Kevin Ryde wrote: > With a perl custom treemodel and a perl connected rows-reordered signal, > there seems to be a gremlin in the way $model->rows_reordered reaches > the signal handler. [... > > I wonder if it might be something like > > * xs/GtkTreeModel.xs (gtk2perl_tree_model_rows_reordered_marshal): > Call gtk_tree_model_iter_n_children before building the stack, in case > it's a Perl class and calls out to ITER_N_CHILDREN. This fixes > clobbering of instance and treepath args to a Perl rows-reordered > handler connected on a Perl model. > * t/GtkTreeModelIface.t: Exercise rows-reordered marshalling.
Yeah, I think you're right. When we call out to gtk_tree_model_iter_n_children and thus eventually ITER_N_CHILDREN, our local stack *pointer* is not modified (since it's local), but the actual stack contents are, and this is what you saw. Committed to both branches. Thanks! (Unfortunately, I didn't see the ChangeLog entry you wrote up early enough to use it for the commit. Sorry.) -- Bye, -Torsten _______________________________________________ gtk-perl-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/gtk-perl-list
