[Bug 28723] Sound stutter in Rage when emulated windows version is set to "Windows 7" (XAudio2 -> mmdevapi sound output path)

http://bugs.winehq.org/show_bug.cgi?id=28723

--- Comment #134 from Jörg Höhle  2012-02-27 04:49:17 CST ---
Initially I thought we need EVENTCALLBACK timing tests.  It turns out that the
current test_worst_case is good enough for reasoning on averages:
render.c:2080: Should play 1000ms continuous tone with fragment size 441.
render.c:2132: Released 44100=100x441 -441 frames at 44100 worth 990ms in
1006ms

- If the events were more than ~10ms apart, total time would be >> 1006ms,
  e.g. ~1200ms on my 12ms callback wakeup Intrepid system.

- If the events were < ~10ms apart, it would take less time or be
  unable to release data 100 times, e.g. 80 times on my Ubuntu system
  with 8ms wakeups, or leave more padding at the end.

Now is the rate 10ms or 10.1587?  I'm researching opportunities for allowing a
regular drain (e.g. decrease padding by one period each time) even in the
presence of irregular or non-matching callback intervals, given that we have a
few frames hidden within ALSA that hopefully will compensate that.  Ideas &
help welcome.