http://bugs.winehq.org/show_bug.cgi?id=28856
--- Comment #20 from Jörg Höhle hoehle@users.sourceforge.net 2012-02-18 01:30:56 CST ---
I'm running Fedora 16 with pulseaudio 1.1 and alsa stack 1.0.25 Interesting! I restarted pulseaudio and ran the test again. This time there was no crackling
Please, please do submit a bug to pulseaudio.org. It is undeniably a bug in either PA or alsa_plugins when restarting the PA server fixes sound. I can't submit a bug report that would be taken seriously because I'm using the outdated Ubuntu Intrepid and Lucid, though I've been liasing with the PA guys: http://lists.freedesktop.org/archives/pulseaudio-discuss/2012-February/01283... But you're using the newest PA and plugins there is and still facing issues. Well, perhaps I should go to the PA bug tracker and post your logs?!?
Then hopefully, there will be less PA issues in the Wine bugtracker.
This PA bug number 2:
render.c:1204: hpctime 460 pcpos 466 render.c:1209: padding 0 position 22500/22500 slept 450ms iteration 0 render.c:1204: hpctime 584 pcpos 590 render.c:1209: padding 10980 position 34020/34020 slept 550ms iteration 1
Within 124ms, PA pretends that the speaker position as reported by snd_pcm_delay advanced by 249ms. That cannot be.
render.c:1217: Test failed: Position delta 12960 not regular
Same here within 103ms.
render.c:1209: padding 6720 position 152100/152100 slept 1250ms iteration 8
Here it pretends to be at 3.168 seconds within the stream after solely 1.327 seconds elapsed. That is PA's buffering, not a speaker position. Please attach the corresponding WINEDEBUG=+alsa,+tid,+timestamp log.
After the patch, each continuous tone emitted was heard completely continuous.
That's the one bug on the Wine side: 1. We need to perform tests to check whether CreateTimerQueueTimer's current behaviour is correct. As observed in bug #28723, comment #126 and #130, unlike timeSetEvent, it does not maintain a constant pace. Should it? 2. Depending on results, either fix TimerQueue or choose another time base.