Created attachment 7598 [details] dots.png around line 33000 (+/- 1000) there is a phase shift that appears. The command line used was: gswin32c -r720 -sDEVICE=png16m -o dots.png dots.pdf The image produced is 1417 pixels wide by 141732 lines I see the first error at raster line 33249/33250 and this also occurs at 98953. The customer reports failure around 66,000 but I can't see these. The location of the phase glitch seems to vary slightly with parameters (the customer used tiff32nc and the error was in a different place). Since this happens with the TIFF device and the png16m device, this is most likely a graphics library issue, possibly related to clist use.
Asking Robin to take a look.
Created attachment 7601 [details] dots2.pdf Even simpler file. gswin32c.exe -dMaxBitmap=10000 -r180 -sDEVICE=png16m -o dots.png dots2.pdf produces a png file of 354x35433 in size, mostly blank. Scroll down to the bottom in your favourite viewer that can cope (I use ImageMagick Display), and you'll see a discontinuity across the small region of the rectangle that is still filled. This corresponds (I believe) to the join between bands 10 and 11. Going to step through the clist code now to see if I can understand where the error is coming from.
Fixed in: commit 4ed123916f3f3373d5648011ea03c3cd969e00c1 Author: Robin Watts <robin@ghostscript.com> Date: Fri Jun 17 17:11:27 2011 +0000 Fix for Bug 692285: phase shift in patterns in long image. The phase for a pattern mask was being held in a short, rather than in an int. Unsurprisingly this caused problems when the phase overflowed. Simple fix is to use an int instead. Cluster testing shows no changes.