Re: Equivalent of the gtk g_object_ref/unref functions in gtkmm?

[Jef Driesen <jefdriesen hotmail com>]

[Repost, forgot the send a copy to the list]

Paul Davis wrote:
> On 3/26/07, Jef Driesen <jefdriesen hotmail com> wrote:
> > Murray Cumming wrote:
> > > On Mon, 2007-03-26 at 16:30 +0200, Jef Driesen wrote:
> > > > Jonathon Jongsma wrote:
> > > > > On 3/26/07, Jef Driesen <jefdriesen hotmail com> wrote:
> > > > > > The application works fine, but closing the main window results in a
> > > > > > segmentation fault (core dump). As far as I can tell (from a gdb
> > > > > > backtrace) the crash is not in my own code, but somewhere inside the
> > > > > > gtk(mm) library.
> > > > > Could you provide the backtrace?  that might be helpful information.
> > > > Sometimes I don't get the segmentation fault, but only a large number of
> > > > errors. See attachment for both the backtrace and the errors.
> > > If you get glib errors, try debuggin with --g-fatal-warnings, or with a
> > > breakpoing on g_log. Then you'll get a backtrace at the relevant point.
> > Backtrace attached, although it doesn't seem to provide much more
> > information to me.
> >
> > > valgrind's memcheck is also incredibly helpful. Trust us on this. You
> > > won't regret trying it.
> > I get lots of errors from valgrind, but I also get them for the trivial
> > program
> >
> > int main (int argc, char *argv[])
> > {
> >      Gtk::Main kit(argc, argv);
> >      Gtk::Window window;
> >      kit.run(window);
> >      return 0;
> > }
> >
> > But to be honest I never used valgrind before, so I could be using it wrong.
>
> By chance, does your vbox3 that you pull out of glade contain a
> Gtk::Menu? I seem to remember reading somewhere about a problem
> pulling the menu out of glade but not the window that contains it.
>
> Although, I can't recall this ever happening to me...

That could be the problem, because the vbox contains an (unused)
menubar. Removing that menubar from the glade file fixes the problem.

> You might try having your Application class have the top level window
> as a member variable that gets pulled out of the glade file instead of
> inheriting from Gtk::Window.  You should be able to test that in about
> 5 minutes and see if it makes a difference.

This also fixes the crash.