On Thu, Apr 19, 2007 at 07:47:51PM +0200, Udo Giacomozzi wrote:
> The old sprite_instance-based implementation stored the result of
> getTarget() and getTargetPath() in a local variable for faster
> subsequent accesses. However, I removed this cache while moving the
> functions to character because:
>
> - changing character's parent (is that allowed?) would void the
> cached target and lead to unexpected results
Changing the character's parent is not allowed.
Changing the character's _parent._name is, which would also be
harder to handle by any cache.
> - this was a known risk and the old implementation included an
> assert() that caused re-calculation of the path anyway
It's incredible the assert() didn't help finding the above possibility.
I committed a testcase for this in MovieClip.as now.
It shows that _target can change, and also (sic) detects a flake in
our current "soft references" ('movieclip' typed as_values)
--strk;
_______________________________________________
Gnash-commit mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/gnash-commit
