[gst-devel] Problems with schedulers and other assorted nonsense
iain at prettypeople.org
Wed Nov 13 06:46:02 CET 2002
More info that I've found on this.
It works in 0.4.2 and it seems to be issues in the cothread_create
routine. Also when I run the program to the halt and do a ps we find out
that the GstThread is a zombie, which would imply that it's crashed?
iain 15358 18.3 0.3 22972 5780 pts/0 S 14:44 0:00 lt-test-sample /h
iain 15371 0.0 0.0 0 0 pts/0 Z 14:44 0:00 [lt-test-sample <
The basic-omega.log.gz file ends while the code is inside the mmap call
in cothreads_create in cothreads.c (line 222?)
running it under gdb doesn't seem to work as gdb then complains about
not being about to connect to the crashed thread.
I'm running out of ideas.
On Sun, 2002-11-10 at 14:21, iain wrote:
> I've got this pipeline, looks like this all inside a GstThread
> filesrc ! spider ! oneton ! f2i ! mysink
> Works fine if I change spider to mad and make sure filesrc is an mp3.
> However when I run it as a spider it breaks. It used to work fine, but
> not anymore.
> I've done some of my own debugging and found that using the basicomega
> scheduler the oneton's connect function never receives fixed caps, and
> so never creates a new src pad, so the f2i is never connected etc...
> With the basicwingo scheduler it gets further, the oneton receives the
> fixed caps, creates two new pads (for a 2 channel audio stream), the f2i
> is connected to both, the mysink is connected as well, the pipeline is
> set to playing and .... nothing. It just sits there.
> I've put the debug and info logs at
> for your perusal.
> Any idea what changes might have been made to make this all stop
>  takes 1 N-channel audio/raw and outputs N 1-channel audio/raw
> In the diagram there are N f2i's which all request a sinkpad from
Television is the opiate of the masses.
More information about the gstreamer-devel