Bug 7856

This one is not going to be easy.

Created attachment 6539 [details]
The said pdf.

Attaching the pdf, who knows if the url will change in the future.

*** Bug 6763 has been marked as a duplicate of this bug. ***

Here's another doc that seems to have the same problem:

  http://www.londontown.com/url/Christmas2005.pdf

Created attachment 10555 [details]
PDF which shows transparency not working in poppler

Another very simple testcase. Created with GIMP (gfig, saved as gif) and OpenOffice PDF export. It correctly displays in Debian etch kpdf, but not evince.

Any chance you could get to try this with poppler 0.5.9?

with 0.5.9 it's only reproducible with cairo backend. It looks like a bug in xpdf 3.01 that was fixed in 3.02. 

(In reply to comment #6)
> Any chance you could get to try this with poppler 0.5.9?

I tried with pdftops, but the resulting postscript is apparently completly broken, as no postscript viewer could show more than white paper.

Created attachment 10573 [details]
the broken postscript file

The problem with pstopdf should be fixed now in cvs head.

Please next time do not attach gz files, i got confused by you saying that binary blob to be a PS file :D

CairoOuputDev doesn't implement transparency group and soft mask methods

//----- transparency groups and soft masks                                                                                                                 
  virtual void beginTransparencyGroup(GfxState * /*state*/, double * /*bbox*/,                                                                               
                                      GfxColorSpace * /*blendingColorSpace*/,                                                                                
                                      GBool /*isolated*/, GBool /*knockout*/,                                                                                
                                      GBool /*forSoftMask*/) {}                                                                                              
  virtual void endTransparencyGroup(GfxState * /*state*/) {}                                                                                                 
  virtual void paintTransparencyGroup(GfxState * /*state*/, double * /*bbox*/) {}                                                                            
  virtual void setSoftMask(GfxState * /*state*/, double * /*bbox*/, GBool /*alpha*/,                                                                         
                           Function * /*transferFunc*/, GfxColor * /*backdropColor*/) {}                                                                     
  virtual void clearSoftMask(GfxState * /*state*/) {}

So, I guess we only have to implement these methods, it should be easy for a cairo guru :-P 

Patches are welcome!

Created attachment 10910 [details] [review]
initial implementation of transparency groups

I've no idea if this still works, or applies or anything. Cairo doesn't support luma masking, only alpha masking. This patch makes no effort to support the luma masked transparency groups.

Created attachment 10916 [details] [review]
Implement transparency groups in the cairo backend

Still really ugly, but it should mostly work. At least it does for the pdf's I've tried.

Fixed in cvs

*** Bug 11886 has been marked as a duplicate of this bug. ***

Reopening this bug, since an attachment at Gnome Bugzilla still shows transparency problems: http://bugzilla.gnome.org/show_bug.cgi?id=516829#c2. Direct link to the PDF file: http://bugzilla.gnome.org/attachment.cgi?id=105391&action=view

I've checked with poppler 0.6.4 and Evince trunk.

(In reply to comment #16)
> Reopening this bug, since an attachment at Gnome Bugzilla still shows
> transparency problems: http://bugzilla.gnome.org/show_bug.cgi?id=516829#c2.
> Direct link to the PDF file:
> http://bugzilla.gnome.org/attachment.cgi?id=105391&action=view
> 
> I've checked with poppler 0.6.4 and Evince trunk.
> 

yes, it works with splash though. With cairo the following message is printed in stdout:

matrix not invertible

In Evince bug http://bugzilla.gnome.org/show_bug.cgi?id=554131 there's another test case:

http://www.sbc.se/EPiBrowser/Dokument%20&%20Blanketter/SBC%20ek%20f%C3%B6r/SBC%20Direkt%20h%C3%A4mta/SBC_Direkt_0608.pdf

I can confirm it's only reproducible with the cairo backend with current git master. 

there are similar issues on 
https://bugs.launchpad.net/bugs/342608
https://bugs.launchpad.net/bugs/342282

could bug #17139 be a duplicate?

(In reply to comment #19)
> there are similar issues on 
> https://bugs.launchpad.net/bugs/342608
> https://bugs.launchpad.net/bugs/342282
> 
> could bug #17139 be a duplicate?
> 

the first one (as well as the document mentioned in Comment #16) has to do with blend modes (will be fixed soon) and the second one might be related to bug #17139, I'll look at it in more detail. Thanks for providing test cases. 

All documents mentioned in this report work now with current poppler (and current cairo). 

Sebastien: document attached to launchpad bug 342282 still fails to render (only page 2, page 11 looks good now) and it doesn't seem to be the same problem than bug #17139. So, could you please file a new bug report for it, please?

(In reply to comment #21)
> All documents mentioned in this report work now with current poppler (and
> current cairo). 
> 
> Sebastien: document attached to launchpad bug 342282 still fails to render
> (only page 2, page 11 looks good now) and it doesn't seem to be the same
> problem than bug #17139. So, could you please file a new bug report for it,
> please?
> 

Nevermind, I've just fixed it (45dae4e66b2b833f95656850b5f1b0ffd1d96f1d)