The customer states and I've verified that the attached PDF file cannot be read by Ghostscript head (r12231) nor 9.01 though earlier version (e.g. 8.54) can read it (I'm searching for the responsible rev now). The command line I'm using for testing: bin/gs -sDEVICE=ppmraw -o test.ppm ./58223.PDF
Ken performed some analysis: The CCITTFaxDecode filter is supplied a parameter dictionary: /DecodeParms<</K -1/Columns 2 0 R/Rows 3 0 R>> The problem is that the Columns and Rows are object references, not values. If I replace these with actual values: /DecodeParms<</K -1/Columns 3296 /Rows 2196 >> Then the PDF file is rendered correctly (or rather, the same as Acrobat). It seems crazy to me to have these as references, but the PDF Reference doesn't seem to say there's any problem with that, and Acrobat opens the file, so I suppose we'll have to support it.
Adobe Acrobat 9.42, Apple Preview 5.0.3, and muPDF all read the file without problem.
This file broke in r11840.
Created attachment 7318 [details] patch Add missing test for /packedarraytype during recursive dereferencing of composite PDF objects.
The patch has been committed as a rev. 12237.