[PATCH inputproto multitouch 10/13] Touch grabs must be sync only

[Peter Hutterer]

On Tue, Aug 23, 2011 at 05:52:05PM -0700, Chase Douglas wrote:
> On Wed, Aug 24, 2011 at 10:47:04AM +1000, Peter Hutterer wrote:
> > On Tue, Aug 23, 2011 at 05:37:39PM -0700, Chase Douglas wrote:
> > > On Wed, Aug 24, 2011 at 09:07:20AM +1000, Peter Hutterer wrote:
> > > > The ownership passing is quite similar to sync grabs, so require touch grabs
> > > > to always be sync.
> > > 
> > > I see two sides of this:
> > > 
> > > Touch grabs are like pointer sync grabs because the owning client can keep other
> > > clients from receiving ownership or cleaning up state if it doesn't
> > > accept/reject.
> > > 
> > > Touch grabs are like pointer async grabs because the client doesn't have to ask
> > > for each event individually. Events are asynchronously streamed to the client.
> > > 
> > > Although it's not technically necessary, for protocol saneness it might make
> > > sense to add a third grab mode like "GrabModeOwner".
> > 
> > how's the receival of ownership events decided then? still with the event
> > mask? should we have two modes? GrabModeTouch, GrabModeTouchOwner?
> 
> Ownership events are still recevied via the event mask. What is the difference
> between GrabModeTouch and GrabModeTouchOwner in your mind?

there's a difference in normal touch event delivery depending whether the
ownership mask is set. Presumably the same rules are valid for touch event
delivery.
If we refer to the touch grab mode as GrabModeOwner this would IMO confuse
the ownership semantics if the actual rules are still dependent on the mask.
i.e., if I select with GrabModeOwner but TouchOwnership _not_ selected in
the event mask, I don't get ownership events. I think GrabModeTouch is just
fine to indicate the different sync/async semantics for touch events.

Cheers,
  Peter