I made further tests on Stretch. My current packages: $ dpkg -l | grep -e ghostscript -e cups-pdf | awk '{print $2, $3}' ghostscript 9.26a~dfsg-0+deb9u3 printer-driver-cups-pdf 2.6.1-22
Test 1: print an UTF-8 text file from Gedit. Works. Test 2: print an ODF document from LibreOffice. Works. Test 3: print an HTML document from Firefox-ESR. Works. This bug thus seems to occur rather randomly. For the record, I indeed encountered a "permission denied" error when I tried printing to PDF the other day, but I no longer remember from which application I did. I probably pressed CTRL-P in whichever application I was using at that moment. Martin-Éric ke 3. heinäk. 2019 klo 15.39 Martin-Éric Racine (martin-eric.rac...@iki.fi) kirjoitti: > > ke 3. heinäk. 2019 klo 14.51 martin f krafft (madd...@debian.org) kirjoitti: > > > > I straced gs run directly (as user "nobody"), as well as gs run from > > cupsd. After replacing all hex addresses with 0xdeadbeef, the two > > traces are nigh identical, except cupsd seems to close FDs 0–2, so > > that when run from cupsd, the first FD used is 0, whereas it's 3 > > otherwise. > > > > Apart from that, I see absolutely no difference in the syscalls > > until one of the fails to write the output file. > > > > Straces attached. Go ahead load them in vimdiff! > > Thanks. Let's see if we can figure out what causes this. > > Volker: please see the whole thread at > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=931363 > > Thanks! > Martin-Éric