[cairo] shading, take 2

[Carl Worth]

On Mar 3, Thomas Hunger wrote:
 > Hm, with cairo I have the following problem on my system:
 > when I compose a surface with an image backend (my gstate->source) against my 
 > gstate->surface which has an X11 backend, it does not use the repeat ==1 
 > although it is explicitly set in the pixman backend. Instead it behaves like 
 > gstate->source->repeat == 0.

If you can provide a minimal test case, I can look at this.

But, even without that, I have a suspicion about what the problem
might be. There is a known bug in libpixman in that the repeat
attribute is not properly handled when a transformation matrix is
applied to the source image. If I'm not mistaken, cairo is always
applying a transformation matrix, (perhaps an identity matrix, but
that should be enough to tickle the bug I think).

-Carl