- recognize all available audio devices
- the audio device is always opened read-write. this results in
full-duplex operation for most OSS implementations, but full-dupplex
needs to be explicitly enabled on OpenBSD.
- don't error out if SNDCTL_DSP_SPEED returns an error. instead
resample based on the returned sample rate.
- the resampling factor was not properly initilized and was returning
random values.
- simply skip samples to down-sample. the average method it uses
sounds horrible.
now ekiga's softphone works for me when using the OSS audio backend
ok ajacoutot
7e624c34a7