[ODE] AutoDisable

[Adam D. Moss]

gl wrote:
>>This would work at least for some cases
> 
> Wouldn't it work for all?  The problem should only occur when bodies are
> resting/touching, so if they don't already intersect AABBs, they should
> after scaling?

I think that this is a sound approach, but it does require
some thought to get right so that the temporary swelling of
the AABBs does for example put the AABB into the right node of a
hashspace, etc, or you could still end up with two geoms whose
expanded BBs definitely overlap but they're never given a chance
to be tested.  It's probably safe to diddle with the AABB directly
as long as it's verified that this causes the geom to be re-inserted
into the right place in its space's structure.  Alternatively instead
of playing with the AABB you could temporarily expand the raw
geom's dimensions (box side lengths, ccylindar radius, etc) which
should safely go through a dirty-AABB path.

--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3