I made a pdf file in inkscape (a solid fill + a horizontal gradient - one that should compress well with DeltaRow but not RLE) for another bug, and this pdf: 1) crashes the X11alpha device 2) rendered wrongly with X11 and ppmraw/png16m (strange scaling/clipping) both for 8.70 and a slightly dated svn-head.
Created attachment 5545 [details] pdf file from inkscape Here is the pdf. xpdf shows blank, while evince (libpoppler-based) shows it correctly (similiar to acroread), which is a bit strange considering that libpoppler forked from xpdf.
libpoppler forked from xpdf a long time ago, and xpdf hasn't seen much development since then.
Created attachment 5553 [details] pdf generated by evince from the older pdf another pdf generated by printing the old one to pdf by evine, which has a slightly different effect - -sDEVICE=x11 and png16m shows as the inkspace pdf (i.e. enlarged and clipped) ; but -sDEVICE=x11alpha shows blank pages.
No crashing now with rev 12380, but some very odd clipping going on. Passing this off to Alex to look at.
Created attachment 7475 [details] Decompressed and modified sample file Uncomment the following line in the file to avoid the clipping effects caused by ignoring the line width during calculation of the actual tile size of a transparent pattern. % 0 0 m 600 800 l % <---- Uncomment this line to avoid clipping
The problem first appears in the trunk with the merge of a branch (see f1b79a336fc7dab3864ae00d45cad2e5989c0fb2 committed 2009-07-22) that re-implements support of patterns with transparency. Apparently, the stroking width is not taken into the account during calculation of the tile size, which causes clipping. Reassigning to the author of the patch.
Works fine in 9.20