pdftopdf doesn't perform software collate

Pages printed are not collated when I print a document with multiple copies set and Collate option enabled on a printer without hardware collate support.

Printer is Samsung ML-2950.

Operating systems affected are: Linux Mint 16, Ubuntu 13.04

CUPS filters pipe for printing a test text file is the following:

D [09/Dec/2013:14:34:00 +0400] [Job 16] texttopdf (text/plain to application/pdf, cost 32)
D [09/Dec/2013:14:34:00 +0400] [Job 16] pdftopdf (application/pdf to application/vnd.cups-pdf, cost 66)
D [09/Dec/2013:14:34:00 +0400] [Job 16] gstoraster (application/vnd.cups-pdf to application/vnd.cups-raster, cost 99)
D [09/Dec/2013:14:34:00 +0400] [Job 16] rastertospl (application/vnd.cups-raster to printer/test, cost 0)

rastertospl is proprietary filter, which converts CUPS raster data to Samsung's SPL PDL.

PPD file is attached.

Further investigation has shown, that pdftopdf filter does not perform software document copy generation with my PPD file set in environment variable PPD, its output contains a single copy, but resulting CUPS raster contains information about multiple copies and rastertospl generates SPL output with mutiple copies for each page (uncollated).

pstops filter, started with the same CUPS job options set and multi-page PostScript input, generates PostScript output with properly collated pages.

pdftopdf has been executed with following parameters:

/usr/lib/cups/filter/pdftopdf 1 user "test" 2 "Collate" texttopdf.pdf > pdftopdf.pdf

When I start pdftopdf with the same parameters, but with PPD environment variable unset, it generates PDF file with document copies created.

When I modify PPD file by changing *cupsManualCopy attribute value to True, pdftopdf produces collated output.

This printer supports hardware page copies, but does not support collate in hardware.

It seems that pdftopdf filter logic used to determine collate method does not handle this case correctly.