[PATCH weston 4/4] input: Detect keyboard capabilities

[Bill Spitzak]

On 05/06/2015 09:24 AM, Derek Foreman wrote:
> On 05/05/15 03:59 PM, Bill Spitzak wrote:
>> It sure seems like it would be better to just provide the list of
>> available keys to the client programs, and let them make up their own
>> categories of device types.
>
> I think desktop-shell is still going to want this information,
> regardless of whether we implement new protocol to feed it on to
> clients.  So it may be the case that clients want it too, but I don't
> think it would be better to only provide it to clients.

I meant clients of libinput. In this case the shell is a client of 
libinput. What I am suggesting is that the desktop shell has the list of 
keys to test, etc, rather than libinput.

It is true that a plan needs to be made so that exact duplicates of 
everything in the libinput api can be delivered to Wayland clients so 
they are not deprived of any ability to use an input device. The obvious 
method is to make a Wayland version for every libinput event and 
request. An alternative is for Wayland to provide the fd and the clients 
talk to libinput directly. Both of these have problems.