Open git-bruh opened 9 months ago
I haven't verified this, but do you think there's a possibility that the calls to alsa_set_stream_state
in alsa_stream_stop
and alsa_run
could be "racing" due to interleaved calls to mutex locks and unlocks?
So even tho the calls are wrapped in mutexes, it might happen that alsa_stream_stop
sets the state to INACTIVE
, and it subsequently gets set to PROCESSING
by the other thread after that thread locks the mutex
Hello, I've discovered a bug where the stream's data callback gets called even after
alsa_stream_stop
has been called. This in-turn causes a crash in Firefox when it tries to capture the microphone from it's WebRTC component (happens very rarely). I've traced it down to cubeb. This does not seem to be reproducible with the example program from the docsStill debugging it, will keep updating here with more details
More details: https://bugzilla.mozilla.org/show_bug.cgi?id=1859796