[cups.development] orientation-requested regression for pdffiles?

[Michael R Sweet]

Kevin Hamilton wrote:
> Jon Peatfield wrote:
>> I currently 'solve' this by having a wrapper round pdftops which removes
>> the orientation-requested option from the command-line so it does nothing
>> wrt rotations letting pstops do the work.
> 
> Thanks for that idea, I've put in a similar workaround for now, wrapping
> pdftops to restore our prior behavior.  But I still feel there is a bug
> or regression here, or at the very least, some documentation changes needed, because the behavior of orientation-requested works differently for different mime types.
> 
> Is Michael Sweet the primary developer of this system?  If so, do you have any thoughts on all this, Mike?

I am the lead developer for CUPS.

The pdftops filter has always been tricky - for years we maintained it
separately so that we could control the PostScript output and prevent
the filter from doing auto-rotation.  Later we allowed it to auto-
rotate to match the paper size, but that introduced other issues.

The current CUPS 1.4 approach is to pass the standard pdftops
utility a portrait or landscape page size (depending on the job
settings) so that we get PostScript data with the expected
orientation.  Then pstops can to the final rotation so the output
comes out right.

That said, current Xpdf and poppler use setpagedevice in the
PostScript output (bad bad bad) and that leads to problems in
orientation - the setpagedevice command specifies a landscape
page size which (sometimes) causes the page to rotate, and then
pstops applies a second rotation which totally hoses things.

Xpdf and poppler need to have a "no setpagedevice" mode that does
not produce setpagedevice commands.

-- 
______________________________________________________________________
Michael R Sweet                        Senior Printing System Engineer