wrong color in PDF export

[Johan Engelen]

Export bug.svg to PDF (0.48+dev, r9841), the sine wave bottom left is colored green instead of black (see bug.pdf).

The cause for this is probably the default color of the sine, AND the default color of the stroke of the object before the sine.

Simplification of the SVG:
  <path style="fill:none;stroke:#097101;stroke-width:2;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-miterlimit:4;stroke-dasharray:none" /> the green line

   <path style="fill:url(#linearGradient6712);fill-opacity:1;stroke:black;stroke-width:0.6;stroke-miterlimit:4;stroke-opacity:1;stroke-dasharray:none;stroke-dashoffset:0" /> the red object

   <path style="fill:none;stroke:#000000;stroke-width:0.56444448000000003;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-opacity:1;stroke-dasharray:none" /> the sine wave

The sine wave in the PDF export has the same color of the green line.
The color is correct when changing any of these:
- change fill of red object to solid color
- change stroke color of red object to something other than black
- change stroke color of sine to something other than black
- change stroke-opacity of red object to e.g. 0.9

After some testing, I think this bug is caused by Cairo. (and has to do something with black (rgba=0,0,0,1) being the default color...)

[su_v (suv-lp)]

reproduced with Inkscape 0.48+devel r9841 on OS X 10.5.8:
sine wave at the bottom left is colored green both with export based on cairo 1.8.10 and based on cairo 1.10.0 (using PDF 1.4 and 1.5)

[Johan Engelen (johanengelen)]

It is a bug in cairo, see my bugreport here: https://bugs.freedesktop.org/show_bug.cgi?id=31140

[Johan Engelen (johanengelen)]

Fixed in Cairo: "It has been fixed in the git master and 1.10 branch so it will appear in the
1.10.2 (stable) and 1.11.2 (development snapshot) releases."

[su_v (suv-lp)]

@Johan - if you use a cairo build from git master, any chance you could test if bug #514696 “group fill color is unset in PDF export in 0.47” is related to your reported issue and fixed with the new cairo version?

[Johan Engelen (johanengelen)]

I do not build Cairo myself, so cannot test without investing a lot of time.

[madbiologist (me-again)]

Cairo 1.10.2 has been released - see http://cairographics.org/news/cairo-1.10.2/

[su_v (suv-lp)]

No longer reproduced with Inkscape 0.48+devel r10000 on OS X 10.5.8 using Cairo 1.10.2

Closing as 'Invalid' in Inkscape since the bug is in Cairo and fixed in its latest stable release.
Thx @Johan for investigating and reporting it upstream with cairo.

[su_v (suv-lp)]

> bug task added: inkscape-devlibs

see <http://article.gmane.org/gmane.comp.graphics.inkscape.devel/35720>

[su_v (suv-lp)]

Note to Windows users:

Due to a regression in Inkscape 0.48.1 with cairo 1.10.2 (bug #707255), the official package (linked from inkscape.org) has been reverted to use the old version (cairo 1.8.8). If you need a newer version of cairo for better PDF export, consider installing the earlier 0.48.1-1 package still available on sf.net:

- cairo 1.8.8: Inkscape-0.48.1-2.exe
- cairo 1.10.2: Inkscape-0.48.1-1.exe
<http://sourceforge.net/projects/inkscape/files/inkscape/0.48.1/>

[jazzynico (jazzynico)]

Tested again with the latest gtk bundle (2.24.8). The PDF issue is fixed as expected, AND the regression found in bug #707255 is not reproduced anymore.

[jazzynico (jazzynico)]

Fix confirmed on Windows XP with Inkscape r10938 and devlibs r27.