As I pointed out on SDL bugzilla
http://bugzilla.libsdl.org/show_bug.cgi?id=298
audio is not behaving correctly when the conversion of sound stream
require more samples on output then the one that are incoming. While
there is actually a bug on SDL, I wonder what should be the number of
samples agreed between user and SDL (and actually the driver), when the
agreed sound rate is not the one asked from the client:
e.g.
I ask for 11025 sample/sec and number of samples 512
If sound driver accept only a 48000 sample/sec, should SDL ask the
driver if it accepts 512 * 48000 / 11025 and report back the scaled
value to the application or what?