[cairo] Different output between Linux and Mac

Neil Mayhew neil_mayhew at users.sourceforge.net
Sun Mar 15 22:19:47 PDT 2009


I have a program that outputs to SVG, PDF, PS, or PNG. It is written 
using cairomm. The same sources build on both Linux and Mac OS X. 
However, I get slightly different outputs on the two platforms. I have 
dumped the coordinates I am giving to move_to and line_to, and these are 
identical on the two platforms (to six significant digits, at least, the 
default precision for std::cout). Also, both platforms are using Intel 
CPUs. However, the cairo output is not identical.

It's easiest to compare the SVG outputs, using diff/meld, but the PS 
output exhibits the same differences. Also the PNG outputs are slightly 
different: I used ImageMagick to subtract one image from the other and 
viewed the resulting image which was all black except for a couple of 
noticeable lines in the same places that I am seeing differences in the 
SVG/PS output.

The SVG/PS differences are localized to an endpoint of just one line 
segment in the middle of two longer paths. Everything else is the same, 
down to the last digit. However, although these differences are not 
visually noticeable (they are in the 4th decimal place) they are causing 
me a problem with my regression tests. I have checked sample output into 
my source repo from Linux, but when I am developing on Mac I get 
different output so I can't check for regressions without manually 
inspecting the output.

Am I correct in thinking that I should get identical output on the 
different platforms? If not, why would almost all the output be the same 
except for just a few places? I noticed that the coordinates I dumped 
from my code and the coordinates in the SVG are the same up to the 3rd 
or 4th digit, but are completely different thereafter.

The package version of cairo on my Linux system (Ubuntu Jaunty alpha) is 
1.8.6-1ubuntu2. On Mac I am using MacPorts and the version is "1.8.6, 
Revision 4". I can send sample output and code if desired.

Thanks for any help offered.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cairographics.org/archives/cairo/attachments/20090315/ed59070c/attachment.html 

More information about the cairo mailing list