[ODE] Flat-ended cylinder to sphere collision unstable on cylinderends

[Adam D. Moss]

John Miles wrote:
> If you hit the 'a' key to show geom AABBs, you'll see that the cylinder
> AABBs are sometimes quite a bit larger than expected in the radial
> dimensions.  As a dCylinder rolls across the landscape, its AABB expands and
> contracts periodically, suggesting some unintended asymmetry in the math.

Note that this is very likely a red herring if you're looking
for the cause of instability/oddness.  A too-large AABB is
harmless and common (albeit suboptimal for the coarse collision
pass), usually because of the expense of calculating the optimal
AABB for some primitives.  A too-small AABB would be a real bug
though.

--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3
"At this point the rocket becomes engorged with astronauts."