Created attachment 306062 [details] [review]
0001-input-settings-x11-don-t-create-non-existing-atoms.patch

Mutter uses a bunch of libinput properties for configuration and bails out if they don't exist on the device. But both get_property() and change_property() *create* the atom if it doesn't exist yet.

This confuses other layers of the stack that have similar checks - once mutter runs the property exists (not on the device, but in the global atom table).

Example for this confusion (though this one doesn't seem to be mutter's fault):
https://bugzilla.redhat.com/show_bug.cgi?id=1199825