Andreas Beck wrote:
>> The stylus uses buttons and valuators, but these are not contigeously
>> enumerated.
>
>
> It is o.k. to have a sparse set of buttons. However it was intended to have
> a packed set of valuators.
>
> It is o.k., though, to "pad" it with dummy valuators that never get sent.
> That is basically the same behaviour as for buttons.
but to find out which valuators are actually valid in an event, I would
have to query the device /
valuator info.
> What should the 'num_axes' member report ? the number of (absolute)
> valuators ? The highest id ?
>
>
> Highest Id. Only way to query it in a reasonable way.
hmm, then you should rename 'num_axes' to make this clear :)
> Even though - if that happens, you can easily just queue multiple events,
> just as you do when queuing noncontigous ones (like when valuators number 1
> and 3 have changed).
>
> Note the "first" member, which can be used to support up to 4 billion
> valuators, which should suffice *g*.
right. Sounds like the right thing to do. Unfortunately, this kind of
information is not contained in any
document. May be I'll just write down what I find out (by code
inspection and by asking you) and
integrate it into the manual...
> So, at the given state of GGI I must ask: is anybody interested in these changes
> at all ?
>
>
> Yeah ! Sure. I don't have any evdev capable device, but it is a good thing
> to do things right, so: YES !
Isn't evdev supposed to become *the* driver for most of the input
devices for newer linux kernels ?
At least the protocol looks very general in scope, so it could easily
absorb keyboard, mice, etc.
Regards, Stefan
