Excerpts from Thomas Freitag's message of lun mar 19 08:05:18 +0100 2012: > Am 16.03.2012 16:08, schrieb Carlos Garcia Campos:
> > The xpdf303 merge introduced some regressions in the cairo backend, > > I'll try to find some time this weekend to send more details to the list. > Seems as if You haven't found the time to take a deeper look into > > it. Yes, sorry. I wanted to run git bisect to find the commits that introduced the regressions, but there are several revisions where the build is broken, which makes a bit more difficult to run git bisect. I haven't still had time to run it, so I'll just give you the pdf files failing. > So > here just hint: if it comes from the part I merged, I see two possibilities: > 1. Radial shading > I wasn't able to merge the changes in the xpdf algorithm in the Gfx area > for radial shading with our algorithm. I took over the xpdf algorithm in > the hope that it fixes some regressions in the cairo backend and made > only a few tests. These tests showed neither an enhancement nor a new > regression, but I let the xpdf algorithm in. If this change causes a > regression in the cairo backend with PDF I haven't tested, we can revert > that change: No one of our backends uses the Gfx algorithm anymore > beside the cairo backend, all other backends implements the radial > shading by itself. Yes, that's seems to be the case of duck and roses, ClarityOCGs and 3e2247elecciones_01p You already have ducks and roses, ClarityOCGs is in http://acroeng.adobe.com/Test_Files/OCG/ and 3e2247elecciones_01p is attached. > 2. Text rendering, especially output text in pattern colorspace > Rendering text in pattern colorspace was now also introduced by xpdf, > and it was a much cleaner and more understandable solution than mine. > But therefore I needed to change every backend to support the new > solution. It was quite easy for me to change it in the postscript and > the splash backend where I understand what these backends are doing. I > thought that I also understood what I need to change in the cairo > backend and did it. I made a complete regtest in the spash backend, but > only a few tests in postscript and cairo backend to see if my changes > work at all, but of course it's possible that there were some > misunderstanding from me especially in the cairo backend. Not sure this is the case, but there are several regressions related to text rendering: pdf_commenting_new.pdf, pdf_drawing_new.pdf and pdf_filler_new.pdf use a different font. You can just google for these files. prosper-tour.pdf renders text in itallic in page 8. You can also google for this file. microtype.pdf, incorrectly places text, you can find this one in http://acroeng.adobe.com/Test_Files/OCG/ too. > Perhaps this helps... You can use poppler-regtest using revision 92ce79 to create the reference images, and current git master to run the tests and see the differences. Thanks, -- Carlos Garcia Campos PGP key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x523E6462
signature.asc
Description: PGP signature
_______________________________________________ poppler mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/poppler
