Re: [xine-devel] still image detection - now in CVS

[Christer P. <pa...@no...>]

Miguel Freitas wrote:

> 
> Christer, i believe you're doing a lot of "stability tests" with xine 
> and are catching bugs similar to the ones i saw some months ago when i 
> did the same. Maybe you could also try to test both cvs xine-lib and 
> stable branch against your streams to summarize any problems that may 
> have being introduced/fixed with latest metronom's and video_out 
> changes. I will do the same here, but more testing is always better...
> 


I see one problem with the latest CVS code that wasn't there before. 
Well, as a matter of fact it was, but it wasn't that serious :-)

The problem is with inserting 0-frames at the very beginning of the 
stream. This problem has been around for a long time, causing xine to 
miss the first 0.5 seconds or so of the stream - just enough to be a 
little annoying. But now I can suddenly get delays in the 10-30 second 
ballpark. When I press play, the seek knob also actually jumps a bit 
into the stream before the delay (again, this is an audio only, mp3 VBR 
file).

What's really weird is that I only see this problem after having left 
xine "idle" for several hours. It seems that it detoriates with 
wall-clock time, regardless of whether I have been playing back or not!

Unfortunately, I'm debugging the X hang problem right now so I can't 
provide any more details, but I wanted to report this anyway. Perhaps 
someone can still get an idea about what the problem could be?
Miguel appears to be a true wizard on fixing problems more or less 
blindly. :-)

-- 
Christer Palm