Pixel brushes with a low density create very noticeable patterns when the mask type is set to either Gaussian or Soft It also happens with the Default mask type if the brush is small enough (less than 11 px). The pattern are the most noticeable with horizontal lines. Reproducible: Always Steps to Reproduce: 1. Choose the Ink_ballpen brush 2. Lower density to 20 % in the brush settings for better visibility of the problem 3. Draw a horizontal line (either freehand or with line tool) Actual Results: The pixels follow a pattern and don't look random at all. Expected Results: Random pixels with no noticeable pattern, like with older Krita version This is a regression in 3.0 because Krita 2.9.11 doesn't have the problem. In fact it is a quite recent regression because the First alpha does not have the problem. I tested a few 3.0 krita: - 1f5cdac (first Alpha) is OK. - f38b47e is broken.
Created attachment 99240 [details] bug demonstration
Created attachment 99241 [details] can't be reproduced on ubuntu 16.04 I can't reproduce this on ubuntu 16.04, with compiled krita git d539a80, not with several settings. I am not sure what might be going on here.
For the initial bug report, I was on Windows 10 with the RC1. I just saw the new 3.0 version (f0cbffc, is this the final version?) and just gave it a try on Windows AND on Ubuntu 16.04. I can say that: - The problem is still present on Windows 10 in f0cbffc - I can't reproduce on Ubuntu 16.04
The bug can be reproduced only on Windows and only with the prebuilt packages. My own build (without Vc) works fine. The latest Alpha2 brebuilt .zip package is broken.
Ok, I might be wrong about Vc. My local branch was at kazakov/memory-optimizations, which is a bit old. Anyway, Alpha2 package is affected.
Git commit 2c0bc2a25f793b7773af271a5d0695794b0d49d0 by Boudewijn Rempt. Committed on 09/10/2016 at 11:20. Pushed by rempt into branch 'master'. Fix the randomness of KisRandomSource The root cause probably was all the places where srand was called, but after this change, textures are random again. This was a problem on Windows only, probably because the resolution of time() is too low. But that's a guess. Related: bug 345196, bug 370246, bug 367873 Closed phabricator task T3887,T3968 M +22 -5 libs/image/brushengine/kis_random_source.cpp http://commits.kde.org/krita/2c0bc2a25f793b7773af271a5d0695794b0d49d0
I still see the problem on the 3.0.91 prebuilt package on Windows 10. The tag does include the commit (and its revert(?)). It does not seem to be fixed, despite what the phabricator tasks say (T3887 and T3968).
I'm unable to reproduce this on Windows in 4.1.1 or git master, could you retest with the latest version? Application Version: 4.2.0 KDE Plasma Version: 5.13.5 KDE Frameworks Version: 5.50.0 Qt Version: 5.11.1
I just retested with several releases. It looks like it is fixed in 4.1.1 . I can reproduce the bug with release 3.0.92, but not with 3.3.3 nor 4.1.1 Good.
Great! Let's close this then, since it appears resolved in the latest versions.