http://bugs.winehq.org/show_bug.cgi?id=25553
--- Comment #5 from Max TenEyck Woodbury max+winebugs@mtew.isa-geek.net 2011-11-02 09:02:15 CDT --- This (and similar problems) comes and goes from one commit set to the next. For any given commit set it is repeatable. It will sometimes persist for a few (on occasion many) subsequent commit sets. Then it will stop happening. In other words, the triggering bug gets fixed. The persistent problem is the hang part.
When the hangs happens, I have to get a separate command line, run 'top c' and set it to show 'zombies'. I can then kill the zombie and the test sequence will continue.
At this moment, the test suite is running to completion without intervention. That does NOT mean the hang problem has been fixed.
It only means that it is not being triggered, but there is no way to reliably trigger the failure. In other words, the bisection procedure in the regression process fails because many of the runs produce indeterminate results for the hang problem..
For the specific audio under-run problem originally mentioned, the problem has not occurred in quite some time. It has probably been fixed, but it has been fixed before and it or something very much like it has reappeared several times now.
Conclusion: The audio problem is not present at the moment. If you want me to isolate the old audio problems that have occurred in the past, let me know and I will run a sequence of regressions on the ones you mention. Otherwise, consider the problem as `fixed', at least for the moment.