Constraining cursor to RandR crtcs

[Egbert Eich]

Nicolas Mailhot writes:
 > 
 > IMHO if you don't allow excursions within dead zones you'll always find a
 > screen setup where your solution suck. If you do allow excursions there

Right. The best (better: the least suboptimal) solution heavily 
depends on the use case, the setup and user preference.
Therefore this is a policy issue: the policy should be adjustable -
even if it's only for experimenting with different solutions.

Policies (proximity, keyboard shortcuts, whatever) are best done on 
the client side by the WM or some dedicated app which do the cursor 
warping.
WMs should have enough information to know the current position of the 
mouse cursor and - thru Xinerama - they  know about the locations of 
visible portions on the screen. So no pieces need to be added.
WMs (and even clients) need to learn how to deal with non-rectangular 
visible areas anyway (window, popup placement, pull downs etc.)
If the user doesn't have any client running that provide a 'cursor stay
in the visible area' policy then it's probably best to leave things as
they are with the cursor disappearing into nowhere.

Cheers,
	Egbert.