[Bug?!] PhysX 3.3x - Rotation breaks overlapMultiple(..) - returns only 1 possible hit


I get sometimes a deterministic “bug”:

I know overlapMultiple is deprecated, but maybe this “bug” is in other PhysX versions too.

The “bug” is that you only get 1 hit on your overlapMultiple(…) as result, which cant be because I have many other shapes which touch it. With normlized values I get the right count of hits.

This is the value which breaks the overlapMultiple(…) (not normalized)
w: 0.50009 x: -0.50000 y: -0.50000 z: 0.50009
(So I guess there should be somewhere marked that only normalized values are valid)

But after I normalizing:
w: 0.50004 x: -4.99996 y: -4.99996 z: 0.50004
With this new value, the overlapMultiple(…) works again.

Should there an assert that you are using a rotation which isnt normalized?

Without normalizing, everything seems to works perfectly, but only sometimes it doesnt.
(After playing around I found this value above - I forgot to normalize my rotation calculation)

I have another question: When will be PhysX 3.3 (non beta) released?
Or maybe another new version?

Its almost 5 months ago where Nvidia released the PhysX 3.3 beta.

( I know you had much to do, but … we would like to see something more in this SDK…)

When will be PhysX 3.3 (non beta) released?
It was released yesterday through RDP.


thank you. After clicking thru the sometimes confusing websites of PhysX, I finally download it.

I didnt read the news before.

(And I like this great news :D)