wl_shell_surface_resize semantics

[Tiago Vignatti]

On 12/07/2011 07:54 PM, Bill Spitzak wrote:
> On 12/07/2011 02:56 AM, Daniel Stone wrote:
>
>> * client requests compositor handle resizing
>> * compositor takes over grab handling, sends button release event to
>> client, noting through a flag that it's a cancel event rather than a
>> physical button release event
>
> I'm sorry, but I still do not understand why a client, that has code in
> it to "request compositor handle resizing" cannot also have code in it
> to know that it may not get a button release event.

device fire event -> kernel read -> compositor fetch -> compositor 
notify client -> client do whatever it wants

but now the compositor is misbehaving when a client steals the grab for 
its own purpose, which I fixed in that patch. Rob though wants to deal 
with a more complex case where a client grips the surface and perform 
the resize till the cursor goes outside it. There's a doubt whether we 
should emit or not a button release outside the surface, and krh said we 
might want a new event announcing when a whole movement is done.

      Tiago