Re: [AD] more 4.1.15 WIP (was: [AD] get_audio_stream_buffer)

[Elias Pschernig <elias@xxxxxxxxxx>]

On Mon, 2004-07-26 at 14:12 +0200, Evert Glebbeek wrote:
> > Hm, I was just thinking - maybe it would be worth a try to try and use
> > the signals version? If the behavior is the same with it, then at least
> > we can be sure it doesn't have much to do with pthreads.
> 
> True. Considering the problems I keep seeing with the sigalrm version, I'm not
> sure this helps much though... ;)
> 

Hm, I wonder if we ever will discover what causes them. Some other race
condition probably.

> > I think not, since we don't even know yet what is causing it. But the
> > asm color converters are broken right now,
> 
> I don't have acces to the relevant messages at this computer, so I can't check,
> but what was the problem exactly? I seem to recall they worked ok the last time
> I tried them...?
> Worst case, could the problem be worked around by reversing an earlier patch? Or
> d we then need to reverse Henrik's patch? That at least should be in the next
> WIP in my opinion.
> 

Well, originally, they don't work with the signals version since they
write outside their stack frame. To fix that - we'd have to partially
reverse Henrik's patch - that is, disable it for the signals version.
Then, to fix the stack frame issue, a patch was applied which breaks
conversion with a 24-bit destination depth (everywhere, even in
windows :P). But Eric said he's trying to find a fix for it.

> > and the yield_timeslice
> > deprecation won't go in before tomorrow as it looks now.
> 
> That would be before the proposed timeframe for CVS freeze.
> 

Oh, I thought 27th..

> > I leave it up to you in any case when to do the release.
> 
> Ok :)
> Say, if I handle this one, will you handle the next WIP release?
> 

Heh, sure. But let's see how long it takes until we are ready for
4.1.16, and who will be around to do it by then :)

-- 
Elias Pschernig