[PATCH weston v2 1/3] toytoolkit: Allow operation without a keymap

[Daniel Stone]

Hi,

On 22 June 2013 01:54, Bill Spitzak <spitzak at gmail.com> wrote:
> Daniel Stone wrote:
>> If the only input API you need is the input method API, then you don't
>> need other input APIs.
>>
>> wl_keyboard is pretty much for people who need to support non-textual
>> keyboard input, e.g. hotkeys/media keys and shortcuts.  As long as
>> those usecases continue to exist - which is forever - we're going to
>> need wl_keyboard in addition to wl_text.
>
> I think I am not asking the right question.
>
> Imagine a client that uses wl_keyboard for "hotkeys/media keys and
> shortcuts" and uses the input api for all text, exactly like you suggest.
>
> Now imagine we feed this client an edited xkb description. It is edited to
> remove as much complexity as possible without breaking the client. For
> instance every single key that produces text and every dead key is changed
> to produce the letter 'a' no matter what the shift state is.
>
> What I suspect is the minimum resulting xkb description where this client
> still works is going to be a LOT smaller, and will be much closer to the
> proposed gamepad evdev mapping.

That doesn't fly for shortcuts, especially when you take into
consideration things like Ctrl+Shift+W needing to trigger the 'close
window' shortcut, even if you have multiple layouts active, and your
current layout doesn't actually have a key which produces W.

If you can think of a way to simplify this, please send concrete
patches to GTK+, Qt and co., and when they've changed, we can later
simplify xkbcommon.  But unfortunately in the last however many years,
no-one's managed to come up with a better solution which actually
works for all the real-world usecases, without the full keymap.

Cheers,
Daniel