Wouldn't it be better to have an utility function that one can use to get a button code for the given Qt::MouseButton?

API-wise, I think we need something like this

Qt::MouseButton buttonToQtMouseButton(uint32_t button);
uint32_t qtToButtonMouseButton(Qt::MouseButton button);

Why add a function nobody needs? The function is already present, only not exposed to the header.

I'm asking you only to replace buttonCodeToQt with the utility function.

// pointer_input.h

uint32_t qtToButtonMouseButton(Qt::MouseButton button);

// pointer_input.cpp

uint32_t qtToButtonMouseButton(Qt::MouseButton button)
{
    return s_buttonToQtMouseButton.key(button);
}

Yea, definitely makes sense. Also that the function for the other direction is not implemented.

I meant "not exposed"

I agree it's better as a function, I was asked to return a hash in this comment here
https://phabricator.kde.org/D25663?id=70691#inline-144704

Maybe I misunderstood, @romangg can you confirm?

Yea, this was a misunderstanding. I meant there the internal representation should be changed to being a hash instead of a switch-statement in a separate patch (as you did in D25763), but I didn't mean that the getter should return the full hash, only that the getter should be exposed as well in a separate patch.

Forgot to remove?