[cairo] perf: long dashed lines

[Vladimir Vukicevic]

Ok, here's the two patches with updated changes:

Patch 1:
-- fixed the 0.0 check to happen before the sqrt(), and check for  
explicit 0.0
-- changed behaviour in stroker_curve_to to behave the same as  
previous code; start/end faces are not initialized correctly in the  
case of 0-length segment.  This is wrong I think, but it's what was  
there before, so we can fix that in a followup (I'd fix it but I'm not  
sure what the correct fix is).

Patch 2:
-- moved point/line tests to cairo-rectangle.c and renamed to  
_cairo_box_*
-- fixed line-rectangle intersection to actually work correctly
-- fixed bounds expansion based on comments; can you guys double-check  
this in stroker_init?  I err on the side of caution and expand by the  
full transformed width/height, or by width * miter_limit.  MAX(width,  
width * miter_limit) might be better, even?

The slight changes in dash-scale output are still there, caused by the  
first patch.  I'm pretty sure this is because the slope values given  
to _compute_face before were converted to fixed-point first and then  
converted back to doubles, whereas with these changes they stay as  
doubles throughout.

How's this look?

     - Vlad

-------------- next part --------------
A non-text attachment was scrubbed...
Name: cairo-stroke-1.patch
Type: application/octet-stream
Size: 11086 bytes
Desc: not available
Url : http://lists.cairographics.org/archives/cairo/attachments/20080204/2bf9e0fc/attachment-0002.obj 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cairo-stroke-2.patch
Type: application/octet-stream
Size: 10639 bytes
Desc: not available
Url : http://lists.cairographics.org/archives/cairo/attachments/20080204/2bf9e0fc/attachment-0003.obj