In mkspecs/features/qt_configure.prf there's

defineReplace(qtIsFeatureEnabled) {
    enable = $$eval($${currentConfig}.features.$${1}.enable)
    !isEmpty(enable) {
        $$qtConfEvaluate($$enable): \
            return(true)
    } else {
        equals(config.input.$${1}, "yes"): \
            return(true)
    }

    return(false)
}

so if input.foo is set to true, feature foo will be considered enabled.

This happens for qmake's 'xcb-native-painting' command line switch which basically sets input.xcb-native-painting=yes, thus enabling the feature.

We are currently lacking this in CMake feature system.
If a user uses the configure script and passes -xcb-native-painting, that will be translated by the script to -DFEATURE_xcb_native_painting=ON

But if a user uses cmake directly, and passes -DINPUT_xcb_native_painting=ON, that will be ignored. If we want parity with qmake, we should consider implementing the same behavior at CMake feature level.