[Scribus] CM problem printing with ICC profile (RGB vs. CMYK?)

[Rob Miller]

Hi,

I recently built some profiles for my printer, HP-1220C, using Argyll 
(http://web.access.net.au/argyll/argyllcms.html).  These work very well 
with Qimage, an 'ICC-aware' printing/interpolation package which runs 
under wine on linux.  Qimage is available from www.ddisoftware.com, 
including a free demo.  Unfortunately, using the profile to print images 
from Scribus doesn't work nearly so well, producing a strong blue cast.

Printing without an ICC profile from both packages produces identical 
images.  The free printer profiles I've found seem to be CMYK, while my 
profile and printer are RGB; Scribus prints successfully (bad colors) 
using e.g. the Euroscale-uncoated profile, but Qimage fails with the lcms 
error "Output profile is operating on wrong colorspace" regardless of 
CMYK/RGB printer driver -- so I haven't been able to compare results using 
another profile.  I'm using the same input profile (sRGB) and same 
rendering intent (Perceptual) for both programs, and I've tried 
with/without black point compensation and UCR.  I'm trying to get Corel 
PhotoPaint for linux to get a third view on my profiles, but it is a big 
download.

I would say that the soft-proof for both programs looks the same (no 
cast).

I'm using the HP open source HP-IJS (foomatic) RGB printer driver, and 
don't yet see a way to make it work with CMYK.  I know that there's a 
gimp-print driver which takes raw CMYK, however the HP driver with their 
'PhotoRet' technology produces much, much better image output.  When I use 
my profile with Scribus and the raw CMYK driver, there's a similar blue 
cast -- for whatever that's worth :-)

So, I'm looking for feedback, and any hints or suggestions on how I might 
improve this situation so I can use Scribus and my printer.  Is the CMYK 
printer driver a non-issue, and others are able to make Scribus print fine 
with RGB profiles to RGB printers?  Alternatively if this is the likely 
problem, is it more like a bug that might be solved with a CVS pull or me 
diving into the code, or is it an unimplemented feature that would take 
lots more work?  Maybe something else?  Any thoughts or info appreciated.

thanks for reading,

rob.