[AD] Patch for polygon3d() and related examples/tests

["Ben Davis" <ben@xxxxxxxxxx>]

This is a multi-file patch, rooted in allegro/ . It replaces the patch I
posted that conflicted with CVS, and contains extra changes.

1. Polygons with two coincident vertices are now always drawn correctly.
2. Setting whole numbers for u and v specifies the top-left corner of a
texel.
3. This change is explained (or rather emphasised) in the docs.
4. ex3d.c and test.c now display the entire bitmap.

I hope the change in the behaviour of u and v doesn't offend anyone - if it
does, maybe we can have some sort of vote. If no-one objects, can someone
commit the patch please?

Not fixed by, and AFAIK not caused by, this patch, there is an error in the
grgb drawing code (at least in polygon3d() - I don't know about
triangle3d()). In very thin bits of the polygon, the deltas seem to do
something funny and some of the polygon doesn't draw - or something like
that. See for yourself - it shows up in the test program. This is only true
for grgb though, so it isn't a fault in the sub-pixel accuracy.

Ben Davis

Attachment: poly3d.diff
Description: Binary data