https://bugs.kde.org/show_bug.cgi?id=442191
Dmitry Kazakov <dimul...@gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |FIXED
Latest Commit| |https://invent.kde.org/grap
| |hics/krita/commit/4d849b266
| |ab32593b25ad1a43d32b2b13236
| |1dae
Status|CONFIRMED |RESOLVED
--- Comment #5 from Dmitry Kazakov <dimul...@gmail.com> ---
Git commit 4d849b266ab32593b25ad1a43d32b2b132361dae by Dmitry Kazakov.
Committed on 24/03/2022 at 13:56.
Pushed by dkazakov into branch 'master'.
Fix bright halos in the transform tool around semi-transparent objects
We had a weird and very old (from 2008) non-linearity in KoMixColorsOp,
which clamped the totalAlpha channel in case it is "not sane". I have
a feeling that this construct was used to just make clamp the alpha
channel value into the destination color space bit depth. Though it
didn't take into accound that with some filters (bicubic, lanczos)
the weight may be negative, basically making this check invalid.
The patch simply removes this change and clamps the alpha channel
separately (since we do already have the same clamping for other
channels).
WARNING: think twice before porting that into 5.0 branch! The patch
changes how KoMixColorsOp works, so it might lead to unexpected
consequences.
M +60 -0 libs/image/tests/kis_filter_weights_applicator_test.cpp
M +2 -0 libs/image/tests/kis_filter_weights_applicator_test.h
M +9 -8 libs/pigment/KoMixColorsOpImpl.h
https://invent.kde.org/graphics/krita/commit/4d849b266ab32593b25ad1a43d32b2b132361dae
--
You are receiving this mail because:
You are watching all bug changes.
