[PULL] input fixes for 1.9, unloved.

[Peter Hutterer]

On Mon, Aug 09, 2010 at 09:00:27PM -0700, Keith Packard wrote:
> On Tue, 10 Aug 2010 12:04:35 +1000, Peter Hutterer <peter.hutterer at who-t.net> wrote:
> 
> > >     xkb: post-fix PointerKeys button events with a DeviceChangedEvent.
> > > exposed by 14327858391ebe929b806efb53ad79e789361883, devices with more
> > > valuators than the XTEST pointer device now skip events.
> 
> Where is the first DCE event coming from? And, how about
> FakePointerMotion? Does that need DCE afterwards as well? 


> You realize these code paths are a disaster, right?

this is a rhetorical question, right?

> 
> > >       xkb: if the button isn't down, don't fake an event.
> > > not strictly necessary for 1.9 but saves us a lot of effort, given that the
> > > ratio of physical releases to required XTEST releases is about a million to
> > > one.
> 
> Seems reasonable to me.
> 
> Reviewed-by: Keith Packard <keithp at keithp.com>

I've amended this one to fit on top of the v2 of the post-fix patch, the
hunks are the same, just the context changed.

> > >       Xi: reset the unused classes pointer after copying
> > > fixes a bug exposed by the first patch above.
> 
> If the other patch exposes a bug, then this patch should be applied
> first.
> 
> Reviewed-by: Keith Packard <keithp at keithp.com>
> 
> (I think I might even understand this -- the unused_classes essentially
> holds structures that might be used when flipping devices around)
> 
> > > Patches have been on the list, rumours have it the few people who looked at
> > > them ran away screaming before throwing Reviewed-by tags my way.
> 
> I'm with them.

thanks for the reviews though, much appreciated.

Cheers,
  Peter