Comment 27 for bug 449255

Cyrillic capital Short I (U+0419), in fonts Arial, Courier New, and Times New Roman, as well as Cyrillic small letter Be (U+0431) in DejaVu Sans and Cyrillic small letter Ef (U+0444) in Antiqua are sooner being losing during pdf-to-postscript conversion in the CUPS pipeline.
This conversion is set up with the package install scripts to be performed by /usr/lib/cups/filter/cpdftocps script. This script uses /usr/lib/cups/filter/pdftops filter which actually use /usr/bin/pdftops poppler front end.

Modeling CUPS pipeline conversions step by step starting with OOo generated postscript file ShortI.ps (with only letter capital Short I in Arial), converting it to PDF using ps2pdf13, and finally converting it back to postscript using /usr/bin/pdftops gives result file ShortI-pdftops.ps that seems to have no Short I included.

Considering explanations from http://bugs.ghostscript.com/show_bug.cgi?id=690935 /usr/bin/pdftops rather not render so called “.notdef” glyphs that are required to be rendered in order to get letters in subject showed/printed.

If the pdf-to-postscript conversion will be performed using ghostscript with RENDERTTNOTDEF option, the result (ShortI-gsps.ps) contain the letters and can be printed successfully.

/usr/lib/cups/filter/pdftops filter is introduced on Jaunty with pdftops-cups-1.4.dpatch, and is contained in cups 1.4 originally. It seems that if either pdftops will be called from the filter somehow to render “.notdef” or conditional compilation in pdftops filter will be switched to using ghostscript and RENDERTTNOTDEF option included, the letters in subject will pass pipeline successfully.