A fractional-second timeout was already supported for the initial timeout of a /TIMER, but
if the timer was recurring then the interval was rounded to a whole number of seconds for
the subsequent timeouts.
Change the type of interval from long to double to fix this.
An absolute wake time is now calculated and updated by set_server_bits(),
TimerTimeout() and tclTimerTimeout(). This is then used to calculate the actual
select() timeout, instead of having those functions calculate a relative timeout
value themselves. This significantly simplifies those functions, since the
underlying values tend to be recorded as absolute times.
It also allows us to disentangle tclTimeTimeout from TimerTimeout(), and move the
responsibility for calculating the QUEUE_SENDS wake time from TimerTimeout() to
set_server_bits() where it belongs.
The QUEUE_SENDS and CONNECT_DELAY wake up times will also now be correctly
calculated, based on the last sent time and connect time respectively.
This is preferable to using time_diff() to compare two timevals, because
time_diff() gives its result as a 'double' which only has 53 bits of
precision.
Also switch a couple of places from using time_diff() to using
this function.
If we immediately handle timers that have already expired, there doesn't
seem to be a compelling reason to add 100ms to the expiry of unexpired timers.
This function never worked - function_timer() was creating a string but then
never throwing it away and returning a part of the argument instead, which
then caused a crash because it couldn't be freed later in the expression
parsing.
git-svn-id: svn://svn.code.sf.net/p/bitchx/code/trunk@472 13b04d17-f746-0410-82c6-800466cd88b0
