http://bugs.winehq.org/show_bug.cgi?id=28723
--- Comment #114 from Jörg Höhle hoehle@users.sourceforge.net 2011-12-22 17:50:37 CST ---
#5 does not speed up sound, but sounds like there is blank space added in the sound playback for just a few ms.
This is desireable behaviour. When an underrun happens, there's nothing better to do but to try and have a blank as short as possible. I'll submit this patch too, even if it's a PA quirk.
Whether an underrun is avoidable is another story.
I'm somewhat frustrated. I've seen random xruns with Ubuntu Lucid booted, running nothing but the render tests (with PA...). Given my comment #51 and comment #61, I now have doubts whether Wine and/or the Linux OS can support the 10ms periods with the current guaranteed underrun when 30-40ms late. Look at the mmdevapi and the damn' 10ms, how many server round-trips, context and thread switches are involved in event signaling and sending data from the app -> dsound -> mmdevapi -> ALSA every 10ms? Do we require a dual core do produce sound nowadays?
BTW, I've attached to bug #27937 my current render & GetPosition tests. I welcome e-mails with logs from them on both native an UNIX machines.
all of the patches for this bug are in Wine now
If Alexey confirms Rage is working fine, this bug could be marked fixed. OTOH I still believe that the current code is too sensitive to underruns due to a missing lead-in in the Rage use-case, and the present bug is the one to discuss this. PA quirks should be discussed elsewhere, even though they are hard to separate from a user POV.