Re: Valgrind errors with UIManager

[Alexander Shaduri <ashaduri gmail com>]

> The bound refptr is taking an additional reference on the Action object and so that object 
> is staying alive longer than it normally would.  One theory is that when the Action object 
> does finally get destroyed in the TestWindow destructor, some piece of GTK+ data that it 
> depended on might have already been destroyed, resulting in the warnings you see.

Thanks for the reply!

(Taking this message back to the mailing list, I hope it's ok with you).

I changed the callback so that it would fetch the Action object from
an associative (action_t -> Action) map, which I made TestWindow's member.
Everything seems to work just fine now.
Just binding Gtk::Action* also seemed to work (I had to use
action.operator->() to get that type).

Anyway, the workaround is known now, thanks for your help!

I wonder if it's a bug in gtkmm/gtk, or just undocumented behaviour.
Should I file a bug report?

Thanks,
Alexander