On Wed, 2012-05-09 at 20:43 +0530, Arun Raghavan wrote: > [I don't expect major objections to this, but since we're in the midst of a > freeze, playing it safe] > > The extended stream creation API introduced subtle breakage at the time > of sink input creation. The sample spec (and thus channel count) is no > longer guaranteed to be available during the PA_SINK_INPUT_NEW hook > (since multiple formats may be specified and a single format might only > be fixed after all hooks are called). > > To deal with this, instead of asserting on mismatched channel maps in > whatever volume is set by a hook, we remap the set volume to the desired > number of channels. > --- Turns out this is all a big lie. The volume setting happens in the fixate hook. I'm going to leave this out for now. I think in the long term we'll need to have a cleaner and better-documented explanation of what happens in what hook for module-writers, but for now, things work fine. -- Arun
