PDF import problem with radial gradients and clip path

Possibly addressed by the fix in r10680 for
Bug #530895 in Inkscape: “Opening EPS with Gradient”
<https://bugs.launchpad.net/inkscape/+bug/530895>
(see comment #4 there)

> Tested with Inkscape 0.48.1, Windows 7 Professional 64 bits.

Please also test with current stable (0.48.2).

Displaced gradients in
- OuvragesMaritimes2010Vf.pdf
- FR_plaquette%20ferroviaire.pdf
reproduced with Inkscape 0.48.2 on OS X Lion (poppler 0.18.2, cairo 1.10.2)

Testing with Inkscape 0.48+devel r10911 on OS X Lion:
- OuvragesMaritimes2010Vf.pdf
gradient is positioned correctly, but scales slightly different than displayed in the PDF file
- FR_plaquette%20ferroviaire.pdf
gradient of clipped path(s) is not visible.

Seems partially fixed in current trunk.

Displaced gradients in
- OuvragesMaritimes2010Vf.pdf
- FR_plaquette%20ferroviaire.pdf
are imported and rendered correctly with the experimental input file format "Adobe PDF via poppler-cairo (*.pdf)" in current trunk.
(See also <http://inkscape.13.n6.nabble.com/pdf-import-again-poppler-native-vs-poppler-cairo-tt2812859.html>)

it appears that there are two (or more) separate bugs shown here.

a partial fix for one has been committed to bzr rev 11094.
this will properly display the gradient in the logo at the bottom right hand corner of page 7 of FR_plaquette ferroviaire.pdf. It also makes visible a long curved horizontal line just above the logo. There is still a problem with a rounded rectangle at the bottom left, which is not displayed completely.

a better fix for one bug has been committed to bzr rev 11112.
 this will properly display the rounded rectangle at the bottom left corner of page 7 of FR_plaquette ferroviaire.pdf. The previous fix was assuming that there were no rotations when a bounding box gets transformed, the current fix allows for the possibility of rotations or skewing.

in the file OuvragesMaritimes2010Vf.pdf, the radial gradient in the logo at the bottom right of page 1 is getting corrupted during the operation saveState() in the file pdfparser.cpp.

Before the saveState operation the parameters are (cx, cy, fx, fy, r) = (0, 0, 0, 0, 1) (these are typical of most radial gradients)
After the saveState operation the parameters are (cx, cy, fx, fy, r) = (0, 1, 0, 0, 1) (this change should not have occurred)

this is causing the observed distortion in the gradient. This appears to be a Poppler bug. In any event, I do not know how to fix it, so the workaround is to not refresh the state. A more elegant solution would be welcomed.

change committed to rev 11184

in the file OuvragesMaritimes2010Vf.pdf there is a remaining problem with the logo at the bottom right on page 1, namely the bright spot is too large. imho, this cannot be solved because it uses a feature not supported in Inkscape.

Using the Python package 'pdfminer 2011', this pdf file was parsed. It was found that this radial gradient uses a FunctionType = 2, which is a non-linear relationship between color and screen position, not supported by Inkscape, or svg as far as I can tell. The function is exponential, and is defined in section 7.10.3 in the document PDF32000_2008.pdf by Adobe.

I believe this addresses all the issues encountered in these pdf files, unfortunately not always successfully.

in the file OuvragesMaritimes2010Vf.pdf, the size of the bright spot at the bottom right on page 1 has been adjusted by implementing a very approximate version of an ExponentialFunction. By evaluating the slope of the exponential function at x = 1 and using this in a linear gradient, and by adding an extra stop, the bright spot now looks similar to the original pdf file.

committed to rev 11256

an improved workaround for the problem with the saveState operation in comment 8 above, has been committed to rev 11340.
This new version does not have the unexpected side effects that were seen in Bug 985182.

It didn't work in 0.48 stabke but IT WORKS NOW with 0.48+r12340 : all my PDF import with multiple and different gradients works fine :)