Revert "KGlobalAccel: port to KKeyServer's new method symXModXToKeyQt, to fix numpad keys"
This reverts commit 2c20ddff. I'm reverting as we don't have a fix yet and the next release is too close to risk any changes. As explained in the original review request for this feature: touching this code base is dangerous. The code is extremely fragile. A month of living in master was not enough to spot the severe regressions we had. As we are now only a few days away from next release we cannot expect to discover regressions any fix would cause. I don't really see a solution to this problem: any change might cause issues especially in KWin, KScreenLocker and KGlobalAccel. All of them have the problem that the maintainer of those components is not running X11 and won't spot the regressions. Futhermore we don't have any test cases for the complete key handing stack on X11. We won't find the regressions. My personal recommendation would be to no longer change the X11 side and instead all together work on Wayland and get it ready. There the world looks better: we have a sane input stack which is completly unit tested. The first regression caused by the changes here was discovered by a KWin test case for Wayland. That's the world where we can change and improve the stack. Let X die, long live Wayland! If someone wants to take on the task to improve the X11 stack here I expect some serious work on the testability of this fragile stack. If we see that the shortcuts will still match I would be way more confident to accept a change. CCMAIL: kde-frameworks-devel@kde.org BUG: 384597
parent
df429c92
Please register or sign in to comment