[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [E-devel] Menu heisenbug caught.

On Fri, 10 Nov 2006 03:33:17 +1000 David Seikel <onefang@gmail.com> babbled:

> I've been trying to catch this one for a while.  It's a wily little
> critter that tends not to trigger when you are looking for it.  I have
> found one way of reproducing it at will, although it might take several
> goes.  I have attached a backtrace.  Others have reported it.  No one
> has managed to get a useful bt, until now.
> Sometimes when bringing up the Applications menu, E will segfault.  If
> you hit the restart button, then you are no longer running from your
> home directory, but what ever directory the icon it crashed on happens
> to be in.  I had discussed this one with raster and others in the past,
> our conclusion was the some graphics library that is handling the
> loading of icon files for us is changing directory while it loads.
> Somehow, things get screwed, and the next time it tries to load that
> icon, it crashes after the change of directory.
> I have never seen it hit when just running E by itself.  It seems to
> require starting up, and maybe closing down some other program.
> Seems that the program being run is important, a couple seem to be
> really good at triggering it.  It never has hit me when running E under
> gdb or xnest, or the combination of the two, I caught this one by
> attaching gdb after the crash.
> The way to reproduce it involves a specific distro (Ubuntu), probably a
> specific version of the distro (Edgy release with all updates), and at
> least one specific program.  Synaptic can cause it to crash I think.
> Globulation 2 is good at causing it.
> I'm going to bed now, others can have a look at this and see what they
> come up with.  It's likely that the only useful answer we can get out
> of this is "don't run synaptic and globulation on ubuntu edgy".

crash inside rsvg->glib->dl stuff...

now.. this could be a rsvg or glib bug.. it could be a bi-product of evas
UNLOADING loader modules and later re-loading them as needed and this
triggering some as-yet undiscovered bug in rsvg/glib... it could also just be
that evas or edje or e or whatever corrupted the heap or stack so that
glib/rsvg/dl don't work right later... hard to tell.

------------- Codito, ergo sum - "I code, therefore I am" --------------
The Rasterman (Carsten Haitzler)    raster@rasterman.com
Tokyo, Japan (東京 日本)