Raab, Andreas wrote:
> 
> Congratulations! You just discovered one of the most serious bugs in the
> Balloon engine. I'm amazed nobody stumbled across that earlier ;-)

Gee, I feel honored.  :)

> Seriously, what happens is that for the sake of speed I'm doing an extremely
> ugly (but very fast!) quadrant interpolation scheme for gradient fills.
> *Usually* this works just nice (as you noticed for everything that is
> symmetric) but for the particular case you've shown it does not.

Okay, it's supposed to work for ovals, but the interpolation doesn't
quite work.  Now, is it also supposed to work when the normal isn't
normal to the fill's direction?  I think that should be equivalent to
some rotated oval, and it does look much the same, but it seems wrong
to violate the implied constraint.  (There were a couple of places
where I got some nice effects without having to do much calculation by
setting the normal at some non-right angle to the fill direction.)

> The only way to work around the problem is using a bitmap fill generated
> from a radial gradient. This should work until I've fixed the problem (which
> will take some time).

Thanks for the suggestion.  Now, is there a way I can make the bitmap
fill use translucent colors?  Aw, what am I asking for, I was going to
play around with those soon, anyway!  :)

-Jesse