to \f(CW\*(C`EV_DEFAULT\*(C'\fR and \f(CW\*(C`EV_DEFAULT_\*(C'\fR, but requires that the
3207default loop has been initialised (\f(CW\*(C`UC\*(C'\fR == unchecked). Their behaviour
3208is undefined when the default loop has not been initialised by a previous
3209execution of \f(CW\*(C`EV_DEFAULT\*(C'\fR, \f(CW\*(C`EV_DEFAULT_\*(C'\fR or \f(CW\*(C`ev_default_init (...)\*(C'\fR.
3210.Sp
3211It is often prudent to use \f(CW\*(C`EV_DEFAULT\*(C'\fR when initialising the first
3212watcher in a function but use \f(CW\*(C`EV_DEFAULT_UC\*(C'\fR afterwards.
1532.PP 3213.PP
3214Example: Declare and initialise a check watcher, utilising the above
3215macros so it will work regardless of whether multiple loops are supported
3216or not.
3217.PP
1533.Vb 6 3218.Vb 5
1534\& myclass::myclass (int fd) 3219\& static void
1535\& : io (this, &myclass::io_cb), 3220\& check_cb (EV_P_ ev_timer *w, int revents)
1536\& idle (this, &myclass::idle_cb)
1537\& { 3221\& {
1538\& io.start (fd, ev::READ); 3222\& ev_check_stop (EV_A_ w);
1539\& } 3223\& }
3224\&
3225\& ev_check check;
3226\& ev_check_init (&check, check_cb);
3227\& ev_check_start (EV_DEFAULT_ &check);
3228\& ev_loop (EV_DEFAULT_ 0);
1540.Ve 3229.Ve
1541.SH "EMBEDDING" 3230.SH "EMBEDDING"
1542.IX Header "EMBEDDING" 3231.IX Header "EMBEDDING"
1543Libev can (and often is) directly embedded into host 3232Libev can (and often is) directly embedded into host
1544applications. Examples of applications that embed it include the Deliantra 3233applications. Examples of applications that embed it include the Deliantra
1545Game Server, the \s-1EV\s0 perl module, the \s-1GNU\s0 Virtual Private Ethernet (gvpe) 3234Game Server, the \s-1EV\s0 perl module, the \s-1GNU\s0 Virtual Private Ethernet (gvpe)
1546and rxvt\-unicode. 3235and rxvt-unicode.
1547.PP 3236.PP
1548The goal is to enable you to just copy the neecssary files into your 3237The goal is to enable you to just copy the necessary files into your
1549source directory without having to change even a single line in them, so 3238source directory without having to change even a single line in them, so
1550you can easily upgrade by simply copying (or having a checked-out copy of 3239you can easily upgrade by simply copying (or having a checked-out copy of
1551libev somewhere in your source tree). 3240libev somewhere in your source tree).
1552.Sh "\s-1FILESETS\s0" 3241.Sh "\s-1FILESETS\s0"
1553.IX Subsection "FILESETS" 3242.IX Subsection "FILESETS"
1554Depending on what features you need you need to include one or more sets of files 3243Depending on what features you need you need to include one or more sets of files
1555in your app. 3244in your application.
1556.PP 3245.PP
1557\fI\s-1CORE\s0 \s-1EVENT\s0 \s-1LOOP\s0\fR 3246\fI\s-1CORE\s0 \s-1EVENT\s0 \s-1LOOP\s0\fR
1558.IX Subsection "CORE EVENT LOOP" 3247.IX Subsection "CORE EVENT LOOP"
1559.PP 3248.PP
1560To include only the libev core (all the \f(CW\*(C`ev_*\*(C'\fR functions), with manual 3249To include only the libev core (all the \f(CW\*(C`ev_*\*(C'\fR functions), with manual
1561configuration (no autoconf): 3250configuration (no autoconf):
1562.PP 3251.PP
1563.Vb 2 3252.Vb 2
1564\& #define EV_STANDALONE 1 3253\& #define EV_STANDALONE 1
1565\& #include "ev.c" 3254\& #include "ev.c"
1566.Ve 3255.Ve
1567.PP 3256.PP
1568This will automatically include \fIev.h\fR, too, and should be done in a 3257This will automatically include \fIev.h\fR, too, and should be done in a
1569single C source file only to provide the function implementations. To use 3258single C source file only to provide the function implementations. To use
1570it, do the same for \fIev.h\fR in all files wishing to use this \s-1API\s0 (best 3259it, do the same for \fIev.h\fR in all files wishing to use this \s-1API\s0 (best
1571done by writing a wrapper around \fIev.h\fR that you can include instead and 3260done by writing a wrapper around \fIev.h\fR that you can include instead and
1572where you can put other configuration options): 3261where you can put other configuration options):
1573.PP 3262.PP
1574.Vb 2 3263.Vb 2
1575\& #define EV_STANDALONE 1 3264\& #define EV_STANDALONE 1
1576\& #include "ev.h" 3265\& #include "ev.h"
1577.Ve 3266.Ve
1578.PP 3267.PP
1579Both header files and implementation files can be compiled with a \*(C+ 3268Both header files and implementation files can be compiled with a \*(C+
1580compiler (at least, thats a stated goal, and breakage will be treated 3269compiler (at least, that's a stated goal, and breakage will be treated
1581as a bug). 3270as a bug).
1582.PP 3271.PP
1583You need the following files in your source tree, or in a directory 3272You need the following files in your source tree, or in a directory
1584in your include path (e.g. in libev/ when using \-Ilibev): 3273in your include path (e.g. in libev/ when using \-Ilibev):
1585.PP 3274.PP
1586.Vb 4 3275.Vb 4
1587\& ev.h 3276\& ev.h
1588\& ev.c 3277\& ev.c
1589\& ev_vars.h 3278\& ev_vars.h
1590\& ev_wrap.h 3279\& ev_wrap.h
1591.Ve 3280\&
1592.PP
1593.Vb 1
1594\& ev_win32.c required on win32 platforms only 3281\& ev_win32.c required on win32 platforms only
1595.Ve 3282\&
1596.PP
1597.Vb 5
1598\& ev_select.c only when select backend is enabled (which is is by default) 3283\& ev_select.c only when select backend is enabled (which is enabled by default)
1599\& ev_poll.c only when poll backend is enabled (disabled by default) 3284\& ev_poll.c only when poll backend is enabled (disabled by default)
1600\& ev_epoll.c only when the epoll backend is enabled (disabled by default) 3285\& ev_epoll.c only when the epoll backend is enabled (disabled by default)
1601\& ev_kqueue.c only when the kqueue backend is enabled (disabled by default) 3286\& ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
1602\& ev_port.c only when the solaris port backend is enabled (disabled by default) 3287\& ev_port.c only when the solaris port backend is enabled (disabled by default)
1603.Ve 3288.Ve
1604.PP 3289.PP
1605\&\fIev.c\fR includes the backend files directly when enabled, so you only need 3290\&\fIev.c\fR includes the backend files directly when enabled, so you only need
1606to compile a single file. 3291to compile this single file.
1607.PP 3292.PP
1608\fI\s-1LIBEVENT\s0 \s-1COMPATIBILITY\s0 \s-1API\s0\fR 3293\fI\s-1LIBEVENT\s0 \s-1COMPATIBILITY\s0 \s-1API\s0\fR
1609.IX Subsection "LIBEVENT COMPATIBILITY API" 3294.IX Subsection "LIBEVENT COMPATIBILITY API"
1610.PP 3295.PP
1611To include the libevent compatibility \s-1API\s0, also include: 3296To include the libevent compatibility \s-1API\s0, also include:
1612.PP 3297.PP
1613.Vb 1 3298.Vb 1
1614\& #include "event.c" 3299\& #include "event.c"
1615.Ve 3300.Ve
1616.PP 3301.PP
1617in the file including \fIev.c\fR, and: 3302in the file including \fIev.c\fR, and:
1618.PP 3303.PP
1619.Vb 1 3304.Vb 1
1620\& #include "event.h" 3305\& #include "event.h"
1621.Ve 3306.Ve
1622.PP 3307.PP
1623in the files that want to use the libevent \s-1API\s0. This also includes \fIev.h\fR. 3308in the files that want to use the libevent \s-1API\s0. This also includes \fIev.h\fR.
1624.PP 3309.PP
1625You need the following additional files for this: 3310You need the following additional files for this:
1626.PP 3311.PP
1627.Vb 2 3312.Vb 2
1628\& event.h 3313\& event.h
1629\& event.c 3314\& event.c
1630.Ve 3315.Ve
1631.PP 3316.PP
1632\fI\s-1AUTOCONF\s0 \s-1SUPPORT\s0\fR 3317\fI\s-1AUTOCONF\s0 \s-1SUPPORT\s0\fR
1633.IX Subsection "AUTOCONF SUPPORT" 3318.IX Subsection "AUTOCONF SUPPORT"
1634.PP 3319.PP
1635Instead of using \f(CW\*(C`EV_STANDALONE=1\*(C'\fR and providing your config in 3320Instead of using \f(CW\*(C`EV_STANDALONE=1\*(C'\fR and providing your configuration in
1636whatever way you want, you can also \f(CW\*(C`m4_include([libev.m4])\*(C'\fR in your 3321whatever way you want, you can also \f(CW\*(C`m4_include([libev.m4])\*(C'\fR in your
1637\&\fIconfigure.ac\fR and leave \f(CW\*(C`EV_STANDALONE\*(C'\fR off. \fIev.c\fR will then include 3322\&\fIconfigure.ac\fR and leave \f(CW\*(C`EV_STANDALONE\*(C'\fR undefined. \fIev.c\fR will then
1638\&\fIconfig.h\fR and configure itself accordingly. 3323include \fIconfig.h\fR and configure itself accordingly.
1639.PP 3324.PP
1640For this of course you need the m4 file: 3325For this of course you need the m4 file:
1641.PP 3326.PP
1642.Vb 1 3327.Vb 1
1643\& libev.m4 3328\& libev.m4
1644.Ve 3329.Ve
1645.Sh "\s-1PREPROCESSOR\s0 \s-1SYMBOLS/MACROS\s0" 3330.Sh "\s-1PREPROCESSOR\s0 \s-1SYMBOLS/MACROS\s0"
1646.IX Subsection "PREPROCESSOR SYMBOLS/MACROS" 3331.IX Subsection "PREPROCESSOR SYMBOLS/MACROS"
1647Libev can be configured via a variety of preprocessor symbols you have to define 3332Libev can be configured via a variety of preprocessor symbols you have to
1648before including any of its files. The default is not to build for multiplicity 3333define before including any of its files. The default in the absence of
1649and only include the select backend. 3334autoconf is documented for every option.
1650.IP "\s-1EV_STANDALONE\s0" 4 3335.IP "\s-1EV_STANDALONE\s0" 4
1651.IX Item "EV_STANDALONE" 3336.IX Item "EV_STANDALONE"
1652Must always be \f(CW1\fR if you do not use autoconf configuration, which 3337Must always be \f(CW1\fR if you do not use autoconf configuration, which
1653keeps libev from including \fIconfig.h\fR, and it also defines dummy 3338keeps libev from including \fIconfig.h\fR, and it also defines dummy
1654implementations for some libevent functions (such as logging, which is not 3339implementations for some libevent functions (such as logging, which is not
1655supported). It will also not define any of the structs usually found in 3340supported). It will also not define any of the structs usually found in
1656\&\fIevent.h\fR that are not directly supported by the libev core alone. 3341\&\fIevent.h\fR that are not directly supported by the libev core alone.
3342.Sp
3343In stanbdalone mode, libev will still try to automatically deduce the
3344configuration, but has to be more conservative.
1657.IP "\s-1EV_USE_MONOTONIC\s0" 4 3345.IP "\s-1EV_USE_MONOTONIC\s0" 4
1658.IX Item "EV_USE_MONOTONIC" 3346.IX Item "EV_USE_MONOTONIC"
1659If defined to be \f(CW1\fR, libev will try to detect the availability of the 3347If defined to be \f(CW1\fR, libev will try to detect the availability of the
1660monotonic clock option at both compiletime and runtime. Otherwise no use 3348monotonic clock option at both compile time and runtime. Otherwise no
1661of the monotonic clock option will be attempted. If you enable this, you 3349use of the monotonic clock option will be attempted. If you enable this,
1662usually have to link against librt or something similar. Enabling it when 3350you usually have to link against librt or something similar. Enabling it
1663the functionality isn't available is safe, though, althoguh you have 3351when the functionality isn't available is safe, though, although you have
1664to make sure you link against any libraries where the \f(CW\*(C`clock_gettime\*(C'\fR 3352to make sure you link against any libraries where the \f(CW\*(C`clock_gettime\*(C'\fR
1665function is hiding in (often \fI\-lrt\fR). 3353function is hiding in (often \fI\-lrt\fR). See also \f(CW\*(C`EV_USE_CLOCK_SYSCALL\*(C'\fR.
1666.IP "\s-1EV_USE_REALTIME\s0" 4 3354.IP "\s-1EV_USE_REALTIME\s0" 4
1667.IX Item "EV_USE_REALTIME" 3355.IX Item "EV_USE_REALTIME"
1668If defined to be \f(CW1\fR, libev will try to detect the availability of the 3356If defined to be \f(CW1\fR, libev will try to detect the availability of the
1669realtime clock option at compiletime (and assume its availability at 3357real-time clock option at compile time (and assume its availability
1670runtime if successful). Otherwise no use of the realtime clock option will 3358at runtime if successful). Otherwise no use of the real-time clock
1671be attempted. This effectively replaces \f(CW\*(C`gettimeofday\*(C'\fR by \f(CW\*(C`clock_get 3359option will be attempted. This effectively replaces \f(CW\*(C`gettimeofday\*(C'\fR
1672(CLOCK_REALTIME, ...)\*(C'\fR and will not normally affect correctness. See tzhe note about libraries 3360by \f(CW\*(C`clock_get (CLOCK_REALTIME, ...)\*(C'\fR and will not normally affect
1673in the description of \f(CW\*(C`EV_USE_MONOTONIC\*(C'\fR, though. 3361correctness. See the note about libraries in the description of
3362\&\f(CW\*(C`EV_USE_MONOTONIC\*(C'\fR, though. Defaults to the opposite value of
3363\&\f(CW\*(C`EV_USE_CLOCK_SYSCALL\*(C'\fR.
3364.IP "\s-1EV_USE_CLOCK_SYSCALL\s0" 4
3365.IX Item "EV_USE_CLOCK_SYSCALL"
3366If defined to be \f(CW1\fR, libev will try to use a direct syscall instead
3367of calling the system-provided \f(CW\*(C`clock_gettime\*(C'\fR function. This option
3368exists because on GNU/Linux, \f(CW\*(C`clock_gettime\*(C'\fR is in \f(CW\*(C`librt\*(C'\fR, but \f(CW\*(C`librt\*(C'\fR
3369unconditionally pulls in \f(CW\*(C`libpthread\*(C'\fR, slowing down single-threaded
3370programs needlessly. Using a direct syscall is slightly slower (in
3371theory), because no optimised vdso implementation can be used, but avoids
3372the pthread dependency. Defaults to \f(CW1\fR on GNU/Linux with glibc 2.x or
3373higher, as it simplifies linking (no need for \f(CW\*(C`\-lrt\*(C'\fR).
3374.IP "\s-1EV_USE_NANOSLEEP\s0" 4
3375.IX Item "EV_USE_NANOSLEEP"
3376If defined to be \f(CW1\fR, libev will assume that \f(CW\*(C`nanosleep ()\*(C'\fR is available
3377and will use it for delays. Otherwise it will use \f(CW\*(C`select ()\*(C'\fR.
3378.IP "\s-1EV_USE_EVENTFD\s0" 4
3379.IX Item "EV_USE_EVENTFD"
3380If defined to be \f(CW1\fR, then libev will assume that \f(CW\*(C`eventfd ()\*(C'\fR is
3381available and will probe for kernel support at runtime. This will improve
3382\&\f(CW\*(C`ev_signal\*(C'\fR and \f(CW\*(C`ev_async\*(C'\fR performance and reduce resource consumption.
3383If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
33842.7 or newer, otherwise disabled.
1674.IP "\s-1EV_USE_SELECT\s0" 4 3385.IP "\s-1EV_USE_SELECT\s0" 4
1675.IX Item "EV_USE_SELECT" 3386.IX Item "EV_USE_SELECT"
1676If undefined or defined to be \f(CW1\fR, libev will compile in support for the 3387If undefined or defined to be \f(CW1\fR, libev will compile in support for the
1677\&\f(CW\*(C`select\*(C'\fR(2) backend. No attempt at autodetection will be done: if no 3388\&\f(CW\*(C`select\*(C'\fR(2) backend. No attempt at auto-detection will be done: if no
1678other method takes over, select will be it. Otherwise the select backend 3389other method takes over, select will be it. Otherwise the select backend
1679will not be compiled in. 3390will not be compiled in.
1680.IP "\s-1EV_SELECT_USE_FD_SET\s0" 4 3391.IP "\s-1EV_SELECT_USE_FD_SET\s0" 4
1681.IX Item "EV_SELECT_USE_FD_SET" 3392.IX Item "EV_SELECT_USE_FD_SET"
1682If defined to \f(CW1\fR, then the select backend will use the system \f(CW\*(C`fd_set\*(C'\fR 3393If defined to \f(CW1\fR, then the select backend will use the system \f(CW\*(C`fd_set\*(C'\fR
1683structure. This is useful if libev doesn't compile due to a missing 3394structure. This is useful if libev doesn't compile due to a missing
1684\&\f(CW\*(C`NFDBITS\*(C'\fR or \f(CW\*(C`fd_mask\*(C'\fR definition or it misguesses the bitset layout on 3395\&\f(CW\*(C`NFDBITS\*(C'\fR or \f(CW\*(C`fd_mask\*(C'\fR definition or it mis-guesses the bitset layout
1685exotic systems. This usually limits the range of file descriptors to some 3396on exotic systems. This usually limits the range of file descriptors to
1686low limit such as 1024 or might have other limitations (winsocket only 3397some low limit such as 1024 or might have other limitations (winsocket
1687allows 64 sockets). The \f(CW\*(C`FD_SETSIZE\*(C'\fR macro, set before compilation, might 3398only allows 64 sockets). The \f(CW\*(C`FD_SETSIZE\*(C'\fR macro, set before compilation,
1688influence the size of the \f(CW\*(C`fd_set\*(C'\fR used. 3399configures the maximum size of the \f(CW\*(C`fd_set\*(C'\fR.
1689.IP "\s-1EV_SELECT_IS_WINSOCKET\s0" 4 3400.IP "\s-1EV_SELECT_IS_WINSOCKET\s0" 4
1690.IX Item "EV_SELECT_IS_WINSOCKET" 3401.IX Item "EV_SELECT_IS_WINSOCKET"
1691When defined to \f(CW1\fR, the select backend will assume that 3402When defined to \f(CW1\fR, the select backend will assume that
1692select/socket/connect etc. don't understand file descriptors but 3403select/socket/connect etc. don't understand file descriptors but
1693wants osf handles on win32 (this is the case when the select to 3404wants osf handles on win32 (this is the case when the select to
1694be used is the winsock select). This means that it will call 3405be used is the winsock select). This means that it will call
1695\&\f(CW\*(C`_get_osfhandle\*(C'\fR on the fd to convert it to an \s-1OS\s0 handle. Otherwise, 3406\&\f(CW\*(C`_get_osfhandle\*(C'\fR on the fd to convert it to an \s-1OS\s0 handle. Otherwise,
1696it is assumed that all these functions actually work on fds, even 3407it is assumed that all these functions actually work on fds, even
1697on win32. Should not be defined on non\-win32 platforms. 3408on win32. Should not be defined on non\-win32 platforms.
3409.IP "\s-1EV_FD_TO_WIN32_HANDLE\s0" 4
3410.IX Item "EV_FD_TO_WIN32_HANDLE"
3411If \f(CW\*(C`EV_SELECT_IS_WINSOCKET\*(C'\fR is enabled, then libev needs a way to map
3412file descriptors to socket handles. When not defining this symbol (the
3413default), then libev will call \f(CW\*(C`_get_osfhandle\*(C'\fR, which is usually
3414correct. In some cases, programs use their own file descriptor management,
3415in which case they can provide this function to map fds to socket handles.
1698.IP "\s-1EV_USE_POLL\s0" 4 3416.IP "\s-1EV_USE_POLL\s0" 4
1699.IX Item "EV_USE_POLL" 3417.IX Item "EV_USE_POLL"
1700If defined to be \f(CW1\fR, libev will compile in support for the \f(CW\*(C`poll\*(C'\fR(2) 3418If defined to be \f(CW1\fR, libev will compile in support for the \f(CW\*(C`poll\*(C'\fR(2)
1701backend. Otherwise it will be enabled on non\-win32 platforms. It 3419backend. Otherwise it will be enabled on non\-win32 platforms. It
1702takes precedence over select. 3420takes precedence over select.
1703.IP "\s-1EV_USE_EPOLL\s0" 4 3421.IP "\s-1EV_USE_EPOLL\s0" 4
1704.IX Item "EV_USE_EPOLL" 3422.IX Item "EV_USE_EPOLL"
1705If defined to be \f(CW1\fR, libev will compile in support for the Linux 3423If defined to be \f(CW1\fR, libev will compile in support for the Linux
1706\&\f(CW\*(C`epoll\*(C'\fR(7) backend. Its availability will be detected at runtime, 3424\&\f(CW\*(C`epoll\*(C'\fR(7) backend. Its availability will be detected at runtime,
1707otherwise another method will be used as fallback. This is the 3425otherwise another method will be used as fallback. This is the preferred
1708preferred backend for GNU/Linux systems. 3426backend for GNU/Linux systems. If undefined, it will be enabled if the
3427headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
1709.IP "\s-1EV_USE_KQUEUE\s0" 4 3428.IP "\s-1EV_USE_KQUEUE\s0" 4
1710.IX Item "EV_USE_KQUEUE" 3429.IX Item "EV_USE_KQUEUE"
1711If defined to be \f(CW1\fR, libev will compile in support for the \s-1BSD\s0 style 3430If defined to be \f(CW1\fR, libev will compile in support for the \s-1BSD\s0 style
1712\&\f(CW\*(C`kqueue\*(C'\fR(2) backend. Its actual availability will be detected at runtime, 3431\&\f(CW\*(C`kqueue\*(C'\fR(2) backend. Its actual availability will be detected at runtime,
1713otherwise another method will be used as fallback. This is the preferred 3432otherwise another method will be used as fallback. This is the preferred
… …
172310 port style backend. Its availability will be detected at runtime, 344210 port style backend. Its availability will be detected at runtime,
1724otherwise another method will be used as fallback. This is the preferred 3443otherwise another method will be used as fallback. This is the preferred
1725backend for Solaris 10 systems. 3444backend for Solaris 10 systems.
1726.IP "\s-1EV_USE_DEVPOLL\s0" 4 3445.IP "\s-1EV_USE_DEVPOLL\s0" 4
1727.IX Item "EV_USE_DEVPOLL" 3446.IX Item "EV_USE_DEVPOLL"
1728reserved for future expansion, works like the \s-1USE\s0 symbols above. 3447Reserved for future expansion, works like the \s-1USE\s0 symbols above.
3448.IP "\s-1EV_USE_INOTIFY\s0" 4
3449.IX Item "EV_USE_INOTIFY"
3450If defined to be \f(CW1\fR, libev will compile in support for the Linux inotify
3451interface to speed up \f(CW\*(C`ev_stat\*(C'\fR watchers. Its actual availability will
3452be detected at runtime. If undefined, it will be enabled if the headers
3453indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
3454.IP "\s-1EV_ATOMIC_T\s0" 4
3455.IX Item "EV_ATOMIC_T"
3456Libev requires an integer type (suitable for storing \f(CW0\fR or \f(CW1\fR) whose
3457access is atomic with respect to other threads or signal contexts. No such
3458type is easily found in the C language, so you can provide your own type
3459that you know is safe for your purposes. It is used both for signal handler \*(L"locking\*(R"
3460as well as for signal and thread safety in \f(CW\*(C`ev_async\*(C'\fR watchers.
3461.Sp
3462In the absence of this define, libev will use \f(CW\*(C`sig_atomic_t volatile\*(C'\fR
3463(from \fIsignal.h\fR), which is usually good enough on most platforms.
1729.IP "\s-1EV_H\s0" 4 3464.IP "\s-1EV_H\s0" 4
1730.IX Item "EV_H" 3465.IX Item "EV_H"
1731The name of the \fIev.h\fR header file used to include it. The default if 3466The name of the \fIev.h\fR header file used to include it. The default if
1732undefined is \f(CW\*(C`<ev.h>\*(C'\fR in \fIevent.h\fR and \f(CW"ev.h"\fR in \fIev.c\fR. This 3467undefined is \f(CW"ev.h"\fR in \fIevent.h\fR, \fIev.c\fR and \fIev++.h\fR. This can be
1733can be used to virtually rename the \fIev.h\fR header file in case of conflicts. 3468used to virtually rename the \fIev.h\fR header file in case of conflicts.
1734.IP "\s-1EV_CONFIG_H\s0" 4 3469.IP "\s-1EV_CONFIG_H\s0" 4
1735.IX Item "EV_CONFIG_H" 3470.IX Item "EV_CONFIG_H"
1736If \f(CW\*(C`EV_STANDALONE\*(C'\fR isn't \f(CW1\fR, this variable can be used to override 3471If \f(CW\*(C`EV_STANDALONE\*(C'\fR isn't \f(CW1\fR, this variable can be used to override
1737\&\fIev.c\fR's idea of where to find the \fIconfig.h\fR file, similarly to 3472\&\fIev.c\fR's idea of where to find the \fIconfig.h\fR file, similarly to
1738\&\f(CW\*(C`EV_H\*(C'\fR, above. 3473\&\f(CW\*(C`EV_H\*(C'\fR, above.
1739.IP "\s-1EV_EVENT_H\s0" 4 3474.IP "\s-1EV_EVENT_H\s0" 4
1740.IX Item "EV_EVENT_H" 3475.IX Item "EV_EVENT_H"
1741Similarly to \f(CW\*(C`EV_H\*(C'\fR, this macro can be used to override \fIevent.c\fR's idea 3476Similarly to \f(CW\*(C`EV_H\*(C'\fR, this macro can be used to override \fIevent.c\fR's idea
1742of how the \fIevent.h\fR header can be found. 3477of how the \fIevent.h\fR header can be found, the default is \f(CW"event.h"\fR.
1743.IP "\s-1EV_PROTOTYPES\s0" 4 3478.IP "\s-1EV_PROTOTYPES\s0" 4
1744.IX Item "EV_PROTOTYPES" 3479.IX Item "EV_PROTOTYPES"
1745If defined to be \f(CW0\fR, then \fIev.h\fR will not define any function 3480If defined to be \f(CW0\fR, then \fIev.h\fR will not define any function
1746prototypes, but still define all the structs and other symbols. This is 3481prototypes, but still define all the structs and other symbols. This is
1747occasionally useful if you want to provide your own wrapper functions 3482occasionally useful if you want to provide your own wrapper functions
… …
1751If undefined or defined to \f(CW1\fR, then all event-loop-specific functions 3486If undefined or defined to \f(CW1\fR, then all event-loop-specific functions
1752will have the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument, and you can create 3487will have the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument, and you can create
1753additional independent event loops. Otherwise there will be no support 3488additional independent event loops. Otherwise there will be no support
1754for multiple event loops and there is no first event loop pointer 3489for multiple event loops and there is no first event loop pointer
1755argument. Instead, all functions act on the single default loop. 3490argument. Instead, all functions act on the single default loop.
3491.IP "\s-1EV_MINPRI\s0" 4
3492.IX Item "EV_MINPRI"
3493.PD 0
3494.IP "\s-1EV_MAXPRI\s0" 4
3495.IX Item "EV_MAXPRI"
3496.PD
3497The range of allowed priorities. \f(CW\*(C`EV_MINPRI\*(C'\fR must be smaller or equal to
3498\&\f(CW\*(C`EV_MAXPRI\*(C'\fR, but otherwise there are no non-obvious limitations. You can
3499provide for more priorities by overriding those symbols (usually defined
3500to be \f(CW\*(C`\-2\*(C'\fR and \f(CW2\fR, respectively).
3501.Sp
3502When doing priority-based operations, libev usually has to linearly search
3503all the priorities, so having many of them (hundreds) uses a lot of space
3504and time, so using the defaults of five priorities (\-2 .. +2) is usually
3505fine.
3506.Sp
3507If your embedding application does not need any priorities, defining these
3508both to \f(CW0\fR will save some memory and \s-1CPU\s0.
1756.IP "\s-1EV_PERIODICS\s0" 4 3509.IP "\s-1EV_PERIODIC_ENABLE\s0" 4
1757.IX Item "EV_PERIODICS" 3510.IX Item "EV_PERIODIC_ENABLE"
1758If undefined or defined to be \f(CW1\fR, then periodic timers are supported, 3511If undefined or defined to be \f(CW1\fR, then periodic timers are supported. If
1759otherwise not. This saves a few kb of code. 3512defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of
3513code.
3514.IP "\s-1EV_IDLE_ENABLE\s0" 4
3515.IX Item "EV_IDLE_ENABLE"
3516If undefined or defined to be \f(CW1\fR, then idle watchers are supported. If
3517defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of
3518code.
3519.IP "\s-1EV_EMBED_ENABLE\s0" 4
3520.IX Item "EV_EMBED_ENABLE"
3521If undefined or defined to be \f(CW1\fR, then embed watchers are supported. If
3522defined to be \f(CW0\fR, then they are not. Embed watchers rely on most other
3523watcher types, which therefore must not be disabled.
3524.IP "\s-1EV_STAT_ENABLE\s0" 4
3525.IX Item "EV_STAT_ENABLE"
3526If undefined or defined to be \f(CW1\fR, then stat watchers are supported. If
3527defined to be \f(CW0\fR, then they are not.
3528.IP "\s-1EV_FORK_ENABLE\s0" 4
3529.IX Item "EV_FORK_ENABLE"
3530If undefined or defined to be \f(CW1\fR, then fork watchers are supported. If
3531defined to be \f(CW0\fR, then they are not.
3532.IP "\s-1EV_ASYNC_ENABLE\s0" 4
3533.IX Item "EV_ASYNC_ENABLE"
3534If undefined or defined to be \f(CW1\fR, then async watchers are supported. If
3535defined to be \f(CW0\fR, then they are not.
3536.IP "\s-1EV_MINIMAL\s0" 4
3537.IX Item "EV_MINIMAL"
3538If you need to shave off some kilobytes of code at the expense of some
3539speed, define this symbol to \f(CW1\fR. Currently this is used to override some
3540inlining decisions, saves roughly 30% code size on amd64. It also selects a
3541much smaller 2\-heap for timer management over the default 4\-heap.
3542.IP "\s-1EV_PID_HASHSIZE\s0" 4
3543.IX Item "EV_PID_HASHSIZE"
3544\&\f(CW\*(C`ev_child\*(C'\fR watchers use a small hash table to distribute workload by
3545pid. The default size is \f(CW16\fR (or \f(CW1\fR with \f(CW\*(C`EV_MINIMAL\*(C'\fR), usually more
3546than enough. If you need to manage thousands of children you might want to
3547increase this value (\fImust\fR be a power of two).
3548.IP "\s-1EV_INOTIFY_HASHSIZE\s0" 4
3549.IX Item "EV_INOTIFY_HASHSIZE"
3550\&\f(CW\*(C`ev_stat\*(C'\fR watchers use a small hash table to distribute workload by
3551inotify watch id. The default size is \f(CW16\fR (or \f(CW1\fR with \f(CW\*(C`EV_MINIMAL\*(C'\fR),
3552usually more than enough. If you need to manage thousands of \f(CW\*(C`ev_stat\*(C'\fR
3553watchers you might want to increase this value (\fImust\fR be a power of
3554two).
3555.IP "\s-1EV_USE_4HEAP\s0" 4
3556.IX Item "EV_USE_4HEAP"
3557Heaps are not very cache-efficient. To improve the cache-efficiency of the
3558timer and periodics heaps, libev uses a 4\-heap when this symbol is defined
3559to \f(CW1\fR. The 4\-heap uses more complicated (longer) code but has noticeably
3560faster performance with many (thousands) of watchers.
3561.Sp
3562The default is \f(CW1\fR unless \f(CW\*(C`EV_MINIMAL\*(C'\fR is set in which case it is \f(CW0\fR
3563(disabled).
3564.IP "\s-1EV_HEAP_CACHE_AT\s0" 4
3565.IX Item "EV_HEAP_CACHE_AT"
3566Heaps are not very cache-efficient. To improve the cache-efficiency of the
3567timer and periodics heaps, libev can cache the timestamp (\fIat\fR) within
3568the heap structure (selected by defining \f(CW\*(C`EV_HEAP_CACHE_AT\*(C'\fR to \f(CW1\fR),
3569which uses 8\-12 bytes more per watcher and a few hundred bytes more code,
3570but avoids random read accesses on heap changes. This improves performance
3571noticeably with many (hundreds) of watchers.
3572.Sp
3573The default is \f(CW1\fR unless \f(CW\*(C`EV_MINIMAL\*(C'\fR is set in which case it is \f(CW0\fR
3574(disabled).
3575.IP "\s-1EV_VERIFY\s0" 4
3576.IX Item "EV_VERIFY"
3577Controls how much internal verification (see \f(CW\*(C`ev_loop_verify ()\*(C'\fR) will
3578be done: If set to \f(CW0\fR, no internal verification code will be compiled
3579in. If set to \f(CW1\fR, then verification code will be compiled in, but not
3580called. If set to \f(CW2\fR, then the internal verification code will be
3581called once per loop, which can slow down libev. If set to \f(CW3\fR, then the
3582verification code will be called very frequently, which will slow down
3583libev considerably.
3584.Sp
3585The default is \f(CW1\fR, unless \f(CW\*(C`EV_MINIMAL\*(C'\fR is set, in which case it will be
3586\&\f(CW0\fR.
1760.IP "\s-1EV_COMMON\s0" 4 3587.IP "\s-1EV_COMMON\s0" 4
1761.IX Item "EV_COMMON" 3588.IX Item "EV_COMMON"
1762By default, all watchers have a \f(CW\*(C`void *data\*(C'\fR member. By redefining 3589By default, all watchers have a \f(CW\*(C`void *data\*(C'\fR member. By redefining
1763this macro to a something else you can include more and other types of 3590this macro to a something else you can include more and other types of
1764members. You have to define it each time you include one of the files, 3591members. You have to define it each time you include one of the files,
1765though, and it must be identical each time. 3592though, and it must be identical each time.
1766.Sp 3593.Sp
1767For example, the perl \s-1EV\s0 module uses something like this: 3594For example, the perl \s-1EV\s0 module uses something like this:
1768.Sp 3595.Sp
1769.Vb 3 3596.Vb 3
1770\& #define EV_COMMON \e 3597\& #define EV_COMMON \e
1771\& SV *self; /* contains this struct */ \e 3598\& SV *self; /* contains this struct */ \e
1772\& SV *cb_sv, *fh /* note no trailing ";" */ 3599\& SV *cb_sv, *fh /* note no trailing ";" */
1773.Ve 3600.Ve
1774.IP "\s-1EV_CB_DECLARE\s0(type)" 4 3601.IP "\s-1EV_CB_DECLARE\s0 (type)" 4
1775.IX Item "EV_CB_DECLARE(type)" 3602.IX Item "EV_CB_DECLARE (type)"
1776.PD 0 3603.PD 0
1777.IP "\s-1EV_CB_INVOKE\s0(watcher,revents)" 4 3604.IP "\s-1EV_CB_INVOKE\s0 (watcher, revents)" 4
1778.IX Item "EV_CB_INVOKE(watcher,revents)" 3605.IX Item "EV_CB_INVOKE (watcher, revents)"
1779.IP "ev_set_cb(ev,cb)" 4 3606.IP "ev_set_cb (ev, cb)" 4
1780.IX Item "ev_set_cb(ev,cb)" 3607.IX Item "ev_set_cb (ev, cb)"
1781.PD 3608.PD
1782Can be used to change the callback member declaration in each watcher, 3609Can be used to change the callback member declaration in each watcher,
1783and the way callbacks are invoked and set. Must expand to a struct member 3610and the way callbacks are invoked and set. Must expand to a struct member
1784definition and a statement, respectively. See the \fIev.v\fR header file for 3611definition and a statement, respectively. See the \fIev.h\fR header file for
1785their default definitions. One possible use for overriding these is to 3612their default definitions. One possible use for overriding these is to
1786avoid the ev_loop pointer as first argument in all cases, or to use method 3613avoid the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument in all cases, or to use
1787calls instead of plain function calls in \*(C+. 3614method calls instead of plain function calls in \*(C+.
3615.Sh "\s-1EXPORTED\s0 \s-1API\s0 \s-1SYMBOLS\s0"
3616.IX Subsection "EXPORTED API SYMBOLS"
3617If you need to re-export the \s-1API\s0 (e.g. via a \s-1DLL\s0) and you need a list of
3618exported symbols, you can use the provided \fISymbol.*\fR files which list
3619all public symbols, one per line:
3620.PP
3621.Vb 2
3622\& Symbols.ev for libev proper
3623\& Symbols.event for the libevent emulation
3624.Ve
3625.PP
3626This can also be used to rename all public symbols to avoid clashes with
3627multiple versions of libev linked together (which is obviously bad in
3628itself, but sometimes it is inconvenient to avoid this).
3629.PP
3630A sed command like this will create wrapper \f(CW\*(C`#define\*(C'\fR's that you need to
3631include before including \fIev.h\fR:
3632.PP
3633.Vb 1
3634\& <Symbols.ev sed \-e "s/.*/#define & myprefix_&/" >wrap.h
3635.Ve
3636.PP
3637This would create a file \fIwrap.h\fR which essentially looks like this:
3638.PP
3639.Vb 4
3640\& #define ev_backend myprefix_ev_backend
3641\& #define ev_check_start myprefix_ev_check_start
3642\& #define ev_check_stop myprefix_ev_check_stop
3643\& ...
3644.Ve
1788.Sh "\s-1EXAMPLES\s0" 3645.Sh "\s-1EXAMPLES\s0"
1789.IX Subsection "EXAMPLES" 3646.IX Subsection "EXAMPLES"
1790For a real-world example of a program the includes libev 3647For a real-world example of a program the includes libev
1791verbatim, you can have a look at the \s-1EV\s0 perl module 3648verbatim, you can have a look at the \s-1EV\s0 perl module
1792(<http://software.schmorp.de/pkg/EV.html>). It has the libev files in 3649(<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
1793the \fIlibev/\fR subdirectory and includes them in the \fI\s-1EV/EVAPI\s0.h\fR (public 3650the \fIlibev/\fR subdirectory and includes them in the \fI\s-1EV/EVAPI\s0.h\fR (public
1794interface) and \fI\s-1EV\s0.xs\fR (implementation) files. Only the \fI\s-1EV\s0.xs\fR file 3651interface) and \fI\s-1EV\s0.xs\fR (implementation) files. Only the \fI\s-1EV\s0.xs\fR file
1795will be compiled. It is pretty complex because it provides its own header 3652will be compiled. It is pretty complex because it provides its own header
1796file. 3653file.
1797.Sp 3654.PP
1798The usage in rxvt-unicode is simpler. It has a \fIev_cpp.h\fR header file 3655The usage in rxvt-unicode is simpler. It has a \fIev_cpp.h\fR header file
1799that everybody includes and which overrides some autoconf choices: 3656that everybody includes and which overrides some configure choices:
1800.Sp 3657.PP
1801.Vb 4 3658.Vb 9
3659\& #define EV_MINIMAL 1
1802\& #define EV_USE_POLL 0 3660\& #define EV_USE_POLL 0
1803\& #define EV_MULTIPLICITY 0 3661\& #define EV_MULTIPLICITY 0
1804\& #define EV_PERIODICS 0 3662\& #define EV_PERIODIC_ENABLE 0
3663\& #define EV_STAT_ENABLE 0
3664\& #define EV_FORK_ENABLE 0
1805\& #define EV_CONFIG_H <config.h> 3665\& #define EV_CONFIG_H <config.h>
1806.Ve 3666\& #define EV_MINPRI 0
1807.Sp 3667\& #define EV_MAXPRI 0
1808.Vb 1 3668\&
1809\& #include "ev++.h" 3669\& #include "ev++.h"
1810.Ve 3670.Ve
1811.Sp 3671.PP
1812And a \fIev_cpp.C\fR implementation file that contains libev proper and is compiled: 3672And a \fIev_cpp.C\fR implementation file that contains libev proper and is compiled:
1813.Sp 3673.PP
1814.Vb 2 3674.Vb 2
1815\& #include "ev_cpp.h" 3675\& #include "ev_cpp.h"
1816\& #include "ev.c" 3676\& #include "ev.c"
1817.Ve 3677.Ve
3678.SH "INTERACTION WITH OTHER PROGRAMS OR LIBRARIES"
3679.IX Header "INTERACTION WITH OTHER PROGRAMS OR LIBRARIES"
3680.Sh "\s-1THREADS\s0 \s-1AND\s0 \s-1COROUTINES\s0"
3681.IX Subsection "THREADS AND COROUTINES"
3682\fI\s-1THREADS\s0\fR
3683.IX Subsection "THREADS"
3684.PP
3685All libev functions are reentrant and thread-safe unless explicitly
3686documented otherwise, but libev implements no locking itself. This means
3687that you can use as many loops as you want in parallel, as long as there
3688are no concurrent calls into any libev function with the same loop
3689parameter (\f(CW\*(C`ev_default_*\*(C'\fR calls have an implicit default loop parameter,
3690of course): libev guarantees that different event loops share no data
3691structures that need any locking.
3692.PP
3693Or to put it differently: calls with different loop parameters can be done
3694concurrently from multiple threads, calls with the same loop parameter
3695must be done serially (but can be done from different threads, as long as
3696only one thread ever is inside a call at any point in time, e.g. by using
3697a mutex per loop).
3698.PP
3699Specifically to support threads (and signal handlers), libev implements
3700so-called \f(CW\*(C`ev_async\*(C'\fR watchers, which allow some limited form of
3701concurrency on the same event loop, namely waking it up \*(L"from the
3702outside\*(R".
3703.PP
3704If you want to know which design (one loop, locking, or multiple loops
3705without or something else still) is best for your problem, then I cannot
3706help you, but here is some generic advice:
3707.IP "\(bu" 4
3708most applications have a main thread: use the default libev loop
3709in that thread, or create a separate thread running only the default loop.
3710.Sp
3711This helps integrating other libraries or software modules that use libev
3712themselves and don't care/know about threading.
3713.IP "\(bu" 4
3714one loop per thread is usually a good model.
3715.Sp
3716Doing this is almost never wrong, sometimes a better-performance model
3717exists, but it is always a good start.
3718.IP "\(bu" 4
3719other models exist, such as the leader/follower pattern, where one
3720loop is handed through multiple threads in a kind of round-robin fashion.
3721.Sp
3722Choosing a model is hard \- look around, learn, know that usually you can do
3723better than you currently do :\-)
3724.IP "\(bu" 4
3725often you need to talk to some other thread which blocks in the
3726event loop.
3727.Sp
3728\&\f(CW\*(C`ev_async\*(C'\fR watchers can be used to wake them up from other threads safely
3729(or from signal contexts...).
3730.Sp
3731An example use would be to communicate signals or other events that only
3732work in the default loop by registering the signal watcher with the
3733default loop and triggering an \f(CW\*(C`ev_async\*(C'\fR watcher from the default loop
3734watcher callback into the event loop interested in the signal.
3735.PP
3736\fI\s-1COROUTINES\s0\fR
3737.IX Subsection "COROUTINES"
3738.PP
3739Libev is very accommodating to coroutines (\*(L"cooperative threads\*(R"):
3740libev fully supports nesting calls to its functions from different
3741coroutines (e.g. you can call \f(CW\*(C`ev_loop\*(C'\fR on the same loop from two
3742different coroutines, and switch freely between both coroutines running the
3743loop, as long as you don't confuse yourself). The only exception is that
3744you must not do this from \f(CW\*(C`ev_periodic\*(C'\fR reschedule callbacks.
3745.PP
3746Care has been taken to ensure that libev does not keep local state inside
3747\&\f(CW\*(C`ev_loop\*(C'\fR, and other calls do not usually allow for coroutine switches as
3748they do not call any callbacks.
3749.Sh "\s-1COMPILER\s0 \s-1WARNINGS\s0"
3750.IX Subsection "COMPILER WARNINGS"
3751Depending on your compiler and compiler settings, you might get no or a
3752lot of warnings when compiling libev code. Some people are apparently
3753scared by this.
3754.PP
3755However, these are unavoidable for many reasons. For one, each compiler
3756has different warnings, and each user has different tastes regarding
3757warning options. \*(L"Warn-free\*(R" code therefore cannot be a goal except when
3758targeting a specific compiler and compiler-version.
3759.PP
3760Another reason is that some compiler warnings require elaborate
3761workarounds, or other changes to the code that make it less clear and less
3762maintainable.
3763.PP
3764And of course, some compiler warnings are just plain stupid, or simply
3765wrong (because they don't actually warn about the condition their message
3766seems to warn about). For example, certain older gcc versions had some
3767warnings that resulted an extreme number of false positives. These have
3768been fixed, but some people still insist on making code warn-free with
3769such buggy versions.
3770.PP
3771While libev is written to generate as few warnings as possible,
3772\&\*(L"warn-free\*(R" code is not a goal, and it is recommended not to build libev
3773with any compiler warnings enabled unless you are prepared to cope with
3774them (e.g. by ignoring them). Remember that warnings are just that:
3775warnings, not errors, or proof of bugs.
3776.Sh "\s-1VALGRIND\s0"
3777.IX Subsection "VALGRIND"
3778Valgrind has a special section here because it is a popular tool that is
3779highly useful. Unfortunately, valgrind reports are very hard to interpret.
3780.PP
3781If you think you found a bug (memory leak, uninitialised data access etc.)
3782in libev, then check twice: If valgrind reports something like:
3783.PP
3784.Vb 3
3785\& ==2274== definitely lost: 0 bytes in 0 blocks.
3786\& ==2274== possibly lost: 0 bytes in 0 blocks.
3787\& ==2274== still reachable: 256 bytes in 1 blocks.
3788.Ve
3789.PP
3790Then there is no memory leak, just as memory accounted to global variables
3791is not a memleak \- the memory is still being referenced, and didn't leak.
3792.PP
3793Similarly, under some circumstances, valgrind might report kernel bugs
3794as if it were a bug in libev (e.g. in realloc or in the poll backend,
3795although an acceptable workaround has been found here), or it might be
3796confused.
3797.PP
3798Keep in mind that valgrind is a very good tool, but only a tool. Don't
3799make it into some kind of religion.
3800.PP
3801If you are unsure about something, feel free to contact the mailing list
3802with the full valgrind report and an explanation on why you think this
3803is a bug in libev (best check the archives, too :). However, don't be
3804annoyed when you get a brisk \*(L"this is no bug\*(R" answer and take the chance
3805of learning how to interpret valgrind properly.
3806.PP
3807If you need, for some reason, empty reports from valgrind for your project
3808I suggest using suppression lists.
3809.SH "PORTABILITY NOTES"
3810.IX Header "PORTABILITY NOTES"
3811.Sh "\s-1WIN32\s0 \s-1PLATFORM\s0 \s-1LIMITATIONS\s0 \s-1AND\s0 \s-1WORKAROUNDS\s0"
3812.IX Subsection "WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS"
3813Win32 doesn't support any of the standards (e.g. \s-1POSIX\s0) that libev
3814requires, and its I/O model is fundamentally incompatible with the \s-1POSIX\s0
3815model. Libev still offers limited functionality on this platform in
3816the form of the \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR backend, and only supports socket
3817descriptors. This only applies when using Win32 natively, not when using
3818e.g. cygwin.
3819.PP
3820Lifting these limitations would basically require the full
3821re-implementation of the I/O system. If you are into these kinds of
3822things, then note that glib does exactly that for you in a very portable
3823way (note also that glib is the slowest event library known to man).
3824.PP
3825There is no supported compilation method available on windows except
3826embedding it into other applications.
3827.PP
3828Not a libev limitation but worth mentioning: windows apparently doesn't
3829accept large writes: instead of resulting in a partial write, windows will
3830either accept everything or return \f(CW\*(C`ENOBUFS\*(C'\fR if the buffer is too large,
3831so make sure you only write small amounts into your sockets (less than a
3832megabyte seems safe, but this apparently depends on the amount of memory
3833available).
3834.PP
3835Due to the many, low, and arbitrary limits on the win32 platform and
3836the abysmal performance of winsockets, using a large number of sockets
3837is not recommended (and not reasonable). If your program needs to use
3838more than a hundred or so sockets, then likely it needs to use a totally
3839different implementation for windows, as libev offers the \s-1POSIX\s0 readiness
3840notification model, which cannot be implemented efficiently on windows
3841(Microsoft monopoly games).
3842.PP
3843A typical way to use libev under windows is to embed it (see the embedding
3844section for details) and use the following \fIevwrap.h\fR header file instead
3845of \fIev.h\fR:
3846.PP
3847.Vb 2
3848\& #define EV_STANDALONE /* keeps ev from requiring config.h */
3849\& #define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
3850\&
3851\& #include "ev.h"
3852.Ve
3853.PP
3854And compile the following \fIevwrap.c\fR file into your project (make sure
3855you do \fInot\fR compile the \fIev.c\fR or any other embedded source files!):
3856.PP
3857.Vb 2
3858\& #include "evwrap.h"
3859\& #include "ev.c"
3860.Ve
3861.IP "The winsocket select function" 4
3862.IX Item "The winsocket select function"
3863The winsocket \f(CW\*(C`select\*(C'\fR function doesn't follow \s-1POSIX\s0 in that it
3864requires socket \fIhandles\fR and not socket \fIfile descriptors\fR (it is
3865also extremely buggy). This makes select very inefficient, and also
3866requires a mapping from file descriptors to socket handles (the Microsoft
3867C runtime provides the function \f(CW\*(C`_open_osfhandle\*(C'\fR for this). See the
3868discussion of the \f(CW\*(C`EV_SELECT_USE_FD_SET\*(C'\fR, \f(CW\*(C`EV_SELECT_IS_WINSOCKET\*(C'\fR and
3869\&\f(CW\*(C`EV_FD_TO_WIN32_HANDLE\*(C'\fR preprocessor symbols for more info.
3870.Sp
3871The configuration for a \*(L"naked\*(R" win32 using the Microsoft runtime
3872libraries and raw winsocket select is:
3873.Sp
3874.Vb 2
3875\& #define EV_USE_SELECT 1
3876\& #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
3877.Ve
3878.Sp
3879Note that winsockets handling of fd sets is O(n), so you can easily get a
3880complexity in the O(nA\*^X) range when using win32.
3881.IP "Limited number of file descriptors" 4
3882.IX Item "Limited number of file descriptors"
3883Windows has numerous arbitrary (and low) limits on things.
3884.Sp
3885Early versions of winsocket's select only supported waiting for a maximum
3886of \f(CW64\fR handles (probably owning to the fact that all windows kernels
3887can only wait for \f(CW64\fR things at the same time internally; Microsoft
3888recommends spawning a chain of threads and wait for 63 handles and the
3889previous thread in each. Great).
3890.Sp
3891Newer versions support more handles, but you need to define \f(CW\*(C`FD_SETSIZE\*(C'\fR
3892to some high number (e.g. \f(CW2048\fR) before compiling the winsocket select
3893call (which might be in libev or elsewhere, for example, perl does its own
3894select emulation on windows).
3895.Sp
3896Another limit is the number of file descriptors in the Microsoft runtime
3897libraries, which by default is \f(CW64\fR (there must be a hidden \fI64\fR fetish
3898or something like this inside Microsoft). You can increase this by calling
3899\&\f(CW\*(C`_setmaxstdio\*(C'\fR, which can increase this limit to \f(CW2048\fR (another
3900arbitrary limit), but is broken in many versions of the Microsoft runtime
3901libraries.
3902.Sp
3903This might get you to about \f(CW512\fR or \f(CW2048\fR sockets (depending on
3904windows version and/or the phase of the moon). To get more, you need to
3905wrap all I/O functions and provide your own fd management, but the cost of
3906calling select (O(nA\*^X)) will likely make this unworkable.
3907.Sh "\s-1PORTABILITY\s0 \s-1REQUIREMENTS\s0"
3908.IX Subsection "PORTABILITY REQUIREMENTS"
3909In addition to a working ISO-C implementation and of course the
3910backend-specific APIs, libev relies on a few additional extensions:
3911.ie n .IP """void (*)(ev_watcher_type *, int revents)""\fR must have compatible calling conventions regardless of \f(CW""ev_watcher_type *""." 4
3912.el .IP "\f(CWvoid (*)(ev_watcher_type *, int revents)\fR must have compatible calling conventions regardless of \f(CWev_watcher_type *\fR." 4
3913.IX Item "void (*)(ev_watcher_type *, int revents) must have compatible calling conventions regardless of ev_watcher_type *."
3914Libev assumes not only that all watcher pointers have the same internal
3915structure (guaranteed by \s-1POSIX\s0 but not by \s-1ISO\s0 C for example), but it also
3916assumes that the same (machine) code can be used to call any watcher
3917callback: The watcher callbacks have different type signatures, but libev
3918calls them using an \f(CW\*(C`ev_watcher *\*(C'\fR internally.
3919.ie n .IP """sig_atomic_t volatile"" must be thread-atomic as well" 4
3920.el .IP "\f(CWsig_atomic_t volatile\fR must be thread-atomic as well" 4
3921.IX Item "sig_atomic_t volatile must be thread-atomic as well"
3922The type \f(CW\*(C`sig_atomic_t volatile\*(C'\fR (or whatever is defined as
3923\&\f(CW\*(C`EV_ATOMIC_T\*(C'\fR) must be atomic with respect to accesses from different
3924threads. This is not part of the specification for \f(CW\*(C`sig_atomic_t\*(C'\fR, but is
3925believed to be sufficiently portable.
3926.ie n .IP """sigprocmask"" must work in a threaded environment" 4
3927.el .IP "\f(CWsigprocmask\fR must work in a threaded environment" 4
3928.IX Item "sigprocmask must work in a threaded environment"
3929Libev uses \f(CW\*(C`sigprocmask\*(C'\fR to temporarily block signals. This is not
3930allowed in a threaded program (\f(CW\*(C`pthread_sigmask\*(C'\fR has to be used). Typical
3931pthread implementations will either allow \f(CW\*(C`sigprocmask\*(C'\fR in the \*(L"main
3932thread\*(R" or will block signals process-wide, both behaviours would
3933be compatible with libev. Interaction between \f(CW\*(C`sigprocmask\*(C'\fR and
3934\&\f(CW\*(C`pthread_sigmask\*(C'\fR could complicate things, however.
3935.Sp
3936The most portable way to handle signals is to block signals in all threads
3937except the initial one, and run the default loop in the initial thread as
3938well.
3939.ie n .IP """long"" must be large enough for common memory allocation sizes" 4
3940.el .IP "\f(CWlong\fR must be large enough for common memory allocation sizes" 4
3941.IX Item "long must be large enough for common memory allocation sizes"
3942To improve portability and simplify its \s-1API\s0, libev uses \f(CW\*(C`long\*(C'\fR internally
3943instead of \f(CW\*(C`size_t\*(C'\fR when allocating its data structures. On non-POSIX
3944systems (Microsoft...) this might be unexpectedly low, but is still at
3945least 31 bits everywhere, which is enough for hundreds of millions of
3946watchers.
3947.ie n .IP """double"" must hold a time value in seconds with enough accuracy" 4
3948.el .IP "\f(CWdouble\fR must hold a time value in seconds with enough accuracy" 4
3949.IX Item "double must hold a time value in seconds with enough accuracy"
3950The type \f(CW\*(C`double\*(C'\fR is used to represent timestamps. It is required to
3951have at least 51 bits of mantissa (and 9 bits of exponent), which is good
3952enough for at least into the year 4000. This requirement is fulfilled by
3953implementations implementing \s-1IEEE\s0 754 (basically all existing ones).
3954.PP
3955If you know of other additional requirements drop me a note.
3956.SH "ALGORITHMIC COMPLEXITIES"
3957.IX Header "ALGORITHMIC COMPLEXITIES"
3958In this section the complexities of (many of) the algorithms used inside
3959libev will be documented. For complexity discussions about backends see
3960the documentation for \f(CW\*(C`ev_default_init\*(C'\fR.
3961.PP
3962All of the following are about amortised time: If an array needs to be
3963extended, libev needs to realloc and move the whole array, but this
3964happens asymptotically rarer with higher number of elements, so O(1) might
3965mean that libev does a lengthy realloc operation in rare cases, but on
3966average it is much faster and asymptotically approaches constant time.
3967.IP "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)" 4
3968.IX Item "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)"
3969This means that, when you have a watcher that triggers in one hour and
3970there are 100 watchers that would trigger before that, then inserting will
3971have to skip roughly seven (\f(CW\*(C`ld 100\*(C'\fR) of these watchers.
3972.IP "Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)" 4
3973.IX Item "Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)"
3974That means that changing a timer costs less than removing/adding them,
3975as only the relative motion in the event queue has to be paid for.
3976.IP "Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)" 4
3977.IX Item "Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)"
3978These just add the watcher into an array or at the head of a list.
3979.IP "Stopping check/prepare/idle/fork/async watchers: O(1)" 4
3980.IX Item "Stopping check/prepare/idle/fork/async watchers: O(1)"
3981.PD 0
3982.IP "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % \s-1EV_PID_HASHSIZE\s0))" 4
3983.IX Item "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))"
3984.PD
3985These watchers are stored in lists, so they need to be walked to find the
3986correct watcher to remove. The lists are usually short (you don't usually
3987have many watchers waiting for the same fd or signal: one is typical, two
3988is rare).
3989.IP "Finding the next timer in each loop iteration: O(1)" 4
3990.IX Item "Finding the next timer in each loop iteration: O(1)"
3991By virtue of using a binary or 4\-heap, the next timer is always found at a
3992fixed position in the storage array.
3993.IP "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)" 4
3994.IX Item "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)"
3995A change means an I/O watcher gets started or stopped, which requires
3996libev to recalculate its status (and possibly tell the kernel, depending
3997on backend and whether \f(CW\*(C`ev_io_set\*(C'\fR was used).
3998.IP "Activating one watcher (putting it into the pending state): O(1)" 4
3999.IX Item "Activating one watcher (putting it into the pending state): O(1)"
4000.PD 0
4001.IP "Priority handling: O(number_of_priorities)" 4
4002.IX Item "Priority handling: O(number_of_priorities)"
4003.PD
4004Priorities are implemented by allocating some space for each
4005priority. When doing priority-based operations, libev usually has to
4006linearly search all the priorities, but starting/stopping and activating
4007watchers becomes O(1) with respect to priority handling.
4008.IP "Sending an ev_async: O(1)" 4
4009.IX Item "Sending an ev_async: O(1)"
4010.PD 0
4011.IP "Processing ev_async_send: O(number_of_async_watchers)" 4
4012.IX Item "Processing ev_async_send: O(number_of_async_watchers)"
4013.IP "Processing signals: O(max_signal_number)" 4
4014.IX Item "Processing signals: O(max_signal_number)"
4015.PD
4016Sending involves a system call \fIiff\fR there were no other \f(CW\*(C`ev_async_send\*(C'\fR
4017calls in the current loop iteration. Checking for async and signal events
4018involves iterating over all running async watchers or all signal numbers.
1818.SH "AUTHOR" 4019.SH "AUTHOR"
1819.IX Header "AUTHOR" 4020.IX Header "AUTHOR"
1820Marc Lehmann <libev@schmorp.de>. 4021Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.

Diff Legend

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