Re: nbmix - curses mixer using audio(4)

[Patrick Mahoney <pat%polycrystal.org@localhost>]

On Tue, Feb 16, 2010 at 11:44:20PM +0100, Rhialto wrote:
> On Mon 15 Feb 2010 at 23:04:11 -0700, Patrick Mahoney wrote:
> > > nbmix - http://polycrystal.org/nbmix
> > 
> > I updated (v0.2) to fix the obvious segfault and to support
> > 'enum' controls, but now there are more untested code paths.
> 
> I've discovered one :-) Writing to the mixer can give a "device busy"
> error on my hardware for some settings (if audio is currently playing;
> with the azalia driver), in which case nbmix currently exists. If it
> just shows the error (and doesn't retry or exit) it would be fine.

Thanks for testing.  It's true, at the moment I treat all
errors equally and simply exit, which is not useful in your
case.

Question:

Neither ioctl(2) nor audio(4) document a possible error of
EBUSY (which is what I assume is the azalia error return;
azalia_codec.c:generic_mixer_set has a couple EBUSY possible
returns).

Is it expected that syscalls document all possible error
returns?  (Is a simple documentation patch for ioctl(2) in
order?).

Or, should an EBUSY return be documented in audio(4) and not
ioctl(2)?

Or, should one expect that any syscall might return any
valid errno?

Thanks.

-- 
Patrick Mahoney <pat%polycrystal.org@localhost>