http://bugs.winehq.org/show_bug.cgi?id=28023
--- Comment #4 from Jörg Höhle hoehle@users.sourceforge.net 2012-01-15 16:26:39 CST --- The leak will be plugged by my next patch submission. (hmm, I just noticed that IAC_Initialize will not release all resources in all error paths.)
However, there are new issues. A deadlock in test_session. hr = IAudioClient_Stop(cap_ac); In my extended render tests, test_session follows test_clock(0)/*exclusive mode*/. The machine will deadlock both cores most of the time with wine-1.3.37, yet I've not managed to throw winegdb at it. When I instead moved test_clock(0) in front of test_audioclient(), the machine managed to run 141 iterations before receiving E_FAIL from IARC_ReleaseBuffer. I suspect OSSpinLock and queue management.