The file Bug687290.ps from the nightly regression test is causing Ghostscript to segfault for the normal raster test in both banded and unbanded modes for pkmraw and r72. Here is the commandline to reproduce: gs -dQUIET -dNOPAUSE -dBATCH -K1000000 -r72 -dMaxBitmap=10000 -sDEVICE=pkmraw -sOutputFile='Bug687290.ps.pkmraw.72.1' -dNOOUTERSAVE -dJOBSERVER -c false 0 startjob pop - < '/home/regression/comparefiles/Bug687290.ps'
This file runs sometimes and not others. I determined that I can get a repeatable failure with a DEBUG build, using the -Z@\$\? option. This appears to be UMR related to the 'sbits' array having different elements initialized in set_ht_colors that are used in set_color_ht (great names, huh?) Elevating to 8.50 release critical since UMR's that cause segfaults are *NOT GOOD*
*** Bug 687774 has been marked as a duplicate of this bug. ***
Copying from Bug 687774 : Running "pkmraw -r72 Bug687290.ps" on Windows got a crash in gx_render_ht_1_level due to b_level = -131058, porder->levels[b_level] runs out the array range.
Created attachment 1046 [details] Patch to resolve umr's See code-review post for more discussion: http://www.ghostscript.com/pipermail/gs-code-review/2004-November/004695.html
Fix committed: http://www.ghostscript.com/pipermail/gs-cvs/2004-November/005030.html
See bug #687805 for followup on where the copying of color procs when making a targeted null device should happen.