Fix color management issue on windows

[Dannny1]

This is just taken from git-db-usr@262b8a6
Since nobody has created PR from it before, i will just open one.

[Dannny1]

Ok, it's updated.

[TurboGit]

Almost there :) TIA.

[TurboGit]

@wpferguson : Would you be able to double check this on your Windows build before I merge? Nothing urgent as the merge windows for 5.6 will be open around Jan 19 at best.

[wpferguson]

@TurboGit I built and it works fine AFAIC see. I can't spin up a second monitor from my VM (maybe libvirt is too old, or I haven't found the correct instructions yet). I could throw the build out on pixls.us and ask for testers.

[TurboGit]

I could throw the build out on pixls.us and ask for testers.

Good idea yes. Thanks.

[Dannny1]

Before i created the PR i tested the original commit (on different system as i use linux).
It was able to get and use the profile. Even when profile was changed in OS during dt already running, after it was moved to different monitor it used correct profile. Entries from log on that computer:
_```
14.9156 [color profile] we got a new screen profile ROG XG279Q_d65_22-05-2023.icm' from the windows color profile api (size: 20412) 126.3981 [color profile] we got a new screen profile M16native_22-05-2023.icm' from the windows color profile api (size: 20552)
164.9649 [color profile] we got a new screen profile `ROG XG279Q_d65_22-05-2023.icm' from the windows color profile api (size: 20412)
256.3814 [color profile] we got a new screen profile `ROG XG279Q_user_p3_22-05-2023.icm' from the windows color profile api (size: 11984)
256.6663 [color profile] we got a new screen profile `M16native_22-05-2023.icm' from the windows color profile api (size: 20552)
258.4953 [color profile] we got a new screen profile `ROG XG279Q_user_p3_22-05-2023.icm' from the windows color profile api (size: 11984)

[TurboGit]

@Dannny1 : That's reassuring, but when a tricky feature like this is introduced and one that I cannot test myself (I'm not using Windows), I always ask for double checking.

[andrewk8]

I’m on W11 25H2. NVidia GeForce GTX 1660. 2 monitors from different companies. Using the vendor-supported Windows profiles.

Working

I don’t tend to move darktable between screens. Sometimes I have my display set to use one monitor only (e.g. Windows Show only on 2 display setting). Sometimes I work with both. With production builds, I noticed exported jpg look darker than darktable when using IrfanView (supposedly color-managed).

With this build, my exported jpg look like darktable. Using the sRGB profile, perceptual intent.

New issue 1

I also use the Windows built-in virtual desktops feature. Sometimes I switch to a different desktop if I am doing something else.

With this build, the virtual desktops will rotate back to the desktop with darktable intermittently. So far I’ve found two scenarios where this is reproducible. You can rotate to the other desktop for ~5-10 seconds before it rotates back to the desktop where darktable is.

• full screen preview from lighttable (f key)
• batch exporting from lighttable

New Issue 2

If you are in thumbnail mode and have a window partially covering darktable, darktable will raise to the front (obsuring the window you were working in) if your mouse accidentally passes into a thumbnail

Here’s a short video showing both issues

• lighttable mode
• pick an image and go full screen
• ctrl+windows+right arrow to rotate to desktop 2
• start working on something else
• after a few seconds the display rotates back to desktop 1
• exit full screen
• bring another window over darktable
• touch a thumbnail with the mouse pointer
• darktable rises to the foreground

2026-01-11.17-10-35.resize.mp4

[git-db-usr]

With this build, my exported jpg look like darktable

I assume you tested on the second screen, otherwise it would be irrelevant to the issue. Also the output profile is irrelevant as color management should take care.

I was able to reproduce those focus issues, but when i executed with debug "-d all" darktable wasn't taking the focus anymore. So this would be tricky to solve. Not that i would be capable of it anyways as i'm not an expert. From my amateur perspective tho, it seems those functions which determine the correct monitor are read-only, reading os info and should not mess with the focus. However in win32 api anything even bug may be possible (the GetICMProfile function was broken previously for months according to the info i found before Microsoft fixed it).
Anyway the monitor info detection can't be avoided if the color mgmt should work on any screen afaik.

[andrewk8]

IrfanView jpg output and darktable look the same on both monitors