The Ghent Output Test Suite v3.0 is freely available from http://www.gwg.org/ghentoutputsuite.phtml. It contains a set of PDF samples specifically designed to expose rendering and/or conversion problems of PDF processing software. Tests included are for knockouts, overprints, font substitution, shadings, output intents, deviceN, etc. Basically I used a commandline of gswin32c.exe ^ -o out.pdf ^ -sDEVICE=pdfwrite ^ -dPDFSETTINGS=/prepress ^ in.pdf to test each of the files included in the Ghent suite. Then I compared if Acroreader rendered out.pdf differently from in.pdf. Current Ghostscript fails on most of these tests. (The guilty component seems to be pdfwrite, since a quick test with 3 files running against DEVICE=tiffsep didn't show up the same problems). If this bug report doesn't serve any other practical purpose, at least you can now decide if you will include the Ghent Test Suite into your routine regression testing activities... :-) [I filed the same bug report against MuPDF as #691702)
(In reply to comment #0) > Basically I used a commandline of > > gswin32c.exe ^ > -o out.pdf ^ > -sDEVICE=pdfwrite ^ > -dPDFSETTINGS=/prepress ^ > in.pdf > > to test each of the files included in the Ghent suite. Then I compared if > Acroreader rendered out.pdf differently from in.pdf. > > Current Ghostscript fails on most of these tests. This is too broad. Which files 'fail' and how do they 'fail' ? Lumping the whole test suite together isn't helpful. using -dPDFSETTINGS also isn't helpful as this alters a large number of the defaults, it would be much more helpful to identify which individual configuration causes a problem. > If this bug report doesn't serve any other practical purpose, at least you > can > now decide if you will include the Ghent Test Suite into your routine > regression testing activities... :-) They have been part of our regression testing for some considerable time. Note that this is regression testing, it simply tests whether the behaviour has changed, not whether it is correct.
OK I'm closing this as 'invalid', not because there definitely isn't a problem, but because the description is too vague. If someone wants to open a bug report or two with specific examples of the problem(s) I'll look at them. However, the Ghent files are specifically intended to test overprint and transparency, and *any* change in colour space will result in the output being 'wrong' as overprint only works for CMYK and transparency can depend on the ProcessColorModel of the device.