The problem I'd expect to see, given what I've seen of the workflow and how the profile was prepared, is one of clipping. Canon raw encodes for about a stop of headroom --- the highest values in the raw file are approximately 200% brightness, not 100%. So if you're only using up to 100% exposure in your scene, then you're only using values up to 50% in the raw. And ICC profiles are usually really bad at dealing with that --- they really, really want everything to be relative to 100% white for the current profile, not relative to 50% white. If they see 50% in the raw, they will map that to grey, not white, in the output profile. That's where pnmnorm comes in: it will expand the "normal" exposure of the raw file to fill the full 16 bits of the pnm file, so that the icc transform really maps white to white. But this is completely the wrong thing to do, because highlights in the scene are going to get fully expanded in the final output, which compresses the tonal range of the rest of the image; in reality, you want to either clip or compress highlights. The default behaviour of pnmnorm also looks totally incorrect in the way it normalises each colour band separately. If the highlights aren't pure white, then that necessarily introduces a colour shift into the image. So with better normalisation/clipping, the profile there may well be usable.On his suggetion, I finally managed to get Cinepaint to compile and run, and the profile works fine there. See the rest of my notes on cinepaint