New version of xfixes with client side cursor support

[David Reveman]

On Fri, 2007-02-16 at 17:34 -0800, Keith Packard wrote: 
> On Fri, 2007-02-16 at 20:27 +0100, David Reveman wrote:
> 
> > It's no worse than NameWindowPixmap, which will give you a pixmap for
> > some random state of the window and not necessarily the pixmap that
> > matches the size indicated by the last ConfigureNotify.
> 
> I disagree; you want to paint the correct cursor at a specific spot on
> the screen and NameCursorPixmap doesn't give you this ability; you'll
> end up with some other cursor entirely.
> 
> > Sure, it might be nice with something better but I don't think that
> > makes sense unless we're interested in improving NameWindowPixmap too.
> 
> Windows have stable names, you're only using 'the current cursor' here,
> which is not naming a stable object.
> 

I still think it's the same problem. Compositing managers can't be sure
that they're painting the correct window contents at a specific location
on the screen in the same way that NameCursorPixmap doesn't allow you to
paint the core pointer cursor contents at the right spot. The object
we're naming with NameCursorPixmap is not the current cursor but the
cursor contents for the core pointer.

Anyhow, I agree with you now that this might not be good enough for
cursors. How about adding the cursor ID to the event and the
NameCursorPixmap request? Only the contents of animated cursors would
change.

- David