[ODE] Question about squashed collision bug

[Russ Smith]

nlin at nlin.net wrote:
> 
> > Russ Smith russ at q12.org
> > [ODE] collision bug found!
> > it has also been the cause of the odd behaviors observed in the
> > boxstack demo, where boxes occasionally "pop up".
> 
> Could this also affect sphere-box collision? I have a demo of my own
> where a sphere collides with planes and boxes, and for testing I
> subject the sphere to certain random arbitrary forces. Sometimes the
> sphere goes flying much, much farther than it should

i think that situation would not be affected by this bug. this bug only
manifests when dCollide returns multiple points (e.g. for box-plane
or a GeomGroup), sphere-box will only return one point max.

> Also, might I ask how you found this error? Knowing how you tracked
> it down might be useful in doing our own debugging.

i noticed stability problems with the new GeomTranform class, i
examined the normals that were being returned from dCollide(), i
noticed that they were not all "swapped" correctly as they should be
in dCollide(), and there it was! so no, i don't have any magic physics
debugging tools, although they would certainly be useful.

russ.

--
Russell Smith
http://www.q12.org