Here is a copy/paste of my conversation with Daniel on IRC and he asked me
to file a bug:

<madmaxer> yesterday i had a very strange crash in my app
<madmaxer> you saw gex some time ago, remember?
<danielk> yes
<madmaxer> so, it has hpaned, inside right pane there's a vbox
<madmaxer> this vbox contains treeview, hbox, and statusbar
<madmaxer> now the hbox: it is a class derived from hbox actually
<madmaxer> in ctor i dynamically create two managed buttons and pack them
into the hbox
<madmaxer> now, when i close the application -> segfault
<madmaxer> so i changed these buttons from dynamically allocated to class
members -> no more crash
<madmaxer> all that with g++ 3.2 and gtkmm 1.3.23
<danielk> ah yes
<danielk> lifetime issues
<madmaxer> but this crash was only if i right clicked the button so it had
focus
<madmaxer> when button wasn't focussed on close there was no crash
<danielk> is there anything connected to the button?
<madmaxer> is that description meaningful for you in any manner?
<danielk> any code accessing it?
<madmaxer> yes, but it didn't matter, i removed all signals and nothing changed
<madmaxer> ah, i forgot, the crash was quite strange
<danielk> that's weird
<danielk> tell
<madmaxer> because actuall segfault was in gtk_paned_calculate_position or
something like that
<madmaxer> i don't remember exact function's name
<madmaxer> this function was called somewhere from the dtor of one of objects
<danielk> hmm
<madmaxer> i tried to recreate a simple sample code showing this bug but i
failed
<danielk> interesting
<madmaxer> but i think if it was some bug in my code changing objects from
dynamic on heap to class members would not solve the problem?
<danielk> did you include the paned in your sample code?
<madmaxer> yes
<madmaxer> but in my application i have many paneds, h/vboxes and other stuff
<madmaxer> my gui is already very complex, so i wasn't patient enough to
find which element exactly causes the problem
<danielk> what's funny about this is that lifetime issues usually occur
with class members and go away if you manage them, not the other way around
<madmaxer> to be honest i had similar case some time ago while i was still
using gcc 2.95
<danielk> I don't think it's a compiler problem
<madmaxer> but then it happened together with other stability problems i
reported
<madmaxer> and then i thought it was a compiler issue
<madmaxer> but now it happened again
<danielk> gtkmm lifetime handling is horribly complex.  Murray worked on it
recently, so I don't even know the current state.
<danielk> It's full of work-arounds and special casing
<madmaxer> not good
<danielk> file a bug
<madmaxer> but murray will say "provide simple as possible code showing the
bug" and i won't be able to do it
<danielk> that's right, but I'd like to have this recorded somewhere
<danielk> even if we don't find a quick fix
<madmaxer> if you can wait some time i'll check something
<danielk> no rush
<madmaxer> i downloaded a release tarball of my app from sf and i will build it
<madmaxer> to see if it also crases in the same way
<madmaxer> i've found the bug very recently, but it may be there since a
long time

I failed to recreate a simple example showing this bug, but if you want to
investigate the problem please go here:
http://sourceforge.net/project/showfiles.php?group_id=52990
and get Gex 0.0.6B or earlier (0.0.6B is for Gtkmm 1.2.23).

To make Gex segfault run it, if command line is not visible choose menu:
View/Command Line. Then right click the small 'Run terminal' icon to focus
it and close the application. It will segfault.