[SOLVED] NewtonCollisionClosestPoint not every returning 0?

[pHySiQuE]

It appears that NewtonCollisionClosestPoint is never returning 0 (for intersection tests) even when the distance between the closest two points is extremely small. I don't know why, because this was previously working.

[Julio Jerez]

collision closest point does not return zero, in newton 300. It used to in newton 200

In the past it use to return zero when the two shape where interpenetrating, with core 300 it always returns the closet distance.
the penetration behavior is now an end user thing.

the function return point p, point q and a normal.
to determine interpenetration you can perform this test

Code: Select all

dist = dostProduct ((q - P), normnal)
if (dist < 0.0) {
   //shapes are interpenetrating. here you can assume return zero
}

[pHySiQuE]

With my sphere test, your code above works correctly. However, when I try getting the closest point with a collision tree I get these huge numbers. The large wall the player is going into is being tested against.

Image4.jpg (105.26 KiB) Viewed 1769 times

[Julio Jerez]

are you tyry to calculated close point on a collision tree vs anoteh collsio tree?
I do not think that works and has never being supported by newton, that's is not a trivial operations.

are you try to call close point between a convex object and a collision tree?
I do not remeber if that works, or has even worked. It looks like it should since the collesion tree is made of convex polygons.

can you provide a test so that I can try?

[pHySiQuE]

It's a cylinder and collision tree.

[pHySiQuE]

Ah, that doesn't make sense, since a collision tree has no volume.

Okay, I just fixed it so I am using a convex hull for those trigger volumes. Problem solved.

(The reason I am doing an intersection test and not simple collision is because this is for volumetric trigger fields, where the trigger will continually be activated when the player is inside it.)