I am using gs 8.71 on Fedora 8 Linux. I have a pdf file with shading that some pdf viewers don't handle right. gs displays it correctly, but when I run it through ps2pdf and view the result with gs, the "customer appreciation" banner at the top is a white hole. The ps2pdf output looks correct with xpdf, so the problem might be viewing in gs rather than the pdf writer. William williambader@hotmail.com
Created attachment 6130 [details] the original pdf
Created attachment 6131 [details] output of ps2pdf with the missing banner when viewed with gs
Acrobat 9 doesn't display the 'text' either. I doubt this is anything to do with shadings, the text appears to be image data overlaid on a white softmask and is therefore most likely a transparency issue. IIRC xpdf has problems with transparency, so I suspect that it is xpdf in error, not GS. Which means the problem is something to do with pdfwrite. A simpler sample file would help if possible, it will take some time to reduce this one to a workable size.
I'm sorry that I don't have a simpler file. I was sent this file in May 2009 because it didn't work in an application, and the problem was with how pdftops from xpdf handled shading. I just tested with pdftops from a poppler snapshot from Jan 12, 2010, and the ps generated from either of the pdf files looks correct in gs. I have been testing pdf compression tools like http://code.google.com/p/pdfsizeopt/ I gathered a bunch of old test files, and I noticed the problem with this file.
This appears to have been fixed sometime in the last 6 months, almost certainly as a result of Michael's hard work on transparency. I'm afraid I don't have a revision number where it was resolved, but I believe the 9.0 release should work correctly with this file.
I confirmed that it works with gs9.00. The only problem is that if I run the resulting pdf through pdftops from poppler, I get a bunch of messages "Error: Illegal entry in bfrange block in ToUnicode CMap" because a character table has 16 bits but has ranges like <45> <46>, while poppler expects that a 16 bit range should have numbers with four digits like <0045> <0046>. The test is around line 300 of poppler/CharCodeToUnicode.cc. I suspect that gs is right, and poppler needs a small change. William