Created attachment 148927 [details] gif krita-nightly-x64-5.1.0-prealpha-d59747b8ad 1.As shown in the figure, if no selection is used, the preview windows will change to a small rectangular area when zooming in and out when the filter is turned on. 2.When the filter is turned on, the preview always has a general "pixelated" effect. If we zoom in and zoom out, it will become normal.
Created attachment 148928 [details] bug2
Confirming with G'MIC 3.1.2. The Preview widget resizes unstably.
Waiting for upstream resolution.
Interesting bug! It stretches to the introduction of the G'MIC-Qt plugin itself, back in 3.0.3: https://invent.kde.org/graphics/krita/-/commit/463213b8e04311f5c7f7c41e24cc198c5a98fb87 What is happening is that the image size command never accounted, in our side, for the layer mode; so we returned the image size each time. This is incorrect whenever a smaller layer is active, so in these cases the preview widget would have inconsistent behaviour due to the mismatching sizes. I also ported this behaviour blindly when I wrote the QMic interface, so this will require an ABI break. I'll add a version definition to the header to help things in the other side, and roll this fix up together with the translations.
A possibly relevant merge request was started @ https://invent.kde.org/graphics/krita/-/merge_requests/1459
Git commit 1dc834c1129ca816a53819b88c1390a6f9524b41 by L. E. Segovia. Committed on 26/05/2022 at 13:29. Pushed by lsegovia into branch 'master'. Fix image size calculation Ever since it was implemented in 463213b8e04311f5c7f7c41e24cc198c5a98fb87, the gmic_qt_get_image_size command never accounted for the selected mode. Instead, the image's size was always returned, which made the preview widget's calculations use wrong data. This fix breaks QMic's ABI, so I am also adding a interface version macro to help detection on the other side. Fixes c-koi/gmic-qt#145 M +78 -2 plugins/extensions/qmic/kis_qmic_interface.cpp M +3 -1 plugins/extensions/qmic/kis_qmic_interface.h https://invent.kde.org/graphics/krita/commit/1dc834c1129ca816a53819b88c1390a6f9524b41
Git commit 73fbcfb137b7a88e5b179fae7da9e4f238a8b109 by L. E. Segovia. Committed on 26/05/2022 at 13:29. Pushed by lsegovia into branch 'master'. Update G'MIC to 3.1.2.2 Related: bug 453945, bug 453949, bug 454100, bug 454286 M +2 -2 3rdparty_plugins/ext_gmic/CMakeLists.txt [INFRASTRUCTURE] https://invent.kde.org/graphics/krita/commit/73fbcfb137b7a88e5b179fae7da9e4f238a8b109
I don't think this problem has been completely solved. Once the "input layers" are changed, it will reappear
Created attachment 149278 [details] gif
A possibly relevant merge request was started @ https://invent.kde.org/graphics/krita/-/merge_requests/1464
Git commit ca0d473bcd570f0332458b296efce25bf6344631 by L. E. Segovia. Committed on 02/06/2022 at 16:18. Pushed by lsegovia into branch 'master'. G'MIC: clean up layer retrieval and properly detect visible paint layers Related: bug 454516 M +0 -1 plugins/extensions/qmic/CMakeLists.txt D +0 -105 plugins/extensions/qmic/kis_input_output_mapper.cpp D +0 -30 plugins/extensions/qmic/kis_input_output_mapper.h M +92 -0 plugins/extensions/qmic/kis_qmic_import_tools.h M +10 -8 plugins/extensions/qmic/kis_qmic_interface.cpp https://invent.kde.org/graphics/krita/commit/ca0d473bcd570f0332458b296efce25bf6344631