[Python-apps-team] Bug#575536: mayavi2 segfaults under VNC

[Gael Varoquaux]

On Sat, Apr 03, 2010 at 09:29:34PM -0400, Varun Hiremath wrote:
> Received this bug in Debian -- mayavi2 segfaults under VNC:
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=575536

> Do you have any idea what the problem could be? Have you ever tried
> running mayavi2 under VNC?

I don't have a terribly good idea of what the problem might be, and I
have not tried running mayavi2 under VNC. I can reproduce the problem
using vnc4server. Ryan Niebur did some pretty interesting debuging, and I
suspect that he is right that a sizer is called with a None parent.
However, to figure out where this is happening in the Python code, even
using the stack trace that Yaroslav attached, is non trivial.

I have tried running under gdb and under vnc4server, but somehow I get a
different segfault that I don't understand (and should investigate).

Yaroslav, could you investigate for me where the segfault is happening in
the Python code. For this, you can zcat
/usr/share/doc/python2.6/gdbinit.gz in ~/.gdbinit, and navigate the stack
in gdb after the segfault. You can use 'pyframev' to something like the
current Python frame. If you can give me an idea of where the sizer with
a None value is initialised, or which sizer it is, it would help (I have
a hunch that it is going to be the frame holding the VTK render window).

Another option would be if you can give me access (say VNC :>) to a box
for which running gdb exhibits the right segfault.

Sorry for not spending more time on that problem alone, I am swamped.

Gaël