There is a slight but significant difference in Testform.v1.0.2.pdf at 72 dpi with Ghostscript r8017 vs r8016. Some of the yellow/black gradient boxes change size such that they are now wider than the black boxes immediately below them. See the attached screenshots for a zoomed in demonstration of this. The command line I used to generate the test files: bin/gs -sDEVICE=ppmraw -sOutputFile=test.ppm -r72 ./Testform.v1.0.2.pdf
Created attachment 3001 [details] r8016.png
Created attachment 3002 [details] r8017.png
Downgrade priority because (1) we think it's an unimportant raster difference, and (2) insufficient engineering resources for important bugs.
Setting P2 for regression bug.
The problem is not related to r8017. That revision occasionally madse it visible due to omitting a conversion of a clipping path to a path and then back to clipping path. While r8017, the new rendering appears worse than the old one. The reason for the worse rendering is that we apply different filling rules when filling and clipping at small resolutions. I opened bug "Bug 690036 clipping differ from Adobe" about that.
I will close this bug with wontfix because it's not a real regression of the algorithm due to Commant #5. The new general bug 690036 will store the continuation.
Created attachment 4305 [details] Simplified.pdf A simplified PDF that was used for analyzis. It is made from Testform.v1.0.2.pdf with removin g unrelated objects with Acrobat, the with unpacking streams and manual editing of one of them (a form stream).
Created attachment 4306 [details] Simplified.pdf A simplified PDF that was used for analyzis. It is made from Testform.v1.0.2.pdf with removin g unrelated objects with Acrobat, the with unpacking streams and manual editing of one of them (a form stream).
Created attachment 4307 [details] Simplified.pdf A simplified PDF that was used for analyzis. It is made from Testform.v1.0.2.pdf with removing unrelated objects with Acrobat, the with unpacking streams and manual editing of one of them (a form stream).