I noticed the working color profile for the raw file is set to lin rec 709. Using this, I also get the different colors with 4.1.0+215. After setting it to lin rec 2020, the problem is gone.
2 Likes
@apostel338 Nice catch - at no point have i changed that working profile - so i’m wondering where it has come from - the JPG output is matching and also has lost the messy gradients.
I’ll have a play to see if its the profile, style or something else changing it.
Thanks
Steve
@apostel338 @priort @danny @kofa @snibgo @g-man
It appears that apostel338 has hit the nail on the head and found the reason I just now need to find the cause of the profile changing.
If I import with the fuji styles script off, the working profile is linear Rec2020 RGB, if i have the script on it changes it to 709. I now need to work out how to change this or ask @bastibe if his script alters the working profile.
I hope I havent wasted your time, I really appreciate your work, efforts and excellent attitude towards newcomers and most of all your help.
Steve
1 Like
As far as I can tell, this has nothing to do with my scripts. They do not touch the working profile, and indeed the input profile is always linear Rec2020 RGB on my computer.
(Edit: Working profile, not input profile)
Hi @bastibe
Good to know your script doesn’t change anything, so I just need to work out what’s going on on import. If I import with the script off, it’s fine, script on and the working profile changes. I’ll investigate further but if anyone has any hints where to start it would be appreciated. I have not played with styles or profiles before, so a bit of learning ahead…
Steve
Wild guess…any chance you have accidentally modified the provia style. Since the script seems to just do matching and apply. What if you just apply the style yourself does it change?? If so maybe you had been playing with input profiles and then also mangaged to save over the style?? No idea but worth a try… Does it happen if you shoot another mode and it matches a different style??
I seem to remember that issues like this could happen if you processed JPEGs. Perhaps if you copied settings from a JPEG to a raw file, or something like that. Or maybe I’m misremembering.
Why don’t you use a profile specific to the camera?
Sorry, typo. Working profile is linear rec2020, not input profile.
Thanks, I feel much better now… 
I was about to go and try it; I’ve done some Nikon-Nikon switching, no surprise there, most of their recent cameras have almost identical spectral sensitivity. Pondering a question at DPReview today, I opened a Canon R5 raw and assigned my Z 6 profile; the color barely shifted hue, but there were some noticeable differences in tone at the low end. Puts real significance in the term “compromise matrix”…
Maybe @bastibe means working profile if not this would not be the best as you say… you can even do a quick test with the profile evaluation in CC module as a rough test of how far off that would be…
Ha, just tried it, linear Rec2020 as the input profile to my flower test shot. Pulled everything reddish; not good.
Well i’m fairly sure I havent changed any profiles or done anything “stupid” by my standards 
Certainly have not processed any JPG’s… or updates styles as I dont know how! Although that opens me up to having accidentally done it - but fairly sure I updated the styles shortly before holiday and hadnt touched DT until now.
So I removed all the images from DT and tried to reimport the suspect one. This time it imported with no problem. I wonder as I have been trying a number of softwares, if one of them had dumped a readable XMP file and it contained the wrong working profile… other images seem to be importing fine now… its very wierd that its now working…
Steve
I was taking some test shots with my Pixel phone and my Spyderchecker 24 the other day. I took one in direct sunlight around the noon hour well maybe later…would have to check the time stamp …never mind and in the full shade and a couple of partial shade/sun shots in and around the garden near object that I could revisit for color assessment . I was going to use them to try and make some rgbCC presets in DT to see how effective it might be. Just as a quick test I thought since I made them I would just swap out the input profile and re-run the calibration.
The funny thing that for this exercise is that while the use of the rec2020 profile as an input initially gives a higher deltaE, for my phone its not too bad from the numbers that the module calculates. And the interesting thing is that running the CC color calibration gives a better color match in terms of delta E and patches marked as poor ie by the X with rec2020 vs the standard profile from the DNG…
Basically I set the initial image to modern WB, no filmic. Ran CC colorchecker , cycled through the options . Neutral or average deltaE were usually the best. I applied the suggested EV to make the correction match and then noted the values. Then I went and changed the input profile to rec2020 and re-ran the calculation using the same settings so that the only difference was the input profile…
I am not suggesting this as a course of action…just an interesting observation… might be interesting to see for other cameras… might check it out with some samples just for giggles one day…
1 Like
Well at least the mystery is revealed and likely you learned a thing or two during the hunt…
I definitly learned a number of things… but i’d still like to get down to the bottom of why… i’m suspecting C1, on1 or Exposure x7 possibly wrote an XMP… but fairly sure the DT uses a differently named XMP to the others. I cant understand why i could reproduce it, and now cant.
Steve
I’ve definitely seen that once! Darktable imported a C1 XMP, which included a weird working profile.
I think it is I deleted all the XMP files in the folder and DT created an XMP the same as the file name - previously it was filename.raf.xmp I assume because one was already there the same.
Annnyway… it proved a good discussion and learning exercise. I need to remember that most raw processors use XMP or external files… as I’m so used to LR keeping things tidy like that.
Steve
You can also turn off darktable’s XMP generation (it will still write edits to the database).
https://docs.darktable.org/usermanual/4.0/en/preferences-settings/storage/#xmp