Hello JML and Luca, Following your advices, I reverted liblarch_reverted to the best state so far.
Only one test is failing so far (it's a new "torture" test). The current state is that filteredtree.py is in fact a double-linked tree reprensented in self.cache_nodes. For each node, we keep the 'parents', the 'children', and the 'paths'. When a path is changed, the node is not modified : it is removed from the old path the added to the new path. I think that's why we had so many crashes before with GTK.Treeview. Gtk.Treeview doesn't support modifying the path of a node : it should be deleted then re-added. (if this is proven true, it would mean that gtk.Treeview is even worst than what I thought, to the point of being useless) The good : + For small datasets, it should works fine. + There's no crashes in GTK.treeview. All crashes are confined into liblarch/filteredtree.py (because it does a lot of sanity check) The bad: - still too many crashes for large dataset - performances are bad but I guess that we don't mind at this stage _______________________________________________ Mailing list: https://launchpad.net/~gtg-contributors Post to : [email protected] Unsubscribe : https://launchpad.net/~gtg-contributors More help : https://help.launchpad.net/ListHelp
