Bug #11667

liquify produces artifacts if OpenCL is used

Added by Alexander Rabtchevich over 2 years ago. Updated over 2 years ago.

Target version:
Start date:
Due date:
% Done:


Estimated time:
Affected Version:
git master branch
other GNU/Linux
hardware architecture:


The liquify module produces thin contours at the edges of its affected area both at viewing (at some scales) and exporting. If OpenCL is turned off, the effect doesn't exist. The arrow is smaller than the modification area.

DSC01244.ARW (23.8 MB) DSC01244.ARW Alexander Rabtchevich, 07/08/2017 01:35 PM
DSC01244.ARW.xmp (19.5 KB) DSC01244.ARW.xmp Alexander Rabtchevich, 07/08/2017 01:36 PM


#1 Updated by Pascal Obry over 2 years ago

I think I see this issue. But for me there is nothing wrong when exporting and activating/deactivating the liquify module or moving the image on the canvas makes the little artifacts disappear! Is that also the case on your side?

#2 Updated by Pascal Obry over 2 years ago

  • % Done changed from 0 to 10
  • Assignee set to Pascal Obry
  • Status changed from New to Confirmed

One question, what is the interpolation you're using (set in the preferences)? I found that when set to bilinear I always have the artifacts with the display. In bicubic and lanczos2 never. In lanczos3 I have from time to time but very rarely. In all cases I never have the artifacts in the export. What was you export setting (jpeg? quality?...) ?

#3 Updated by Alexander Rabtchevich over 2 years ago

Hello, Pascal. Here are the settings:

High quality interpolation at export: always
Interpolation method: Lancros3
jpeg 90% quality, 1600 pixels x 1600 pixels image size.

It's strange, but during scaling image view in darkroom mode, the viewing artifacts can appear or disappear depending on scale. And now I tried to export the image with the same settings and the resulting image has no artifacts :(. So I'll try to reproduce it again.

#4 Updated by Pascal Obry over 2 years ago

Thanks for the data! The bad news is that this issue on my side is playing with me. It is in fact quite difficult to reproduce systematically. I didn't find a pattern to reproduce, and today it seems that it is reproducible with the same probability with all interpolation.

As you find out it is quite difficult to reproduce. It would be nice to have a systematic way to reproduce to work on a patch. Let me know if you find such procedure to reproduce... Thanks.

Also available in: Atom PDF

Go to top