Re: [grt-talk] textures

[Nikodemus Siivola]

On Tue, 15 Apr 2003, Simon Adameit wrote:

The logic for intersection normals was originally (and the definition for
inside-p was slightly different as well):

  (if (inside-p obj-a point)
      (normal obj-b point)
    (normal obj-a point))

So I was using inside-p as a kind-of not-on-surface test. And a lot of time
when the point was really on the surface of obj-a the inside-p would return
true anyways: so I got a lot of bogus normals. Testing inside/proximity to
be within grt-epsilon of 0.0 seemes more reliable: if I do get the normal
of the "wrong" object, the surfaces are so close to each other that it
doesn't really matter.

Cheers,

  -- Nikodemus