Alexandre,
You wrote:
>It doesn't work here:
Backtrace:
=>0 get_audio_session+0xbd() in winealsa
1 AudioClient_Initialize+0x5ae(iface=0x30c80, mode=AUDCLNT_SHAREMODE_SHARED,
flags=0, duration=<is not available>, period=0x186a0, fmt=0x30f00,
sessionguid=0x22fa00) [dlls/winealsa.drv/mmdevdrv.c:1075] in winealsa
2 test_session+0x114() [dlls/mmdevapi/tests/render.c:1298] in mmdevapi_test
Could you please investigate this?
The crash happens in test_session, before the one change to this
function by my patch.
The tests have been running thousands of iterations on my dual core
MacOS (with my wip lock-free patches) and a single core Linux
ALSA/dmix box.
Several people have run the tests on native, including 64bit MS-Windows.
I can't rule out that one of my latest additions may have introduced a
bug, but so far I'd rather believe that my renderer tests have
uncovered a bug in mmdevapi that shows up in your environment only. Why?
On my side, I think I'll add ok(ref=0) checks around every IAC_Release.
Thank you,
Jörg Höhle
