Well, I have found the culprit.  In a way, it *IS* an accuracy problem -- more
specifically, it is a division by zero problem.  The following code produces the
problem a little more handily:

#version unofficial MegaPov 0.4;
isosurface{
  function{(1/0)}
  contained_by {sphere{0,1}}
  threshold .5 pigment {rgb 0} finish {ambient 0}
}
camera {location <2,0,0> up y look_at <0,0,0>}

Apparently the Solaris and Linux compiles freak out upon seeing the division by
zero and return white -- no matter the color they are "supposed" to return.
Interestingly, adding "method 2" to the above isosurface "fixes" the problem, but
probably only because of the low value of MaxGradient.  Is there a test which
could be added to the method 1 section to stop infinite values from "whiting out"
the isosurface?
-Alex V.

Post a reply to this message