Constraining cursor to RandR crtcs

[Dirk Thierbach]

On Wed, Apr 04, 2007 at 12:51:35AM +0200, Alexej Davidov wrote:

> But as for a default: allow the pointer to move through dead zones.
> While it can be annoying to lose the pointer, I think this is the most
> intuitive solution. 

And it may actually be useful to move the pointer through such a 
dead zone: Sometimes I "park" windows at the edge of the screen, so
that they are only partially visible. Suppose I'd "park" a window at
a corner such as has been described before, and then I want to
select some text in this window, where both endpoints of the text
to select are visible on different screens. I could do that without
having to bring the window back to full visible if I can just move 
the pointer through a dead zone. And there's no danger of loosing the
pointer, because I get visual feedback by the changing selection, which
is still partially visible.

> Pointer position indicators at screen edges when the pointer is in a
> dead zone, as suggested in another post, would be useful, though.

It may be useful to decouple the pointer for the whole Xinerama
screen from the cursors of the single screens. Then a WM could use
easily use those cursors as position indicators if the pointer is
in a dead zone, if that is the desired policy.

That could actually integrate with the multi-pointer extension MPX
(which I don't know anything about other than it exists, so I may be
wrong). Something along the lines of: There's a number of pointers,
and a number of cursors. The latter are bound to screens. Default
policy is to assign a cursor to a pointer whenever that pointer is
inside a screen, but the WM can override this by listening to suitable
events and do whatever he likes.

- Dirk