http://bugs.winehq.org/show_bug.cgi?id=28723
Jörg Höhle hoehle@users.sourceforge.net changed:
What |Removed |Added ---------------------------------------------------------------------------- Severity|minor |normal
--- Comment #96 from Jörg Höhle hoehle@users.sourceforge.net 2011-12-16 11:35:26 CST --- I once wrote: Ever since I've seen EAGAIN returned by ALSA when trying to write into an almost full buffer, I'm convinced that the audio devices maintain a safety margin between the write and read positions and will *not* let you write as much as the upper limit buffer size - written frames says.
That might explain the underruns with a max buffer 3x period_size. Perhaps 4 times would be better. I'll try and analyse logs over the week-end.
Could you try set_buffer_min(period) // not sure 3x should be the minimum, try 2x? set buffer_near(3xperiod)? set buffer_near(4xperiod)?
I don't know how to test for the improved PA behavior
I'm simply running my render.c tests and looking at GCP and GP values.