Re: [Gnash-commit] gnash ChangeLog server/character.h server/dlist...

[Sandro Santilli]

On Tue, Oct 30, 2007 at 10:33:26AM +0100, Udo Giacomozzi wrote:
> Hello Sandro,
> 
> Tuesday, October 30, 2007, 10:17:16 AM, you wrote:
> SS> Yes, makes sense, thanks.
> SS> "top of each other" is not necessarely true though....
> SS> if the mask and the maskee do not overlap moving the mask out of the
> SS> maskee bounding box  shouldn't make the maskee invalidated.
> 
> The maskee should *never* become invalidated unless it changes itself.
> 
> The situation "mask and maskee do not overlap" is handled by
> add_invalidated_bounds(), btw. It intersects the mask bounds with the
> maskee bounds leading to NULL ranges and thus not adding anything to
> the invalidated *bounds*. This intelligence has been added some days
> ago. But anyway this case is *not* related to set_invalidated()
> because it is only about *avoiding* useless re-rendering (which is a
> performance issue in contrast to incorrect rendering).

Agreed, switched to performance-wise now.

> SS> Maybe more then talking about invalidated *state* the TODO is about
> SS> the invalidated *bounds*.
> SS> Like, get_invalidated_bounds for a masked object might only returns
> SS> the invalidated bounds resulting from intersection of own 
> SS> invalidated bounds and maks's invalidated bounds.
> 
> Yeah, that's what it does already. The relevant patch is:
> http://cvs.savannah.gnu.org/viewvc/gnash/server/dlist.cpp?root=gnash&r1=1.96&r2=1.97

If you remember I asked you "do you have setMask in mind" when you added
the new intelligence. Maybe time to prove it against setMask now ?
Note that dynamic masks (set by setMask) do not have to be in the same 
DisplayList at all !
(but I belive we don't have an automated test for this yet)

--strk;