[PATCH v2 weston] input: Don't test keyboard/pointer/touch pointers

[Derek Foreman]

On 02/02/15 04:26 PM, Hardening wrote:
> Le 02/02/2015 22:24, Derek Foreman a écrit :
>> Thanks for looking at this rather cumbersome patch! :)
>>
>> On 02/02/15 01:40 PM, Bill Spitzak wrote:
>>> Is there a reason it does not just clear the pointer when
>>> keyboard_device_count is changed to zero? That would seem like a smaller
>>> patch.
>>
>> Pulled Jonas Ådahl in on the CC for that question...  I understand why
>> weston_pointer isn't freed when released, but is there a need for the
>> same behavior for weston_keyboard and weston_touch?
>>
> 
> I had a discussion on this on IRC, and the reason is that we wanna save
> the last position of the pointer. I'm not sure it's a valid argument, as
> the counter part is that the cleaning of pointer, keyboard and touch is
> not done, or done partially. Not so nice for weston, a reference
> implementation, the code should be straightforward and here it's not.

Yeah, I stepped on that strangeness a little while ago and that's why
this patch exists...

> My opinion is that this last position should be stored in the seat, and
> restored when a new pointer is bound to the seat.

I can't personally think of any reason why that wouldn't work...  Did
anyone disagree?

I wonder if there are dangling pointers we get to ignore if we don't
free things, and we're depending on that...

> I will propose patches on seat releasing, that appears to be hard if not
> impossible to do safely right now.
> 
> Regards.
>