[AD] Timer accuracy

["Eric Botcazou" <ebotcazou@xxxxxxxxxx>]

The Windows timer driver has used its own code in the 4.0.x series to handle
the ticks, a big part of which was merely duplicated from src/timer.c; as
this situation is bug-prone (Ben Davis fixed a discrepancy between the codes
some months ago), I made the Windows timer driver reuse the code in
src/timer.c for the 4.1.x series.

In the process I lifted a limitation on the accuracy of timers for the
Windows port. This modification has just bitten me back:
http://sourceforge.net/tracker/index.php?func=detail&aid=606246&group_id=5665&atid=105665

The fix is simply to re-enable the limitation:

--- /home/Eric/cvsroot/allegro/src/timer.c Wed May 15 08:39:02 2002
+++ allegro/src/timer.c Thu Sep 12 14:28:40 2002
@@ -102,6 +102,12 @@
    timer_delay -= d;
    timer_semaphore = FALSE;

+#ifdef ALLEGRO_WINDOWS
+   /* fudge factor to prevent interrupts from coming too close to each other */
+   if (new_delay < MSEC_TO_TIMER(1))
+      new_delay = MSEC_TO_TIMER(1);
+#endif
+
    return new_delay;
 }


Should we enable the limitation for the Windows port only or on a
platform-independent basis ?

- Eric