Re: [Gzz-commits] gzz/gfx demo/paperbasis.py libpaper/texcomb_NV1...

[Tuomas Lukka]

On Sat, Sep 14, 2002 at 04:32:28PM +0300, Janne Kujala wrote:
> On Sat, Sep 14, 2002 at 03:35:05PM +0300, Tuomas Lukka wrote:
> > On Sat, Sep 14, 2002 at 06:22:28AM -0400, Janne V. Kujala wrote:
> > > Log message:
> > >   Create clamped, bordered versions of textures for BasisPaperQuad on 
> > > demand; mipmapping is disabled for clamped textures because I couldn't 
> > > get it to work correctly (should GL_SGIS_generate_mipmap allow reading 
> > > the generated lower LOD texture images?); 
> > 
> > Why do that and not just turn on GL_SGIS_generate_mipmap on the new
> > textures?
> 
> Because the border needs to be one pixel wide on every mipmap level
> (I'm drawing the border to the actual texture image data).
>
> I tried using the border parameter to glTexImage2D, but always got
> totally one-colored textures. Furthermore, the border feature isn't
> very well documented. For example, the border needs to be added to 
> width and height parameters but there is no mention whether the
> actual image data should also include the borders.

Jvk: yes, it should. The borders are extra texels that work differently
from the others in mipmapping.

> > > there is still the problem of choosing the texture color outside the 
> > > clamped box so that the something is visible when the two textures do not 
> > > overlap
> > 
> > Yes, it's true that it's not really meaningful; how about just
> > (.5,.5,.5)?
> 
> I tried 0, 0.5, and 1.0 but there wasn't much difference.
> 
> It would be nice to be able to show the actual dot products
> when the textures do not overlap and do the color interpolation
> only inside the intersection. This would require new combiner code
> or some stencil tricks.

Would be good.

> The band-like texture only uses one dot product 
> (and the dot product of the two textures inside the intersection).
> 
> For both types, the actual 3-component textures should also be
> shown in addition to the dot products.

Yes, make it so it's easy to flash through them.

> > And maybe drawing clearly colored lines for the boundaries of
> > the basis quads?
> 
> That should help but not with the above problems.

No, simply with showing things clearer.

        Tuomas