Skip to content

tests: use rgb_diff_stat more and consistently

Pekka Paalanen requested to merge pq/weston:mr/diffstat into main

This is another series polishing the existing color management related tests.

rgb_diff_stat gets improved reporting and debug dumping.

Then alpha-blending and color-icc-output tests are migrated to use rgb_diff_stat for their test acceptance criteria instead of the open-coded previous check. Finally they are both changed to use two-norm (Euclidean distance) error measure between reference and screenshot colors. I believe two-norm is a better error measure than just maximum as it won't completely ignore the two "less wrong" color channels on a pixel.

@vitalyp, what do you think?

I found the color-icc-output debug dumps as visualized in Octave really... interesting. Eye-brow raising if I dare say. I hadn't looked at those before.

Merge request reports