https://bugs.kde.org/show_bug.cgi?id=415320
Bug ID: 415320 Summary: Conversion from openraster to png gives different results unpredictably Product: krita Version: unspecified Platform: Kubuntu Packages OS: Linux Status: REPORTED Severity: normal Priority: NOR Component: File formats Assignee: krita-bugs-n...@kde.org Reporter: inklab...@gmail.com Target Milestone: --- Created attachment 124571 --> https://bugs.kde.org/attachment.cgi?id=124571&action=edit Example ORA file that seems to cause the issue. SUMMARY When saving a certain .ora file as a png using Krita batch mode, the output file will be slightly different seemingly randomly between two versions. Out of a sample of 100 batch mode runs, I got 56 of one png file and 44 of the other, not quite an even split but almost. In this case the ora file is two layers in "dest-in" composite mode. STEPS TO REPRODUCE 1. Place the ora file attached to a path, or make a new similar one 2. run $ krita <path> --export --export-filename <somepath>.png 3. run $ sha1sum <somepath>.png 4. Do steps 2-3 a number of times, and take note of the hash values OBSERVED RESULT The hash values are different seemingly at random. EXPECTED RESULT The hash values (and therefore the output image) should always be the same for the same input. SOFTWARE/OS VERSIONS Linux/KDE Plasma: KUbuntu 19.04 (available in About System) KDE Plasma Version: 5.15.4 Krita 4.1.7.101 from Kubuntu apt ADDITIONAL INFORMATION I could only attach one file, so here are the example of png 1 and png 2 possibilities that were exported. When loading them in Krita and comparing, it seems that one is slightly 'eroded' compared to the other. https://pasteboard.co/ILQ01nD.png https://pasteboard.co/ILPZKpO.png Here is python code I used to get stats and also save text files of the difference in pixel values between the images. It looks like there are differences in the RGB channels but not the A channel: https://www.pastiebin.com/5dfa4668c4098 -- You are receiving this mail because: You are watching all bug changes.