darktable Update Can't Read File Properly

I am fairly inexperienced and infrequent user of darktable and today noticed I was using an old version (3.0.4). I decided to update to the newest version (3.6.1). When I opened the updated software, a file I was recently working on is now 100% multi-color noise. Going through the history stack the problem seems to originate at an exposure adjustment. Before the update I had viewed this file with no problems.

Any ideas what went wrong or how to fix it?

image

Hi @CrossthreadLoctite,

A swift idea: make a copy of the troublesome RAW file.
Start from history == empty.
Make a few adjustments.
How does it look?

Have fun!
Claes in Lund, Sweden

Thanks for your quick reply! The RAW file itself looks good and I can make edits to it no problem. The issue seems to be with the new version of darktable displaying the edits made in the old version. I actually found another file that had more extensive editing done to it that appears the same way, though this second file seems to have trouble at the second instance of color balance in the history stack compared to exposure on the first file.

Actually, I am not surprised: going from
a very old version to a relatively recent one
“ought” to create problems.

If you like, you could upload the RAW file and
its companion .xmp, so that someone perhaps
could give some advice…

Also: what operating system are you using?

Have fun!
Claes in Lund, Sweden

I’m not entirely surprised either. Luckily, I have .jpgs saved of the edits I had made previously so its not a huge deal but I’d like to be able to access the full edits on the RAW file if possible.

I’ve attached the RAW file and .xmp. I have Windows 10 version 20H2 installed on this computer.

DSC03250.ARW (19.8 MB)
DSC03250.ARW.xmp (201.2 KB)

In the past I had this happen a few times…I cannot recall the exact version or sequence but it was always when I had multiple module instances. It seems some time they are not correctly handled moving to newer versions…In your case you found it…seems to be in that exposure instance…if you do a selective copy with shift ctrl c and exclude exp 1 and exp 2 and then paste to a new duplicate original…the edit seems fine…I used to use multiple local contrast modules in the past more so than now…and I recall this would cause an issue…

Given how much darktable tries to handle older edits
(to the point of keeping deprecated modules only for
that purpose ) I’m a bit surprised that an edit from the
same major version causes such problems…

One thing I noticed in the history stack was a number of
modules with duplicate names (e.g. twice “exposure 1”,
and there are more like that).

I will try the selective copy and see how it works out. Thank you @priort!

The duplicate names are mostly because I’m very inexperienced and have a tendency to play around with the modules quite a bit, sometimes going back to modules I’ve already used. I’m sure that doesn’t help things either.

For me I recall now it was the old channel mixer….and esp if I had 2 instances but disabled the first in the course of an edit but kept the second one…that is going back a long way now….there was a window where a lot of work was done on versions of xmp files etc and I think you have gone back that far…I have been using versions that are no more than 2 weeks old for the last year and touch wood no major problems……

I have versioning set up to backup my xmp files so I can go back to any point in time…this is not DT but via a third party solution….

You will enjoy the new version …if you recognize it…. :blush:

It’s not the presence of duplicated names in itself, that can happen easily enough. What is unusual is duplicates following one another directly in the history stack. Normally, there should be at least one other module between two copies.
So that looks like there’s a problem in reading or restoring the history stack from the XMP file…

Another hint something is wrong appears when compressing the history stack (at least on resetting, it empties completely)

If this has worked before and isn’t now, it’s probably a regression. You can report issue here: