[gst-devel] valgrind reports
Vincent.Torri at iecn.u-nancy.fr
Sun Apr 17 00:24:06 CEST 2005
I'm writing a program that tries get the frames of a video. First, i need
to know the informations of the streams. So, I basically create a pipeline
with a file source and a decodebin elements, i iterate and i wait for the
pads to be created.
My test file is an .ogm one with one xvid video stream and 2 vorbis
I have installed gstreamer, gst-plugins and gst-ffmpeg cvs, branch 0.8
If I just iterate the pipeline with no pad detection, there's no crash
If I iterate the pipeline with the pad detection, it crashes always at the
same time (28 iterations), with apparently a call of malloc(-1)
A better description of what i do :
1) i iterate the pipeline to find all the streams and i set the pipeline
to READY. I don't build the whole pipeline. As parameters, I have an audio
and video stream index. If these parameters can't correspond to a stream,
they are set to 0 (that is, the first streams).
2) I iterate again to build the pipeline according to these parameters (i
check if the streams that i find correspond to the parameters).
The program crashes after 28 iterations in the 2nd step.
so i've tested on the video file with a simple program that i've done
(gspot like, called esi) and which doesn't crash. But when I run valgrind
on this program, there's a lot of messages.
valgrind logfile :
program that i use :
part of the program that crashes :
Also note that there is a memory leak with xml (line 1458 of the logfile).
xmlCleanupParser() should be called when xml is not used anymore.
My questions are:
1) Are the problems reported by valgrind known ?
2) Do I do something wrong when I try to get the informations (in esi, or
in the program that crashes) ? (for esi, see src/lib/esi.c and
More information about the gstreamer-devel