Sphere/Box collision creates contacts for bounding boxes

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Place a kinematic sphere and a kinematic box
2. Let the box touch the AABB of the sphere but not the sphere itself
3. Iterate in the TickCallback over the contacts. A contact is stored with a 
positive pt.m_distance1 

What is the expected output? What do you see instead?
No contact should be stored when only an AABB is touched, but only when the box 
really touches the sphere.

Original issue reported on code.google.com by linz...@gmx.de on 21 Apr 2011 at 10:30

[GoogleCodeExporter]

Bullet generally reports contacts with positive distance for various reasons 
(stability, contact predictions etc). 

The getContactBreakingThreshold/getContactProcessingThreshold might not be 
taken into account though.

Original comment by erwin.coumans on 11 May 2011 at 9:15

[GoogleCodeExporter]

Yes i know, but I thought such positive contacts would only be remembered when 
they lie within the shapes margin? Also it seems the issue happens only for 
sphere AABBs but not for rotated boxes AABBs.

Original comment by linz...@gmx.de on 12 May 2011 at 12:46

[GoogleCodeExporter]

Contacts might be reported at a positive distance, it depends on the collision 
shape type indeed.

So it is best to add an additional check for the distance, if needed.
The latest trunk reverts some behavior (introduced in 2.78) that caused 
additional positive distance contacts.

Original comment by erwin.coumans on 2 Sep 2011 at 4:35

• Changed state: Fixed