Re: [Plib-users] Patch for plib and alsa oss emulation

[Felix <fx...@gm...>]

Hmm, looks like I should clarify this ...

On Sat, 27 Jul 2002 13:17:15 -0500
Steve Baker <sjb...@ai...> wrote:

> Felix Kühling wrote:
> 
> > The cause turned out to be that plib writes more data to the
> > (non-blocking) sound device with the default safety_margin than it can
> > take. The attached patch fixes this problem for linux, both Tuxkart and
> > Flightgear work fine :) and it doesn't break Torcs which was ok before.
> 
> But PLIB asks the OSS API how many bytes are free in the input buffer and
> always writes that number.  I think there must be something wrong with
> the slDSP::secondsRemaining function for this device.

Maybe we are talking about different versions, I was talking about the
latest stable version 1.4.2. In slScheduler::realUpdate audio data is
written in fragments using slDSP::play. The fragment size was determined
in slScheduler::initBuffers. However, realUpdate keeps writing to the
audio device while ( secondsUsed() <= safety_margin ).
slScheduler::safety_margin is not related to the buffer memory available
in the audio device. It is initialized to 1 second by default. Torcs
which sets this value to 0.128 works fine.

BTW, slDSP::secondsRemaining is never called in slDSP.cxx and
slScheduler.cxx.

> > However, it introduces a new function slDSP::fragmentsRemaining which
> > would have to be implemented (as dummy?) for other platforms.
> 
> But that's exactly what 'secondsRemaining' is doing (except that it's
> supposed to be implemented on every platform and it doesn't export the
> concept of 'fragments' which isn't meaningful at this level.

I tried some magic with secondsRemaining first but I never got it to
work reliably. Also this approach would require a new function like
secondsFragment in order to be sure that there is enough buffer memory
left for an entire fragment.

> 
> I don't think this is the correct patch for whatever your problem is.

I think the fact that other people don't have the problem (like me on my
other computer with a real OSS driver ;) is that there is usually enough
buffer available for the default safety margin of 1 sec.

The patch works for me. Maybe the problem does not exist in the latest
unstable version but this is not a solution for me since I don't want to
recompile all the programmes and I'm not even sure that they would work
with plib 1.5.x.

Regards,
    Felix Kühling

               __\|/__    ___     ___     ___
__Tschüß_______\_6 6_/___/__ \___/__ \___/___\___You can do anything,___
_____Felix_______\Ä/\ \_____\ \_____\ \______U___just not everything____
  fx...@gm...    >o<__/   \___/   \___/        at the same time!