Update to wl_seat v7

[romangg]

We currently are on v5. We should update to the currently highest version available v7.

New in v6

• New wl_pointer::axis_source::wheel_tilt enum value.
• New wl_touch::shape and wl_touch::orientation events.

New in v7

• wl_keyboard::keymap fd must be mapped with MAP_PRIVATE by clients.

In our Server library we don't need to change anything for the MAP_PRIVATE fd change immediately, because we copy the keymap for each wl_keyboard instance. We could do something similar like Weston with sealing off the fd. But I don't know if it's worth the effort as for example wlroots also has just provided copies for a long time and did a formal switch to v7. Since recently it doesn't do copies anymore though and instead sends read-only keymaps, which is independent of v7 though.

[romangg]

On a side-note after talking to @emersion it is clear that an implementation without copies could be done rather simply by sending the file as read-only. Sealing off the file like Weston does is not necessary. One can even do this with clients only supporting wl_seat v6 and below.

[12:27] <romangg> Since https://github.com/swaywm/wlroots/pull/3092 keymaps are sent without copies as readonly and that even for clients lower wl_seat v7. If that's possible why was v7's MAP_PRIVATE change necessary at all?
[12:28] <romangg> I thought the MAP_PRIVATE change was explicitly to allow compositors to share the same keymap with all clients and not to have separate copies for each.
[12:29] <emersion> yeah it was a mistake
[12:29] <emersion> MAP_PRIVATE isn't necessary
[12:29] <emersion> it's just that people didn't realize files could be opened twice
[12:30] <emersion> once in read-write mode, once in read-only mode
[12:30] <romangg> Ah, interesting. Well, mistakes happen. But thanks for clearing that up.
[12:30] <romangg> So the sealing off stuff in weston is also not necessary.
[12:30] <emersion> correct