[LightDM] On keyboard layouts

[Christian Seiler]

Hi,

I've just pushed a merge request for updating the Qt bindings so that
they also understand the keyboard layout business:

https://code.launchpad.net/~christian-w/lightdm/qt-binding-keyboard-layouts/+merge/205356

However, while experimenting with that, I've made a couple of
interesting observations:

1. All greeters I've looked at (Gtk, unmaintained Qt, KDE, Razor-QT,
Unity) don't seem to implement this at all... Ok, the Qt-based greeters
are no surprise. But for Gtk/Unity: I see some references to old bug
reports online if I search for this, but they reference C code, whereas
the Unity greeter is now written completely in Vala. (At least current
bzr trunk.) And ok, Vala is compiled to C, but the Vala code I looked
at contains nothing that I could find that contains anything resembling
keyboard layouts.

Am I missing something here? Currently, it appears to me that nobody is
actually using that code. Nevertheless, I've seen screenshots of the
Unity greeter with a keyboard symbol, is that perhaps an old version?

2. I get a huge list of layouts from the backend, it appears to me that
those are all layouts that are installed on my system. (472 entries) If
I want to use that for writing some kind of layout switcher for a
greeter, I don't see how that could be made usable in a reasonable way.
The only greeter outside of lightdm of which I know that it supports
layouts is gdm3, and that uses the current XkbLayouts X setting to
restrict the available layouts to those, making it much easier for a
system administrator to choose a sane policy of available layouts.

3. The gobject API exposes three properties for keyboard layouts: name,
short description and description. From reading the API, this seems
perfectly reasonable, but looking at the data the contents seems
*really* weird:

  - Description is fine, that is something like 'English (US)' or
   'Berber (Morocco, Tifinagh extended)'.

  - Short description ist empty *most* of the time, but sometimes it
    contains a ISO 639-2 (or -3?) language code, but *only* on a
    minority of the layouts. (Ok, can probably live with that, but the
    name 'short description' is misleading IMHO.)

  - Then we come to name, which is the really weird part. For the most
    basic layouts, name is simply the X layout code like 'us', 'de',
    'at', 'fr', etc. However, if there is a variant (no dead keys, euro
    sign, etc.), then the name field consists of name "\t" variant, so
    for example for the Belgian ISO alternative layout, name is equal
    to "be\tiso-alternative".

Funny thing is, the API itself works fine, getting the current layout
and then switching to something else is not a problem.

Regards,
Christian