[cairo] Path fill bug

[Carl Worth]

On Fri, 06 Aug 2004 15:29:56 -0600, Charles Tuckey wrote:
> We have discovered a bug in the way cairo (or its backends) fill a 
> particular concave polygon. Using glitz this polygon has some extra fill 
> on its left side; using xlib or pixman the polygon lower interior is not 
> filled. Stroking the polygon works fine.
> 
> I have attached three cairo snippets that illustrate the problem:

Charles,

Thank you for the excellent bug report along with precise snippets for
duplication.

The problem you've seen is a known bug[1] in the current CVS version of
the code. It's been there much longer than I would like while we are in
the midst of re-writing the tessellation routine within cairo.

To work around your problem, I suggest you back off to the most recent
cairo snapshot which, as I just checked, does not exhibit this bug with
your test cases:

	http://cairographics.org/snapshots/cairo-0.1.23.tar.gz

I don't know if you're using glitz or not, but if so then you'll need to
back up a few snapshots before you find one that is compatible with
cairo 0.1.23:

	http://cairographics.org/snapshots/glitz-0.1.3.tar.gz

The current skew between glitz and cairo is precisely because I do not
plan to make a new snapshot until the new tessellator is finished and
this bug is licked.

-Carl

[1] As seen in cairo/BUGS:

The polygon tessellation routine has problems. It appears that the
following paper has the right answers:

        http://cm.bell-labs.com/cm/cs/doc/93/2-27.ps.gz

        [Hobby93c] John D. Hobby, Practical Segment Intersection with
        Finite Precision Output, Computation Geometry Theory and
        Applications, 13(4), 1999.

Recent improvements to make the intersection code more robust (using
128-bit arithmetic where needed), have exposed some of the weakness in
the current tessellation implementation. So, for now, filling some
polygons will cause "leaking" until we implement Hobby's algorithm.