After some extensive testing, including compiling my own version of POVray,
I have a conclusion. The 'slope bug' is not a bug but a feature, once again
the common case of the computer doing exactly as told, instead of doing as
wanted.

The easy demonstration about how the 'misfeature' occurs is to render the
'slope bug' scene with quality setting lower than 6, say, +Q3. This way both
sides look equal. With +Q6 or higher, the backfacing intersection kicks in,
and this time the intersection normal points inwards! Therefore, near the
top of the sphere, we first get totally transmitting blue on the incoming
ray, which is basically invisible. With the outgoing ray, however, the
normal points inwards, which in this case is almost entirely downwards, thus
giving the opaque yellow at the other end of the color map.

The 'equivalent function' does not have this behavior since it does not care
about the surface normal at all, but is a pattern based on the intersection
point coordinate.

I'll  let someone else decide whether this is a bug or not.

Another thing I noticed was that 'slope' function has no respect for any
object transformations at all, but always operates on the world-space
surface normal. Just a point of reference.

-- MeowCat

Post a reply to this message