On Mon, Feb 11, 2008 at 11:26:11AM +0100, Udo Giacomozzi wrote: > Hello strk, > > Friday, February 8, 2008, 5:40:16 PM, you wrote: > s> I attached the .as files and .swf files to this bug item: > s> https://savannah.gnu.org/bugs/index.php?22258 > > Seems that the problem is solved now?
Yes thanks. The loading mechanism was bogus. It turns out when you call loadMovie agains a sprite and then unload the sprite, the movie is still loaded only if a new sprite with same target as old one is available at load time (one frame after the request). > I anyway think that this code is useless, so I removed it from > sprite_instance and improved generic_character instead. This > simplifies the code and probably offers slightly better performance > for movies with complex display lists. Did you take care of masks ? both dynamic and layer ones ? I mean, if a mask is out of the clipping area, we still want to display it IFF the display is meant to construct the "mask", right ? OR, we may skip it's rendering as far as we also skip rendering of any maskee.. > A question about do_display_callback(): What is it's purpose? With the > old implementation it wasn't called unless the character has been > rendered (changed) in that frame. Now it is called in each frame for > each character (seems more correct to me since it would be done that > way without invalidated bounds optimization). It is probably unused old code, I'd drop it if nobody else has an hint about it. > Anyway, I would like to know the effects of missing calls to > NON-rendring calls in display() implementations. IIRC I tested this a bit in the past with the 'skip-rendering-if-late' patch. Actually, it should still be in head, switched by a compile-time macro in gui.cpp. It would skip movie_root::display as a whole.. --strk; _______________________________________________ Gnash-dev mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnash-dev
