Brian and others,
Brian Paul writes:
> I'm familiar with the history but I thought the 1.2.1 spec was
> supposed to be the final word.
Maybe this particular issue was hashed out in an ARB meeting.
If so, the resolution never got back into the man pages
and the sample implementation.
As you note, all the other new commands that take an
internalformat use GLenum for their type. That would imply to
me that there was a conscious effort to use GLenum as
the internalformat type for new commands and that the
glTexImage3D routine was simply an oversight. I subscribe
to the "what's the most likely mistake and what's the
easiest to fix" theory.
The 1.2.1 spec is pretty good, but it has enough mistakes
and oversights. As I mention, the LOD clamping stuff is
badly mis-specified.
And the following will not generate an error if you read
the spec literally:
glEnable(GL_CLIP_PLANE0+6);
glDisable(GL_CLIP_PLANE0+6);
Light sources do not have this problem.
As another counter-example to the "final word" theory for
the 1.2.1 spec, consider Table 4.5. Where did the
GetTexImage1D, GetTexImage2D, and GetTexImage3D commands
come from? And why is GetPolygonStipple missing from
the list?
- Mark
