Artifacts when importing / opening old edit in current master

Hi all,

I’m quite sure there already was some link to a GitHub ticket regarding this issue, but I can’t find it. Neither here nor at GitHub. So please forgive me to raise it once again, but any help is really appreciated.

So I’m on current master, Windows 11, openCL (but it doesn’t matter, if openCL is on or off). I opened / imported an old edit the other day. Already in lighttable, the picture is totally screwed up.

Left: after import, right: after History reset
import_vs_resetHistory

Switching to darkroom I face the same situation. The white luminamce setting in filmic are weird and lead to that result:

filmic

Currently after importing the file, filmic is the last entry in the history. When I go one step before filmic down in the history, the picture returns to normal.

Any clue, any fix in sight? Of course, I will be happy to provide more information as needed to debug that, but didn’t want to push each and everything now, since I’m sure that it’s already available somewhere…

Thanks in advance for any help - bye,
Lars.

Updated the XMP, uploaded RAW file:
D75_3622.NEF (22.7 MB)
D75_3622.NEF.xmp (11.3 KB)

Hi Lars,
I had this problem happen with me. I am struggling to remember the solution, but I believe it may have been because with my current version of DT I run sigmoid as the default processing method, but my previous edit was filmic based. I believe both filmic and sigmoid were active. Take a look and see it that is the case with yours.

1 Like

Hi Terry,
thanks for your hint, I just checked on my side - my workflow default is scene referred (filmic) and Sigmoid is not a part of the history of the imported picture.
So seems to be something different…

The default values of the contrast in highlights and contrast in shadows settings (options tab) was set to safe in some version (previously, only hard and soft were available). Later, Aurélien reverted this to hard (I think that was when the latitude was set to 0). Check what your setting is. High contrast with high latitude (and/or high dynamic range) may require setting the safe option.

1 Like

Just this one photo or many…

1 Like

All pictures of this current film roll. But this is just a couple of six or so. I need to check, if more edits are affected and also if the information that @kofa mentioned applies here. I will come back with my findings.

How old was the edit… I feel like a few years ago there was a window in time where old edits broke and also maybe a certain rendition of filmic. To produce that result which you show you need a high contrast ,that weird value that popped up in display white and as kofa noted the setting for transition to shadows and highlights could also exaggerate it… But white luminance should be 100 percent not 0.17… that is a weird result

1 Like

OK, @priort the edit was made with dt 3.2.1 and in reply to @kofa => filmic settings are version 4 (2020) and contrast in highlights/shadows are both set to “hard”. Changing the contrast to soft or safe has no impact. Setting display target white luminance to 100% doesn’t change either.

But: I was able to drill it down further. Changing the contrast (being 1.0 after importing the picture) to the same value of 1.0, then everything “realigns” and the picture looks fine. So it seems to be some kind of mismatch between persisted values in the XMP and the value after reading the XMP or writing to the database?

I’m still wondering, if this is a real “widely spread” issue that may be good to sort out for the next release, or just a very special one. Still need to browse through my pictures here to see, if more of them are affected.

You should open a bug report, so the developers can check the migration code.

1 Like

I will do - thanks to both of you @kofa and @priort for helping here. I will post the link to the GitHub issue here.

1 Like

Be sure to share the xmp file.

1 Like

I’ll look back but as I said you might be in that window where the pipeline went through a couple of quick iterations and for sure at least once or twice edits were broken

1 Like

I ran into this about a week ago, opening an old edit with the “windows insider” build, but can’t find that photo now. The used “colour science” was v. 4; I made a duplicate and after some fiddling I have reset the filmic module, to get rid of the weird orange curve. Sorry, not much of a help.

Be well!

1 Like
1 Like

Just to let you know, as I’m quite happy since the abovementioned issue has recently been fixed in current master :muscle: thanks to the Dev team once again :pray:. Have a wonderful day :wave:

3 Likes