Starting with r11723 the following command line seg faults: ./main/obj/pcl6 -sOutputFile=test.pcl -sDEVICE=ppmraw -r600 -dNOPAUSE -dBATCH ./mlc0495.pcl
It's not entirely clear how the change I made can have caused this, but it does indeed appear to have done so. Valgrind shows only 1 uninitialised read, that I will open details of in a new bug. My working hypothesis is that if I set the padding bits to 0, the compression given for the bitmap is better, so the clist ends up a different size, and this triggers an unrelated bug elsewhere. I'll test this shortly.
I've opened a bug (691665) with the valgrind problem and a workaround. Oddly, and for no reason I can understand, the workaround for that bug does seem to solve this SEGV too. On 64bit linux the file now runs to completion with no errors. On 32bit linux it exits with interpreter error code -12. I'll wait to hear what Henry says about 691665 before continuing.
Fixed by the fix for bug 691665.