Bug#797227: segfault - gst_memory_unmap, libgstreamer

[jnqnfe]

On Mon, 2015-08-31 at 12:39 +0300, Sebastian Dröge wrote:
> Hi,
> 
> On Sun, 30 Aug 2015 00:04:02 +0100 jnqnfe <jnqnfe at gmail.com> wrote:
> 
> > Upgrading to gstreamer1.0-plugins-bad from experimental as someone
> > suggested resulted in the following package changes:
> > 
> > The following NEW packages will be installed:
> > [...]
> > Iceweasel indeed no longer crashes now. Running in gdb I instead 
> > get 
> > a SIGPIPE failure.
> 
> Can you get a backtrace of that SIGPIPE? And without running in gdb,
> everything works fine for you now?

Yes having switched some of the gstreamer packages to 1.5.90 from
experimental, iceweasel no longer crashes during normal use.

The SIGPIPE failure doesn't seem to relate to gstreamer, so I'll stick
a bt for that in a separate bug report, if one doesn't already exist
for it.

> The vimeo link, https://vimeo.com/55640554, pasted in this bug report
> earlier here does not cause any crashes or anything for me, it just
> works fine.

Did you play the video or just load the page? I believe someone said
that you need to play it for the crash to occur.

I should perhaps explain that I did not try the supplied link; I have
iceweasel set to load my previous tabs on startup (of which I currently
have quiet a lot). After upgrading iceweasel along with gstreamer
-plugins-bad and dependencies, loading iceweasel then resulted in the
crash. I did not alter the set of tabs during testing. The crash
occured running iceweasel in safe mode in gdb. Switching gstreamer
-plugins-bad to v1.5.90 from experimental, loading iceweasel with that
same exact set of tabs as before, I had no issues. I checked in gdb
just for the hell of it and got the SIGPIPE failure (and still do), but
it's fine in normal operation.

> Can someone who is still able to reproduce this with 1.5.90 from
> experimental also install debug symbols for libc6, libglib2.0-0, all
> the GStreamer packages and then
>  a) run with valgrind --track-origins=yes
>  b) get a new backtrace with gdb?
> 
> Thanks!

I do not believe anyone has reported being able to reproduce the issue
with 1.5.90 packages.

Do you want me to revert to the troublesome packages and get a more
complete bt for you?