Pointer grabs causing accessibility issues! Why not deprecate them?

[Eamon Walsh]

Daniel Stone wrote:
> On Tue, Apr 29, 2008 at 09:43:57PM -0400, Eamon Walsh wrote:
>   
>> Daniel Stone wrote:
>>     
>>> ... unless it's a security-sensitive app like PolicyKit, which
>>> explicitly says 'please do not let anyone else get any kind of access to
>>> input, because I'm a special case application'.  Hence not letting
>>> anyone else get any kind of access to input is kind of a feature, in
>>> terms of actually being protocol-compliant, and respecting the
>>> application's wishes.
>>>       
>> Except XQueryKeymap and XQueryPointer allow complete recovery of the  
>> input regardless of grabs.  Are the PolicyKit people aware of this?
>>     
>
> I think the way to go is to just return blank for the keymap/button
> map/co-ords for those two, unless the client requesting is the
> currently-focussed client, and there are no active grabs.  Sound sane?
>   

Works for me.  However if we're going to change server behavior, there 
are a few other requests to deal with:

<opens up the spreadsheet...>


XkbGetState and XiQueryDeviceState: same idea as above should work.
MPXQueryPointer: same idea here?  Not sure what this is for.

GetMotionEvents: client must own the window?
XiGetDeviceMotionEvents: no clue - it's a complete motion history dump.

I would note that the motion history is an optional server feature, 
according to the spec.  A motion history buffer size of zero is valid.



-- 
Eamon Walsh <ewalsh at tycho.nsa.gov>
National Security Agency