Comment 7 for bug 370591

I could reproduce the bug now.

It is not a CUPS problem. If you send a multiple-page PostScript or PDF file directly to CUPS (with the "lp" or "lpr" command) the printout is always correct.

The problem is in Evince or even in the GTK printing framework. If the user checks "Reverse" on the first tab of the printing dialog, the PostScript data sent to CUPS are all pages in the reverse order. This means the 3rd page is sent at first, then the 2nd, and in the end the first. So CUPS prints the first two pages it gets onto the first sheet, page 3 in fron and page 2 on the back. Then it prints page 1 onto the front page of the second sheet and leaves the back side blank.

So Evince/GTK should as the best solution simple always send the pages in the original order and add an IPP attribute for the reverse printing. So CUPS can arrange it correctly with its pdftopdf or pstops filter.

Alternatively, Evince/GTK can add a blank page to the end of a job with an odd number of pages and duplex set, and reverse the page order only after that, but note that it gets complicated if N-up comes to the place. So my recommendation is to leave all page management to CUPS and so simply send the original document with all page management option settings (options on the first two tabs of the dialog) sent as IPP attributes to CUPS.

If the same problem occurs for OpenOffice.org it should be fixed the same way.