New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Artifacts when importing / opening old edit in current master #15489
Comments
No idea but applying your side car to a variety of my images leads to no weird issues ...setting seem intact... maybe copy those images to a new location with the sidecars... then remove them from the db and reimport...any issue... it might have just been a database hiccup at some point along the way.... |
That's quite strange. |
I also tried your image and it looks beautiful. believe you should not have added Tumbleweed openSUSE 20231023 nice image |
While importing from V4 the spline correction is not good. That's "deep inside filmicrgb dev history", i am currently not sure how we can handle that correctly. |
Sorry for not replying earlier to your suggestions. But now I have some more time to come back to this. Thanks for pointing me there @todd-prior . I just tried with --configdir and a non-existing path. Well, everything was initialized, but unfortunately the import shows exactly the same issue. Meeeee... So I just kept on playing and had a closer look at your screenshot (thanks for sharing,). I noticed that after re-applying just the contrast value of "1", the filmic "look only" spline looked quite different on my system from what your screenshot shows. This was also reflected in the overall look when comparing your screenshot to my dt darkroom view. So I adjusted my filmic settings to match my spline with yours. And - hm - there are indeed some interesting value changes: Please ignore the highlight reconstruction, which I simply turned off without having any influence. But the rest is perhaps interesting
That all still sounds to me like a persistence / reading issue. At the same time it seems as if under Linux there is no issue at all with this picture (am I right that you are also running dt on Linux as @ptilopteri does?) So I really have no clue what might be a Windows specific problem here. But another user in the pixls-thread mentioned, that he also experienced the same issue with one of the "Windows Insider" test-builds one / two weeks ago. His picture had also filmic v4 applied. |
This issue has been marked as stale due to inactivity for the last 60 days. It will be automatically closed in 300 days if no update occurs. Please check if the master branch has fixed it and report again or close the issue. |
I tried again with current master [0c2ad29] to please the github-actions bot 😁 and as I expected, the issue is still there. @jenshannoschwalm, I was not sure, if you already had the opportunity to read my last comment on this in which I have tried to summarize my findings after successfully restoring the filmic RGB settings manually after the faulty import. Best regards and wishes for the advancing New Year 😃 |
I don't know the filmicrgb code well enough to understand how parameters would have to be adapted in some module version change. |
I was still trying to gather some more supporting information on this effect to help narrowing things down a little bit more... So I tried latest Ansel nightly [4d185b4] today and was indeed able to import the picture on my system without any of the above mentioned artifacts resulting after import 😃 Hopefully, these findings may help you... |
That would point to Filmic's |
Think so too but couldn't spot the problem... |
Thanks a lot for testing also on your side @TurboGit. Did you use Linux or Windows for the test? I'm wondering, since it seems, as if Linux works fine, whereas on Windows me and another user (from a pixls.us discussion) can see this effect. |
@Macchiato17 : I'm using Linux. Hard to believe that this could be a Windows only issue... but everything is possible. |
I added the "reproduced" label some time ago and as @TurboGit couldn't reproduce i just checked again on this mornings master and can still confirm the issue by
and the logs from
|
@jenshannoschwalm : Seeing the error could this be a |
I have tested with English locale instead of French and I have:
So not problem with FilmicRGB, something really fishy here. |
Yes, i checked this too. After compressing history it goes away, we basically check parameters for every history step i think. |
Well i can confirm the issue can not be reproduced if importing the original xmp with "german" being the dt language. VERY stangely - after switching back to english i can not confirm :-( |
maybe something related to #16044 where the aperture filter didn't work because of locale issues (comma vs dot in floats)? |
Fix part of issue reported in darktable-org#15489.
For lens see #16079 |
@Macchiato17 : Can you still reproduce with 4.6.0 (or using branch darktable-4.6.x) or master ? |
First of all, thank you all for your great commitment. I really appreciate it. So I built dt on master [420bc45] and was unfortunately still able to see this issue. Following your discussion, I did a There are several occurrences of an error related to
I'm on a German Windows 11, dt GUI is set to English. But even when changing dt GUI to German, this issue still resist. |
Since you are building from source, can you add this patch and report what is displayed for the wrong value?
TIA. |
So we get this output now:
Hope that helps 🙂 have a nice day👋 |
Somehow it helps, the value 0.0000 is not a correct one as latitude (and this since first implementation of FilmicRGB) should be between 0.01 and 99.0. So somehow the params is wrong, yet the float value is 'sane' (not inf or nan) and so looks like a real value not some corrupted one. But I have no idea on how 0.0 could have been set here, and so maybe some corruption after all... As you see, it helps just a little bit :) Side question, if you set latitude to 0.01 manually does it fixes the FilmicRGB curve and image look? What I really do not understand is why I don't see this on my side! |
I tried on a second computer with CLang instead of GCC and using a Release build instead of a Debug one... And still cannot reproduce :( |
Another try, can you test with this change:
It is just changing CLAMP lower value to |
I checked filmicrgb code 3.0: we didn't have any default,min,max settings via introspection and there was no clamping for safe data, so some direct input or magic bauhaus stuff might have inserted a zero? |
@TurboGit after adding your patch for
Don't know, if this will be of any help - I added a PDF file that shows the resulting settings in filmic RGB directly after the import in Ansel (abovementioned version) and in darktable (with both patches applied) |
The problems is definitely in |
Well, to show variable values, I compiled as Debug, and the problem disappeared.... I did notice an unrelated error in the v3->v6 conversion: |
This issue has been marked as stale due to inactivity for the last 60 days. It will be automatically closed in 300 days if no update occurs. Please check if the master branch has fixed it and report again or close the issue. |
Describe the bug
see also:
https://discuss.pixls.us/t/artifacts-when-importing-opening-old-edit-in-current-master/40097
After opening / importing a darktable 3.2.1 edit, the picture is totally screwed up already in lighttable (see Screenshot 1 below). Same for darkroom.
Things I noted:
Finally: leaving everything as is after importing the picture and then changing the contrast (being 1.0 after importing) to the same value of 1.0, then everything “realigns” and the picture looks fine.
May this be some kind of mismatch between persisted values in the XMP and the value after reading the XMP (see attached file) or writing to the database?
Steps to reproduce
see description
if this can just be debugged with the RAW file available, please drop a note and I will add it to this ticket.
=> added together with the related XMP
Expected behavior
No response
Logfile | Screenshot | Screencast
Screenshot 1
Screenshot 2
RAW file
XMP file
Commit
No response
Where did you install darktable from?
self compiled
darktable version
darktable-4.5.0+958~ga37b7fb789-win64
What OS are you using?
Windows
What is the version of your OS?
Windows 11
Describe your system?
Laptop 8GB RAM
NVIDIA GeForce GTX 1050 Max-Q (4GB)
Intel(R) Core(TM) i5-8265U CPU @ 1.60GHz
Are you using OpenCL GPU in darktable?
Yes
If yes, what is the GPU card and driver?
NVIDIA GeForce GTX 1050 Max-Q (4GB), driver 31.0.15.3699 (04.08.2023)
Please provide additional context if applicable. You can attach files too, but might need to rename to .txt or .zip
No response
The text was updated successfully, but these errors were encountered: