[cairo] Fix for get-path-extents test cases

[Chris Wilson]

On Wed, Jun 19, 2013 at 10:51:34PM +0000, Bryce W. Harrington wrote:
> Chris,
> 
> I ran across bug 62375 about the get-path-extents test case breakage.  You posted a patch there a few months back, to switch from _cairo_path_fixed_stroke_polygon_to_traps() to just _cairo_path_fixed_stroke_to_polygon(), which makes a couple dozen tests pass (below) and none fail.  The former appears to be running _cairo_bentley_ottmann_tessellate_polygon over the polygon and I gather this was what breaks the test.  Unfortunately the output files for this test seem to be empty when it passes, so I'm not sure what's happening here.
> 
> Could you elaborate on your thoughts regarding that patch?  Is it just a workaround, or would it be suitable as a fix?

Avoiding the intermediate traps for determining tesselated extents is
the right approach. The other question though is whether the error for
the traps path is less than tolerance - the test should be fixed not to
expect answers to higher precision than tolerance and should also test
with varying tolerance. It would be worth checking that first.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre