[anjuta-devel] Glade-anjuta integration crashes on raspberry pi

[Marco Diego Aurélio Mesquita <marcodiegomesquita gmail com>]

Hi devs!

I'm cross posting it because it has been a couple of months
investigating this problem with very little progress and this problem
may be related to anjuta as much as glade.

Please, take a look at bug
https://bugzilla.gnome.org/show_bug.cgi?id=787738, specially trace
https://bugzilla.gnome.org/page.cgi?id=traceparser/trace.html&trace_id=238242.

I'm not sure if it is a bug in anjuta, libglade, don't know if it is
caused by stack corruption or if it is thread related.

I find very likely that it is thread related because while I was
investigating it, I added some breakpoints inside GTK+ and after some
breaks, continue, steps and stepi's, I could successfully load a .ui
file in anjuta. Unfortunately, I could not reproduce the success
again.

The segfault happens when, deep into GTK+, a _gtk_widget_get_toplevel
is invoked. Since it loops depending on widget->parent, it seems clear
that the parent of a widget is being corrupted. I don't know if it
because of a race condition in threads or some corrupted stack is
changing its value. Glade is known to change widget parenting in fancy
was, so it may even be another totally unrelated thing.

In anjuta source code, the crash happens when
gtk_notebook_prepend_page_menu is called from
anjuta_docman_add_document. If the call to gtk_widget_show in
glade_plugin_add_project is moved to after the call to
ianjuta_document_add_document, then the crash happens when
gtk_widget_show is called. In both cases it  segfaults in
_gtk_widget_get_toplevel.

Please help me, I'm running out of ideas. Does anyone has a line that
can be useful to investigate this problem? Any suggestion, tool or
experiment that I can perform or use to gather more data about it?