Not so sure why phabricator didn't set you as the commit author... sorry about that.

In D11061#219841, @davidedmundson wrote:

FWIW, there's a lovely Qt-GConf wrapper in plasma-pa/gconfitem

May I ask when this gsettings is introduced? Is it required after certain ibus version? Is there any case that this will not work?

I sent the comment before you corrected the diff branch. LGTM.

Oh, ok.. now the diff looks normal.

Emm, mind to get rid of the .desktop part for this review?

In D8529#186000, @broulik wrote:

Good to go, no?

In D8798#175698, @subdiff wrote:

Sorry for the long delay on a review.

Can you give a short overview on how the switching works between Xlib and libinput backend on X? So what's the runtime detection if libinput is available or not? Somebody above said mouse kcm is not available without libinput anymore with this patch. But this should not happen of course.

Naming the config option Pointer Acceleration is problematic, because libinput names another property "pointer acceleration". I think we should not change the name. Unneeded deviation from upstream. Maybe better add a tooltip with a description if people complain.

In the touchpad kcm the backends for libinput and xlib are in different files (in the same backend/x11 subfolder). Is this possible here as well?

In D8798#175698, @subdiff wrote:

Sorry for the long delay on a review.

Can you give a short overview on how the switching works between Xlib and libinput backend on X? So what's the runtime detection if libinput is available or not? Somebody above said mouse kcm is not available without libinput anymore with this patch. But this should not happen of course.

Naming the config option Pointer Acceleration is problematic, because libinput names another property "pointer acceleration". I think we should not change the name. Unneeded deviation from upstream. Maybe better add a tooltip with a description if people complain.

In the touchpad kcm the backends for libinput and xlib are in different files (in the same backend/x11 subfolder). Is this possible here as well?

@jriddell ping?

In D8798#170892, @jriddell wrote:

This disables the input kcm if libinput x11 plugin is not found. I expect this is fine, it is for ubuntu distros at least, but it might be worth asking distros if it causes them a problem.

revert the accidentally changed currentIndex in ui file when save in designer.

Whether it shows up reallys depends on your system. Especially when there's multiple mouse driver available and you may using evdev.

@apol got a chance to test it?

Also just FYI, My PR of fcitx's gtk im module recently got merged to freedesktop runtime. https://github.com/flatpak/freedesktop-sdk-images/pull/58

In D8798#169156, @apol wrote:

In D8798#169155, @ngraham wrote:

$ kcmshell5 touchpad
kcm_touchpad: Using X11 backend

The file on disk is at /usr/lib/x86_64-linux-gnu/qt5/plugins/kded_touchpad.so (on KDE Neon and other Ubuntu-based distros)

Are you sure? because this is modifying files in kcms/input that creates a kcm_input.so... Is it a kcm that extends kcm_touchpad.so?

Fix based on comment

Use the new kcoreaddons function to cal the length.

forget to use else..

Revert the accident checking length change. Also make weightedLength linkage local.

revert the reordering of code in unrelated part

Add a version tag instead of always using master.

Since those class are called "X11....", can you avoid using those class in the caller instead of make themselves check the platform?

Sorry for being lazy to review this.
I noticed that there's no devicePixelRatio in your code change so I think there's a problem in hidpi.
AFAIK, QScreen returns size after scaled with hidpi. You need to restore it to make it work on hidpi correctly.

Just noticed that FindPkgConfig has import_module now.. Personally I just port my code to use that.

Being split into small patches

.. Sorry click on the wrong one.

To be split into small patches.

In D8460#161613, @subdiff wrote:

In D8460#160642, @xuetianweng wrote:

Also, I doubt if our X11 backend work on Xwayland. AFAIK Xwayland will only
get a virtual cursor. Thus I just move the condition to check the name of qpa
platform for now.

Can you explain this more? And why is checking the qpa platform name now different to the previous check? But I think you're right that the backend is not used in the Wayland session. The workaround in KWin reads in the X mouse acceleration directly from the config file.

Remove backend check so xwayland settings may be changed.

update based on comments.

@mart I understand what you want to say, but for this specific case:

Update accroding to subdiff's comment.

Merge kmousedlg.ui into kcmmouse.ui.

ping

using a UI file and add libinput support.