4.11

4 files changed, 319 insertions, 114 deletions

diff --git a/Changes b/Changes
index 5c3771a..bf3bb77 100644
--- a/Changes
+++ b/Changes
@@ -1,9 +1,9 @@
 Revision history for libev, a high-performance and full-featured event loop.
 
 TODO: ev_loop_wakeup
-
 TODO: EV_STANDALONE == NO_HASSEL (do not use clock_gettime in ev_standalone)
-4.11
+
+4.11 Sat Feb  4 19:52:39 CET 2012
 	- INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
           was documented already, but not implemented in the repeating case.
         - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
diff --git a/configure.ac b/configure.ac
index 1728e36..5ad67f2 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1,7 +1,7 @@
 AC_INIT
 AC_CONFIG_SRCDIR([ev_epoll.c])
 
-AM_INIT_AUTOMAKE(libev,4.09) dnl also update ev.h!
+AM_INIT_AUTOMAKE(libev,4.11) dnl also update ev.h!
 AC_CONFIG_HEADERS([config.h])
 AM_MAINTAINER_MODE
 
diff --git a/ev.3 b/ev.3
index f2a4514..2e3365e 100644
--- a/ev.3
+++ b/ev.3
@@ -1,4 +1,4 @@
-.\" Automatically generated by Pod::Man 2.22 (Pod::Simple 3.07)
+.\" Automatically generated by Pod::Man 2.23 (Pod::Simple 3.14)
 .\"
 .\" Standard preamble:
 .\" ========================================================================
@@ -124,7 +124,7 @@
 .\" ========================================================================
 .\"
 .IX Title "LIBEV 3"
-.TH LIBEV 3 "2011-02-16" "libev-4.04" "libev - high performance full featured event loop"
+.TH LIBEV 3 "2012-02-04" "libev-4.11" "libev - high performance full featured event loop"
 .\" For nroff, turn off justification.  Always turn off hyphenation; it makes
 .\" way too many mistakes in technical documents.
 .if n .ad l
@@ -246,7 +246,7 @@ loop mechanism itself (\f(CW\*(C`ev_idle\*(C'\fR, \f(CW\*(C`ev_embed\*(C'\fR, \f
 limited support for fork events (\f(CW\*(C`ev_fork\*(C'\fR).
 .PP
 It also is quite fast (see this
-<benchmark> comparing it to libevent
+benchmark <http://libev.schmorp.de/bench.html> comparing it to libevent
 for example).
 .SS "\s-1CONVENTIONS\s0"
 .IX Subsection "CONVENTIONS"
@@ -296,12 +296,18 @@ library in any way.
 Returns the current time as libev would use it. Please note that the
 \&\f(CW\*(C`ev_now\*(C'\fR function is usually faster and also often returns the timestamp
 you actually want to know. Also interesting is the combination of
-\&\f(CW\*(C`ev_update_now\*(C'\fR and \f(CW\*(C`ev_now\*(C'\fR.
+\&\f(CW\*(C`ev_now_update\*(C'\fR and \f(CW\*(C`ev_now\*(C'\fR.
 .IP "ev_sleep (ev_tstamp interval)" 4
 .IX Item "ev_sleep (ev_tstamp interval)"
-Sleep for the given interval: The current thread will be blocked until
-either it is interrupted or the given time interval has passed. Basically
-this is a sub-second-resolution \f(CW\*(C`sleep ()\*(C'\fR.
+Sleep for the given interval: The current thread will be blocked
+until either it is interrupted or the given time interval has
+passed (approximately \- it might return a bit earlier even if not
+interrupted). Returns immediately if \f(CW\*(C`interval <= 0\*(C'\fR.
+.Sp
+Basically this is a sub-second-resolution \f(CW\*(C`sleep ()\*(C'\fR.
+.Sp
+The range of the \f(CW\*(C`interval\*(C'\fR is limited \- libev only guarantees to work
+with sleep times of up to one day (\f(CW\*(C`interval <= 86400\*(C'\fR).
 .IP "int ev_version_major ()" 4
 .IX Item "int ev_version_major ()"
 .PD 0
@@ -555,7 +561,7 @@ example) that can't properly initialise their signal masks.
 .el .IP "\f(CWEVFLAG_NOSIGMASK\fR" 4
 .IX Item "EVFLAG_NOSIGMASK"
 When this flag is specified, then libev will avoid to modify the signal
-mask. Specifically, this means you ahve to make sure signals are unblocked
+mask. Specifically, this means you have to make sure signals are unblocked
 when you want to receive them.
 .Sp
 This behaviour is useful when you want to do your own signal handling, or
@@ -603,10 +609,10 @@ This backend maps \f(CW\*(C`EV_READ\*(C'\fR to \f(CW\*(C`POLLIN | POLLERR | POLL
 Use the linux-specific \fIepoll\fR\|(7) interface (for both pre\- and post\-2.6.9
 kernels).
 .Sp
-For few fds, this backend is a bit little slower than poll and select,
-but it scales phenomenally better. While poll and select usually scale
-like O(total_fds) where n is the total number of fds (or the highest fd),
-epoll scales either O(1) or O(active_fds).
+For few fds, this backend is a bit little slower than poll and select, but
+it scales phenomenally better. While poll and select usually scale like
+O(total_fds) where total_fds is the total number of fds (or the highest
+fd), epoll scales either O(1) or O(active_fds).
 .Sp
 The epoll mechanism deserves honorable mention as the most misdesigned
 of the more advanced event mechanisms: mere annoyances include silently
@@ -619,19 +625,22 @@ forks then \fIboth\fR parent and child process have to recreate the epoll
 set, which can take considerable time (one syscall per file descriptor)
 and is of course hard to detect.
 .Sp
-Epoll is also notoriously buggy \- embedding epoll fds \fIshould\fR work, but
-of course \fIdoesn't\fR, and epoll just loves to report events for totally
-\&\fIdifferent\fR file descriptors (even already closed ones, so one cannot
-even remove them from the set) than registered in the set (especially
-on \s-1SMP\s0 systems). Libev tries to counter these spurious notifications by
-employing an additional generation counter and comparing that against the
-events to filter out spurious ones, recreating the set when required. Last
+Epoll is also notoriously buggy \- embedding epoll fds \fIshould\fR work,
+but of course \fIdoesn't\fR, and epoll just loves to report events for
+totally \fIdifferent\fR file descriptors (even already closed ones, so
+one cannot even remove them from the set) than registered in the set
+(especially on \s-1SMP\s0 systems). Libev tries to counter these spurious
+notifications by employing an additional generation counter and comparing
+that against the events to filter out spurious ones, recreating the set
+when required. Epoll also erroneously rounds down timeouts, but gives you
+no way to know when and by how much, so sometimes you have to busy-wait
+because epoll returns immediately despite a nonzero timeout. And last
 not least, it also refuses to work with some file descriptors which work
 perfectly fine with \f(CW\*(C`select\*(C'\fR (files, many character devices...).
 .Sp
-Epoll is truly the train wreck analog among event poll mechanisms,
-a frankenpoll, cobbled together in a hurry, no thought to design or
-interaction with others.
+Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
+cobbled together in a hurry, no thought to design or interaction with
+others. Oh, the pain, will it ever stop...
 .Sp
 While stopping, setting and starting an I/O watcher in the same iteration
 will result in some caching, there is still a system call per such
@@ -719,11 +728,11 @@ hacks).
 .Sp
 On the negative side, the interface is \fIbizarre\fR \- so bizarre that
 even sun itself gets it wrong in their code examples: The event polling
-function sometimes returning events to the caller even though an error
+function sometimes returns events to the caller even though an error
 occurred, but with no indication whether it has done so or not (yes, it's
-even documented that way) \- deadly for edge-triggered interfaces where
-you absolutely have to know whether an event occurred or not because you
-have to re-arm the watcher.
+even documented that way) \- deadly for edge-triggered interfaces where you
+absolutely have to know whether an event occurred or not because you have
+to re-arm the watcher.
 .Sp
 Fortunately libev seems to be able to work around these idiocies.
 .Sp
@@ -944,7 +953,9 @@ with something not expressible using other libev watchers (i.e. "roll your
 own \f(CW\*(C`ev_run\*(C'\fR"). However, a pair of \f(CW\*(C`ev_prepare\*(C'\fR/\f(CW\*(C`ev_check\*(C'\fR watchers is
 usually a better approach for this kind of thing.
 .Sp
-Here are the gory details of what \f(CW\*(C`ev_run\*(C'\fR does:
+Here are the gory details of what \f(CW\*(C`ev_run\*(C'\fR does (this is for your
+understanding, not a guarantee that things will work exactly like this in
+future versions):
 .Sp
 .Vb 10
 \&   \- Increment loop depth.
@@ -1069,10 +1080,11 @@ overhead for the actual polling but can deliver many events at once.
 By setting a higher \fIio collect interval\fR you allow libev to spend more
 time collecting I/O events, so you can handle more events per iteration,
 at the cost of increasing latency. Timeouts (both \f(CW\*(C`ev_periodic\*(C'\fR and
-\&\f(CW\*(C`ev_timer\*(C'\fR) will be not affected. Setting this to a non-null value will
+\&\f(CW\*(C`ev_timer\*(C'\fR) will not be affected. Setting this to a non-null value will
 introduce an additional \f(CW\*(C`ev_sleep ()\*(C'\fR call into most loop iterations. The
 sleep time ensures that libev will not poll for I/O events more often then
-once per this interval, on average.
+once per this interval, on average (as long as the host time resolution is
+good enough).
 .Sp
 Likewise, by setting a higher \fItimeout collect interval\fR you allow libev
 to spend more time collecting timeouts, at the expense of increased
@@ -1134,7 +1146,7 @@ each call to a libev function.
 .Sp
 However, \f(CW\*(C`ev_run\*(C'\fR can run an indefinite time, so it is not feasible
 to wait for it to return. One way around this is to wake up the event
-loop via \f(CW\*(C`ev_break\*(C'\fR and \f(CW\*(C`av_async_send\*(C'\fR, another way is to set these
+loop via \f(CW\*(C`ev_break\*(C'\fR and \f(CW\*(C`ev_async_send\*(C'\fR, another way is to set these
 \&\fIrelease\fR and \fIacquire\fR callbacks on the loop.
 .Sp
 When set, then \f(CW\*(C`release\*(C'\fR will be called just before the thread is
@@ -1491,7 +1503,7 @@ transition between them will be described in more detail \- and while these
 rules might look complicated, they usually do \*(L"the right thing\*(R".
 .IP "initialiased" 4
 .IX Item "initialiased"
-Before a watcher can be registered with the event looop it has to be
+Before a watcher can be registered with the event loop it has to be
 initialised. This can be done with a call to \f(CW\*(C`ev_TYPE_init\*(C'\fR, or calls to
 \&\f(CW\*(C`ev_init\*(C'\fR followed by the watcher-specific \f(CW\*(C`ev_TYPE_set\*(C'\fR function.
 .Sp
@@ -1873,10 +1885,11 @@ monotonic clock option helps a lot here).
 .PP
 The callback is guaranteed to be invoked only \fIafter\fR its timeout has
 passed (not \fIat\fR, so on systems with very low-resolution clocks this
-might introduce a small delay). If multiple timers become ready during the
-same loop iteration then the ones with earlier time-out values are invoked
-before ones of the same priority with later time-out values (but this is
-no longer true when a callback calls \f(CW\*(C`ev_run\*(C'\fR recursively).
+might introduce a small delay, see \*(L"the special problem of being too
+early\*(R", below). If multiple timers become ready during the same loop
+iteration then the ones with earlier time-out values are invoked before
+ones of the same priority with later time-out values (but this is no
+longer true when a callback calls \f(CW\*(C`ev_run\*(C'\fR recursively).
 .PP
 \fIBe smart about timeouts\fR
 .IX Subsection "Be smart about timeouts"
@@ -1968,68 +1981,84 @@ In this case, it would be more efficient to leave the \f(CW\*(C`ev_timer\*(C'\fR
 but remember the time of last activity, and check for a real timeout only
 within the callback:
 .Sp
-.Vb 1
+.Vb 3
+\&   ev_tstamp timeout = 60.;
 \&   ev_tstamp last_activity; // time of last activity
+\&   ev_timer timer;
 \&
 \&   static void
 \&   callback (EV_P_ ev_timer *w, int revents)
 \&   {
-\&     ev_tstamp now     = ev_now (EV_A);
-\&     ev_tstamp timeout = last_activity + 60.;
+\&     // calculate when the timeout would happen
+\&     ev_tstamp after = last_activity \- ev_now (EV_A) + timeout;
 \&
-\&     // if last_activity + 60. is older than now, we did time out
-\&     if (timeout < now)
+\&     // if negative, it means we the timeout already occured
+\&     if (after < 0.)
 \&       {
 \&         // timeout occurred, take action
 \&       }
 \&     else
 \&       {
-\&         // callback was invoked, but there was some activity, re\-arm
-\&         // the watcher to fire in last_activity + 60, which is
-\&         // guaranteed to be in the future, so "again" is positive:
-\&         w\->repeat = timeout \- now;
-\&         ev_timer_again (EV_A_ w);
+\&         // callback was invoked, but there was some recent 
+\&         // activity. simply restart the timer to time out
+\&         // after "after" seconds, which is the earliest time
+\&         // the timeout can occur.
+\&         ev_timer_set (w, after, 0.);
+\&         ev_timer_start (EV_A_ w);
 \&       }
 \&   }
 .Ve
 .Sp
-To summarise the callback: first calculate the real timeout (defined
-as \*(L"60 seconds after the last activity\*(R"), then check if that time has
-been reached, which means something \fIdid\fR, in fact, time out. Otherwise
-the callback was invoked too early (\f(CW\*(C`timeout\*(C'\fR is in the future), so
-re-schedule the timer to fire at that future time, to see if maybe we have
-a timeout then.
+To summarise the callback: first calculate in how many seconds the
+timeout will occur (by calculating the absolute time when it would occur,
+\&\f(CW\*(C`last_activity + timeout\*(C'\fR, and subtracting the current time, \f(CW\*(C`ev_now
+(EV_A)\*(C'\fR from that).
 .Sp
-Note how \f(CW\*(C`ev_timer_again\*(C'\fR is used, taking advantage of the
-\&\f(CW\*(C`ev_timer_again\*(C'\fR optimisation when the timer is already running.
+If this value is negative, then we are already past the timeout, i.e. we
+timed out, and need to do whatever is needed in this case.
+.Sp
+Otherwise, we now the earliest time at which the timeout would trigger,
+and simply start the timer with this timeout value.
+.Sp
+In other words, each time the callback is invoked it will check whether
+the timeout cocured. If not, it will simply reschedule itself to check
+again at the earliest time it could time out. Rinse. Repeat.
 .Sp
 This scheme causes more callback invocations (about one every 60 seconds
 minus half the average time between activity), but virtually no calls to
 libev to change the timeout.
 .Sp
-To start the timer, simply initialise the watcher and set \f(CW\*(C`last_activity\*(C'\fR
-to the current time (meaning we just have some activity :), then call the
-callback, which will \*(L"do the right thing\*(R" and start the timer:
+To start the machinery, simply initialise the watcher and set
+\&\f(CW\*(C`last_activity\*(C'\fR to the current time (meaning there was some activity just
+now), then call the callback, which will \*(L"do the right thing\*(R" and start
+the timer:
 .Sp
 .Vb 3
-\&   ev_init (timer, callback);
-\&   last_activity = ev_now (loop);
-\&   callback (loop, timer, EV_TIMER);
+\&   last_activity = ev_now (EV_A);
+\&   ev_init (&timer, callback);
+\&   callback (EV_A_ &timer, 0);
 .Ve
 .Sp
-And when there is some activity, simply store the current time in
+When there is some activity, simply store the current time in
 \&\f(CW\*(C`last_activity\*(C'\fR, no libev calls at all:
 .Sp
-.Vb 1
-\&   last_activity = ev_now (loop);
+.Vb 2
+\&   if (activity detected)
+\&     last_activity = ev_now (EV_A);
+.Ve
+.Sp
+When your timeout value changes, then the timeout can be changed by simply
+providing a new value, stopping the timer and calling the callback, which
+will agaion do the right thing (for example, time out immediately :).
+.Sp
+.Vb 3
+\&   timeout = new_value;
+\&   ev_timer_stop (EV_A_ &timer);
+\&   callback (EV_A_ &timer, 0);
 .Ve
 .Sp
 This technique is slightly more complex, but in most cases where the
 time-out is unlikely to be triggered, much more efficient.
-.Sp
-Changing the timeout is trivial as well (if it isn't hard-coded in the
-callback :) \- just change the timeout and invoke the callback, which will
-fix things for you.
 .IP "4. Wee, just use a double-linked list for your timeouts." 4
 .IX Item "4. Wee, just use a double-linked list for your timeouts."
 If there is not one request, but many thousands (millions...), all
@@ -2063,11 +2092,49 @@ rather complicated, but extremely efficient, something that really pays
 off after the first million or so of active timers, i.e. it's usually
 overkill :)
 .PP
+\fIThe special problem of being too early\fR
+.IX Subsection "The special problem of being too early"
+.PP
+If you ask a timer to call your callback after three seconds, then
+you expect it to be invoked after three seconds \- but of course, this
+cannot be guaranteed to infinite precision. Less obviously, it cannot be
+guaranteed to any precision by libev \- imagine somebody suspending the
+process with a \s-1STOP\s0 signal for a few hours for example.
+.PP
+So, libev tries to invoke your callback as soon as possible \fIafter\fR the
+delay has occurred, but cannot guarantee this.
+.PP
+A less obvious failure mode is calling your callback too early: many event
+loops compare timestamps with a \*(L"elapsed delay >= requested delay\*(R", but
+this can cause your callback to be invoked much earlier than you would
+expect.
+.PP
+To see why, imagine a system with a clock that only offers full second
+resolution (think windows if you can't come up with a broken enough \s-1OS\s0
+yourself). If you schedule a one-second timer at the time 500.9, then the
+event loop will schedule your timeout to elapse at a system time of 500
+(500.9 truncated to the resolution) + 1, or 501.
+.PP
+If an event library looks at the timeout 0.1s later, it will see \*(L"501 >=
+501\*(R" and invoke the callback 0.1s after it was started, even though a
+one-second delay was requested \- this is being \*(L"too early\*(R", despite best
+intentions.
+.PP
+This is the reason why libev will never invoke the callback if the elapsed
+delay equals the requested delay, but only when the elapsed delay is
+larger than the requested delay. In the example above, libev would only invoke
+the callback at system time 502, or 1.1s after the timer was started.
+.PP
+So, while libev cannot guarantee that your callback will be invoked
+exactly when requested, it \fIcan\fR and \fIdoes\fR guarantee that the requested
+delay has actually elapsed, or in other words, it always errs on the \*(L"too
+late\*(R" side of things.
+.PP
 \fIThe special problem of time updates\fR
 .IX Subsection "The special problem of time updates"
 .PP
-Establishing the current time is a costly operation (it usually takes at
-least two system calls): \s-1EV\s0 therefore updates its idea of the current
+Establishing the current time is a costly operation (it usually takes
+at least one system call): \s-1EV\s0 therefore updates its idea of the current
 time only before and after \f(CW\*(C`ev_run\*(C'\fR collects new events, which causes a
 growing difference between \f(CW\*(C`ev_now ()\*(C'\fR and \f(CW\*(C`ev_time ()\*(C'\fR when handling
 lots of events in one iteration.
@@ -2086,6 +2153,40 @@ If the event loop is suspended for a long time, you can also force an
 update of the time returned by \f(CW\*(C`ev_now ()\*(C'\fR by calling \f(CW\*(C`ev_now_update
 ()\*(C'\fR.
 .PP
+\fIThe special problem of unsynchronised clocks\fR
+.IX Subsection "The special problem of unsynchronised clocks"
+.PP
+Modern systems have a variety of clocks \- libev itself uses the normal
+\&\*(L"wall clock\*(R" clock and, if available, the monotonic clock (to avoid time
+jumps).
+.PP
+Neither of these clocks is synchronised with each other or any other clock
+on the system, so \f(CW\*(C`ev_time ()\*(C'\fR might return a considerably different time
+than \f(CW\*(C`gettimeofday ()\*(C'\fR or \f(CW\*(C`time ()\*(C'\fR. On a GNU/Linux system, for example,
+a call to \f(CW\*(C`gettimeofday\*(C'\fR might return a second count that is one higher
+than a directly following call to \f(CW\*(C`time\*(C'\fR.
+.PP
+The moral of this is to only compare libev-related timestamps with
+\&\f(CW\*(C`ev_time ()\*(C'\fR and \f(CW\*(C`ev_now ()\*(C'\fR, at least if you want better precision than
+a second or so.
+.PP
+One more problem arises due to this lack of synchronisation: if libev uses
+the system monotonic clock and you compare timestamps from \f(CW\*(C`ev_time\*(C'\fR
+or \f(CW\*(C`ev_now\*(C'\fR from when you started your timer and when your callback is
+invoked, you will find that sometimes the callback is a bit \*(L"early\*(R".
+.PP
+This is because \f(CW\*(C`ev_timer\*(C'\fRs work in real time, not wall clock time, so
+libev makes sure your callback is not invoked before the delay happened,
+\&\fImeasured according to the real time\fR, not the system clock.
+.PP
+If your timeouts are based on a physical timescale (e.g. \*(L"time out this
+connection after 100 seconds\*(R") then this shouldn't bother you as it is
+exactly the right behaviour.
+.PP
+If you want to compare wall clock/system timestamps to your timers, then
+you need to use \f(CW\*(C`ev_periodic\*(C'\fRs, as these are based on the wall clock
+time, where your comparisons will always generate correct results.
+.PP
 \fIThe special problems of suspended animation\fR
 .IX Subsection "The special problems of suspended animation"
 .PP
@@ -2138,18 +2239,28 @@ keep up with the timer (because it takes longer than those 10 seconds to
 do stuff) the timer will not fire more than once per event loop iteration.
 .IP "ev_timer_again (loop, ev_timer *)" 4
 .IX Item "ev_timer_again (loop, ev_timer *)"
-This will act as if the timer timed out and restart it again if it is
-repeating. The exact semantics are:
-.Sp
-If the timer is pending, its pending status is cleared.
-.Sp
-If the timer is started but non-repeating, stop it (as if it timed out).
+This will act as if the timer timed out, and restarts it again if it is
+repeating. It basically works like calling \f(CW\*(C`ev_timer_stop\*(C'\fR, updating the
+timeout to the \f(CW\*(C`repeat\*(C'\fR value and calling \f(CW\*(C`ev_timer_start\*(C'\fR.
 .Sp
-If the timer is repeating, either start it if necessary (with the
-\&\f(CW\*(C`repeat\*(C'\fR value), or reset the running timer to the \f(CW\*(C`repeat\*(C'\fR value.
+The exact semantics are as in the following rules, all of which will be
+applied to the watcher:
+.RS 4
+.IP "If the timer is pending, the pending status is always cleared." 4
+.IX Item "If the timer is pending, the pending status is always cleared."
+.PD 0
+.IP "If the timer is started but non-repeating, stop it (as if it timed out, without invoking it)." 4
+.IX Item "If the timer is started but non-repeating, stop it (as if it timed out, without invoking it)."
+.ie n .IP "If the timer is repeating, make the ""repeat"" value the new timeout and start the timer, if necessary." 4
+.el .IP "If the timer is repeating, make the \f(CWrepeat\fR value the new timeout and start the timer, if necessary." 4
+.IX Item "If the timer is repeating, make the repeat value the new timeout and start the timer, if necessary."
+.RE
+.RS 4
+.PD
 .Sp
 This sounds a bit complicated, see \*(L"Be smart about timeouts\*(R", above, for a
 usage example.
+.RE
 .IP "ev_tstamp ev_timer_remaining (loop, ev_timer *)" 4
 .IX Item "ev_tstamp ev_timer_remaining (loop, ev_timer *)"
 Returns the remaining time until a timer fires. If the timer is active,
@@ -2279,9 +2390,12 @@ Another way to think about it (for the mathematically inclined) is that
 \&\f(CW\*(C`ev_periodic\*(C'\fR will try to run the callback in this mode at the next possible
 time where \f(CW\*(C`time = offset (mod interval)\*(C'\fR, regardless of any time jumps.
 .Sp
-For numerical stability it is preferable that the \f(CW\*(C`offset\*(C'\fR value is near
-\&\f(CW\*(C`ev_now ()\*(C'\fR (the current time), but there is no range requirement for
-this value, and in fact is often specified as zero.
+The \f(CW\*(C`interval\*(C'\fR \fI\s-1MUST\s0\fR be positive, and for numerical stability, the
+interval value should be higher than \f(CW\*(C`1/8192\*(C'\fR (which is around 100
+microseconds) and \f(CW\*(C`offset\*(C'\fR should be higher than \f(CW0\fR and should have
+at most a similar magnitude as the current time (say, within a factor of
+ten). Typical values for offset are, in fact, \f(CW0\fR or something between
+\&\f(CW0\fR and \f(CW\*(C`interval\*(C'\fR, which is also the recommended range.
 .Sp
 Note also that there is an upper limit to how often a timer can fire (\s-1CPU\s0
 speed for example), so if \f(CW\*(C`interval\*(C'\fR is very small then timing stability
@@ -3333,9 +3447,6 @@ of \*(L"global async watchers\*(R" by using a watcher on an otherwise unused
 signal, and \f(CW\*(C`ev_feed_signal\*(C'\fR to signal this watcher from another thread,
 even without knowing which loop owns the signal.
 .PP
-Unlike \f(CW\*(C`ev_signal\*(C'\fR watchers, \f(CW\*(C`ev_async\*(C'\fR works with any event loop, not
-just the default loop.
-.PP
 \fIQueueing\fR
 .IX Subsection "Queueing"
 .PP
@@ -3439,13 +3550,16 @@ signal or similar contexts (see the discussion of \f(CW\*(C`EV_ATOMIC_T\*(C'\fR
 embedding section below on what exactly this means).
 .Sp
 Note that, as with other watchers in libev, multiple events might get
-compressed into a single callback invocation (another way to look at this
-is that \f(CW\*(C`ev_async\*(C'\fR watchers are level-triggered, set on \f(CW\*(C`ev_async_send\*(C'\fR,
-reset when the event loop detects that).
-.Sp
-This call incurs the overhead of a system call only once per event loop
-iteration, so while the overhead might be noticeable, it doesn't apply to
-repeated calls to \f(CW\*(C`ev_async_send\*(C'\fR for the same event loop.
+compressed into a single callback invocation (another way to look at
+this is that \f(CW\*(C`ev_async\*(C'\fR watchers are level-triggered: they are set on
+\&\f(CW\*(C`ev_async_send\*(C'\fR, reset when the event loop detects that).
+.Sp
+This call incurs the overhead of at most one extra system call per event
+loop iteration, if the event loop is blocked, and no syscall at all if
+the event loop (or your program) is processing events. That means that
+repeated calls are basically free (there is no need to avoid calls for
+performance reasons) and that the overhead becomes smaller (typically
+zero) under load.
 .IP "bool = ev_async_pending (ev_async *)" 4
 .IX Item "bool = ev_async_pending (ev_async *)"
 Returns a non-zero value when \f(CW\*(C`ev_async_send\*(C'\fR has been called on the
@@ -3503,7 +3617,7 @@ Example: wait up to ten seconds for data to appear on \s-1STDIN_FILENO\s0.
 .IP "ev_feed_fd_event (loop, int fd, int revents)" 4
 .IX Item "ev_feed_fd_event (loop, int fd, int revents)"
 Feed an event on the given fd, as if a file descriptor backend detected
-the given events it.
+the given events.
 .IP "ev_feed_signal_event (loop, int signum)" 4
 .IX Item "ev_feed_signal_event (loop, int signum)"
 Feed an event as if the given signal occurred. See also \f(CW\*(C`ev_feed_signal\*(C'\fR,
@@ -3587,6 +3701,49 @@ real programmers):
 \&       (((char *)w) \- offsetof (struct my_biggy, t2));
 \&   }
 .Ve
+.SS "\s-1AVOIDING\s0 \s-1FINISHING\s0 \s-1BEFORE\s0 \s-1RETURNING\s0"
+.IX Subsection "AVOIDING FINISHING BEFORE RETURNING"
+Often you have structures like this in event-based programs:
+.PP
+.Vb 4
+\&  callback ()
+\&  {
+\&    free (request);
+\&  }
+\&
+\&  request = start_new_request (..., callback);
+.Ve
+.PP
+The intent is to start some \*(L"lengthy\*(R" operation. The \f(CW\*(C`request\*(C'\fR could be
+used to cancel the operation, or do other things with it.
+.PP
+It's not uncommon to have code paths in \f(CW\*(C`start_new_request\*(C'\fR that
+immediately invoke the callback, for example, to report errors. Or you add
+some caching layer that finds that it can skip the lengthy aspects of the
+operation and simply invoke the callback with the result.
+.PP
+The problem here is that this will happen \fIbefore\fR \f(CW\*(C`start_new_request\*(C'\fR
+has returned, so \f(CW\*(C`request\*(C'\fR is not set.
+.PP
+Even if you pass the request by some safer means to the callback, you
+might want to do something to the request after starting it, such as
+canceling it, which probably isn't working so well when the callback has
+already been invoked.
+.PP
+A common way around all these issues is to make sure that
+\&\f(CW\*(C`start_new_request\*(C'\fR \fIalways\fR returns before the callback is invoked. If
+\&\f(CW\*(C`start_new_request\*(C'\fR immediately knows the result, it can artificially
+delay invoking the callback by e.g. using a \f(CW\*(C`prepare\*(C'\fR or \f(CW\*(C`idle\*(C'\fR watcher
+for example, or more sneakily, by reusing an existing (stopped) watcher
+and pushing it into the pending queue:
+.PP
+.Vb 2
+\&   ev_set_cb (watcher, callback);
+\&   ev_feed_event (EV_A_ watcher, 0);
+.Ve
+.PP
+This way, \f(CW\*(C`start_new_request\*(C'\fR can safely return before the callback is
+invoked, while not delaying callback invocation too much.
 .SS "\s-1MODEL/NESTED\s0 \s-1EVENT\s0 \s-1LOOP\s0 \s-1INVOCATIONS\s0 \s-1AND\s0 \s-1EXIT\s0 \s-1CONDITIONS\s0"
 .IX Subsection "MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS"
 Often (especially in \s-1GUI\s0 toolkits) there are places where you have
@@ -3610,7 +3767,7 @@ triggered, using \f(CW\*(C`EVRUN_ONCE\*(C'\fR:
 \&   while (!exit_main_loop)
 \&     ev_run (EV_DEFAULT_ EVRUN_ONCE);
 \&
-\&   // in a model watcher
+\&   // in a modal watcher
 \&   int exit_nested_loop = 0;
 \&
 \&   while (!exit_nested_loop)
@@ -3819,7 +3976,7 @@ called):
 .PP
 That basically suspends the coroutine inside \f(CW\*(C`wait_for_event\*(C'\fR and
 continues the libev coroutine, which, when appropriate, switches back to
-this or any other coroutine. I am sure if you sue this your own :)
+this or any other coroutine.
 .PP
 You can do similar tricks if you have, say, threads with an event queue \-
 instead of storing a coroutine, you store the queue object and instead of
@@ -3917,7 +4074,7 @@ Aliases to the same types/functions as with the \f(CW\*(C`ev_\*(C'\fR prefix.
 For each \f(CW\*(C`ev_TYPE\*(C'\fR watcher in \fIev.h\fR there is a corresponding class of
 the same name in the \f(CW\*(C`ev\*(C'\fR namespace, with the exception of \f(CW\*(C`ev_signal\*(C'\fR
 which is called \f(CW\*(C`ev::sig\*(C'\fR to avoid clashes with the \f(CW\*(C`signal\*(C'\fR macro
-defines by many implementations.
+defined by many implementations.
 .Sp
 All of those classes have these methods:
 .RS 4
@@ -4058,7 +4215,7 @@ watchers in the constructor.
 \&   class myclass
 \&   {
 \&     ev::io   io  ; void io_cb   (ev::io   &w, int revents);
-\&     ev::io2  io2 ; void io2_cb  (ev::io   &w, int revents);
+\&     ev::io   io2 ; void io2_cb  (ev::io   &w, int revents);
 \&     ev::idle idle; void idle_cb (ev::idle &w, int revents);
 \&
 \&     myclass (int fd)
@@ -4107,20 +4264,20 @@ makes rev work even on mingw.
 .IP "Haskell" 4
 .IX Item "Haskell"
 A haskell binding to libev is available at
-<http://hackage.haskell.org/cgi\-bin/hackage\-scripts/package/hlibev>.
+http://hackage.haskell.org/cgi\-bin/hackage\-scripts/package/hlibev <http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
 .IP "D" 4
 .IX Item "D"
 Leandro Lucarella has written a D language binding (\fIev.d\fR) for libev, to
-be found at <http://proj.llucax.com.ar/wiki/evd>.
+be found at <http://www.llucax.com.ar/proj/ev.d/index.html>.
 .IP "Ocaml" 4
 .IX Item "Ocaml"
 Erkki Seppala has written Ocaml bindings for libev, to be found at
-<http://modeemi.cs.tut.fi/~flux/software/ocaml\-ev/>.
+http://modeemi.cs.tut.fi/~flux/software/ocaml\-ev/ <http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
 .IP "Lua" 4
 .IX Item "Lua"
 Brian Maher has written a partial interface to libev for lua (at the
 time of this writing, only \f(CW\*(C`ev_io\*(C'\fR and \f(CW\*(C`ev_timer\*(C'\fR), to be found at
-<http://github.com/brimworks/lua\-ev>.
+http://github.com/brimworks/lua\-ev <http://github.com/brimworks/lua-ev>.
 .SH "MACRO MAGIC"
 .IX Header "MACRO MAGIC"
 Libev can be compiled with a variety of options, the most fundamental
@@ -4165,7 +4322,11 @@ suitable for use with \f(CW\*(C`EV_A\*(C'\fR.
 .el .IP "\f(CWEV_DEFAULT\fR, \f(CWEV_DEFAULT_\fR" 4
 .IX Item "EV_DEFAULT, EV_DEFAULT_"
 Similar to the other two macros, this gives you the value of the default
-loop, if multiple loops are supported (\*(L"ev loop default\*(R").
+loop, if multiple loops are supported (\*(L"ev loop default\*(R"). The default loop
+will be initialised if it isn't already initialised.
+.Sp
+For non-multiplicity builds, these macros do nothing, so you always have
+to initialise the loop somewhere.
 .ie n .IP """EV_DEFAULT_UC"", ""EV_DEFAULT_UC_""" 4
 .el .IP "\f(CWEV_DEFAULT_UC\fR, \f(CWEV_DEFAULT_UC_\fR" 4
 .IX Item "EV_DEFAULT_UC, EV_DEFAULT_UC_"
@@ -4330,6 +4491,14 @@ supported). It will also not define any of the structs usually found in
 .Sp
 In standalone mode, libev will still try to automatically deduce the
 configuration, but has to be more conservative.
+.IP "\s-1EV_USE_FLOOR\s0" 4
+.IX Item "EV_USE_FLOOR"
+If defined to be \f(CW1\fR, libev will use the \f(CW\*(C`floor ()\*(C'\fR function for its
+periodic reschedule calculations, otherwise libev will fall back on a
+portable (slower) implementation. If you enable this, you usually have to
+link against libm or something equivalent. Enabling this when the \f(CW\*(C`floor\*(C'\fR
+function is not available will fail, so the safe default is to not enable
+this.
 .IP "\s-1EV_USE_MONOTONIC\s0" 4
 .IX Item "EV_USE_MONOTONIC"
 If defined to be \f(CW1\fR, libev will try to detect the availability of the
@@ -4451,16 +4620,30 @@ If defined to be \f(CW1\fR, libev will compile in support for the Linux inotify
 interface to speed up \f(CW\*(C`ev_stat\*(C'\fR watchers. Its actual availability will
 be detected at runtime. If undefined, it will be enabled if the headers
 indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
+.IP "\s-1EV_NO_SMP\s0" 4
+.IX Item "EV_NO_SMP"
+If defined to be \f(CW1\fR, libev will assume that memory is always coherent
+between threads, that is, threads can be used, but threads never run on
+different cpus (or different cpu cores). This reduces dependencies
+and makes libev faster.
+.IP "\s-1EV_NO_THREADS\s0" 4
+.IX Item "EV_NO_THREADS"
+If defined to be \f(CW1\fR, libev will assume that it will never be called
+from different threads, which is a stronger assumption than \f(CW\*(C`EV_NO_SMP\*(C'\fR,
+above. This reduces dependencies and makes libev faster.
 .IP "\s-1EV_ATOMIC_T\s0" 4
 .IX Item "EV_ATOMIC_T"
 Libev requires an integer type (suitable for storing \f(CW0\fR or \f(CW1\fR) whose
-access is atomic with respect to other threads or signal contexts. No such
-type is easily found in the C language, so you can provide your own type
-that you know is safe for your purposes. It is used both for signal handler \*(L"locking\*(R"
-as well as for signal and thread safety in \f(CW\*(C`ev_async\*(C'\fR watchers.
+access is atomic and serialised with respect to other threads or signal
+contexts. No such type is easily found in the C language, so you can
+provide your own type that you know is safe for your purposes. It is used
+both for signal handler \*(L"locking\*(R" as well as for signal and thread safety
+in \f(CW\*(C`ev_async\*(C'\fR watchers.
 .Sp
 In the absence of this define, libev will use \f(CW\*(C`sig_atomic_t volatile\*(C'\fR
-(from \fIsignal.h\fR), which is usually good enough on most platforms.
+(from \fIsignal.h\fR), which is usually good enough on most platforms,
+although strictly speaking using a type that also implies a memory fence
+is required.
 .IP "\s-1EV_H\s0 (h)" 4
 .IX Item "EV_H (h)"
 The name of the \fIev.h\fR header file used to include it. The default if
@@ -4488,6 +4671,10 @@ will have the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument, and you can
 additional independent event loops. Otherwise there will be no support
 for multiple event loops and there is no first event loop pointer
 argument. Instead, all functions act on the single default loop.
+.Sp
+Note that \f(CW\*(C`EV_DEFAULT\*(C'\fR and \f(CW\*(C`EV_DEFAULT_\*(C'\fR will no longer provide a
+default loop when multiplicity is switched off \- you always have to
+initialise the loop manually in this case.
 .IP "\s-1EV_MINPRI\s0" 4
 .IX Item "EV_MINPRI"
 .PD 0
@@ -4594,6 +4781,19 @@ when you use \f(CW\*(C`\-Wl,\-\-gc\-sections \-ffunction\-sections\*(C'\fR) func
 your program might be left out as well \- a binary starting a timer and an
 I/O watcher then might come out at only 5Kb.
 .RE
+.IP "\s-1EV_API_STATIC\s0" 4
+.IX Item "EV_API_STATIC"
+If this symbol is defined (by default it is not), then all identifiers
+will have static linkage. This means that libev will not export any
+identifiers, and you cannot link against libev anymore. This can be useful
+when you embed libev, only want to use libev functions in a single file,
+and do not want its identifiers to be visible.
+.Sp
+To use this, define \f(CW\*(C`EV_API_STATIC\*(C'\fR and include \fIev.c\fR in the file that
+wants to use libev.
+.Sp
+This option only works when libev is compiled with a C compiler, as \*(C+
+doesn't support the required declaration syntax.
 .IP "\s-1EV_AVOID_STDIO\s0" 4
 .IX Item "EV_AVOID_STDIO"
 If this is set to \f(CW1\fR at compiletime, then libev will avoid using stdio
@@ -4980,7 +5180,7 @@ model. Libev still offers limited functionality on this platform in
 the form of the \f(CW\*(C`EVBACKEND_SELECT\*(C'\fR backend, and only supports socket
 descriptors. This only applies when using Win32 natively, not when using
 e.g. cygwin. Actually, it only applies to the microsofts own compilers,
-as every compielr comes with a slightly differently broken/incompatible
+as every compiler comes with a slightly differently broken/incompatible
 environment.
 .PP
 Lifting these limitations would basically require the full
@@ -5126,8 +5326,12 @@ The type \f(CW\*(C`double\*(C'\fR is used to represent timestamps. It is require
 have at least 51 bits of mantissa (and 9 bits of exponent), which is
 good enough for at least into the year 4000 with millisecond accuracy
 (the design goal for libev). This requirement is overfulfilled by
-implementations using \s-1IEEE\s0 754, which is basically all existing ones. With
-\&\s-1IEEE\s0 754 doubles, you get microsecond accuracy until at least 2200.
+implementations using \s-1IEEE\s0 754, which is basically all existing ones.
+.Sp
+With \s-1IEEE\s0 754 doubles, you get microsecond accuracy until at least the
+year 2255 (and millisecond accuracy till the year 287396 \- by then, libev
+is either obsolete or somebody patched it to use \f(CW\*(C`long double\*(C'\fR or
+something like that, just kidding).
 .PP
 If you know of other additional requirements drop me a note.
 .SH "ALGORITHMIC COMPLEXITIES"
@@ -5191,8 +5395,9 @@ watchers becomes O(1) with respect to priority handling.
 .IX Item "Processing signals: O(max_signal_number)"
 .PD
 Sending involves a system call \fIiff\fR there were no other \f(CW\*(C`ev_async_send\*(C'\fR
-calls in the current loop iteration. Checking for async and signal events
-involves iterating over all running async watchers or all signal numbers.
+calls in the current loop iteration and the loop is currently
+blocked. Checking for async and signal events involves iterating over all
+running async watchers or all signal numbers.
 .SH "PORTING FROM LIBEV 3.X TO 4.X"
 .IX Header "PORTING FROM LIBEV 3.X TO 4.X"
 The major version 4 introduced some incompatible changes to the \s-1API\s0.
diff --git a/ev.h b/ev.h
index 72587d5..54b4f00 100644
--- a/ev.h
+++ b/ev.h
@@ -1,7 +1,7 @@
 /*
  * libev native API header
  *
- * Copyright (c) 2007,2008,2009,2010,2011 Marc Alexander Lehmann <libev@schmorp.de>
+ * Copyright (c) 2007,2008,2009,2010,2011,2012 Marc Alexander Lehmann <libev@schmorp.de>
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without modifica-
@@ -199,7 +199,7 @@ struct ev_loop;
 /*****************************************************************************/
 
 #define EV_VERSION_MAJOR 4
-#define EV_VERSION_MINOR 4
+#define EV_VERSION_MINOR 11
 
 /* eventmask, revents, events... */
 enum {