[cairo] FP errors in cairo

[Carl Worth]

On Sun, 01 May 2005 23:17:54 -0400, John Ellson wrote:
> I see from the comments that there are other concerns about this code.   
> Is there a rewite in the works
> or should I take a shot at improving the stability of this version?

Yes, we want to throw that particular function away.

Keith wrote a new intersection routine that is much more careful about
accounting for every single bit in the computations. His code is
already in that same file. You can enable it by changing the define to
read:

	#define CAIRO_TRAPS_USE_NEW_INTERSECTION_CODE 0

And when you do this, you should find that some paths fill improperly,
(cairo/test/fill-rule.c should flip from PASS to FAIL with that
change). And see some comments in that test for more details.

I don't know if the remaining failure is due to more bugs in Keith's
new intersection code, or if his code is just exposing bugs in the
higher-level tessellation code, (which is also scheduled for a rewrite
--- partially completed in nickle and waiting for me to finish all
this API rework).

Any help with any of this would be greatly appreciated.

-Carl
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.freedesktop.org/archives/cairo/attachments/20050509/7252d160/attachment.pgp