[mythtv] Re: [PATCH] mono support, take two

[Jim Radford]

On Tue, Nov 19, 2002 at 08:49:53PM -0500, Isaac Richards wrote:
> Is there really a need for the audio_ioctl_int stuff?  It's, well, ugly, IMO.

It is there, more or less, just to gather the debugging prints in one
location so I could turn them off at once, so that I didn't keep
having cut and paste errors :-), and so that the error checking was
uniform.  I left them in so you could see why the other code is
necessary (see below).  You can replace audio_ioctl_int with ioctl and
everything should work fine.

> There shouldn't be a need for detecting if the samplerate/channels changed in 
> the middle of the stream, either..

The information seems to only be available "in the middle of the
stream", and it only runs once as far as I can tell.  The code can be
moved out-of-line if you're worried about cache footprint.  The
problem is that lame, like oss, doesn't use what you tell it to.
Specifically when I tell it to do one channel and mono it doesn't do
it, it converts it to two channels.  So the myth header says 1 channel
(which it would be without compression) and the lame header says 2.
So first we set up one channel and then lame comes along and says, no
I really used two.

The overall architectural thing that could be fixed here would be to
figure out what the input/output/stream support before starting up and
then choosing the appropriate mode and possibly conversion routines.

-Jim