[PATCH] protocol: Add a request and event to identify the visible area for a surface

[Bill Spitzak]

It would be nice if there was a clear way to find out *all* the visible 
area. Right now it looks like I have to try a really large rectangle and 
just hope I have managed to cover all of the current output. It seems to 
me that getting rid of the rectangle and returning the result as though 
it was infinite in size would be simpler and cover what I would imagine 
a toolkit will actually do. Is there a reason for the rectangle?

Another idea is to allow it to return a clip that is *larger* than the 
rectangle. This may be useful if the visible area is discontiguous: only 
the areas that intersect the rectangle are returned, and if none do then 
the "closest" one is returned. May be useful if a surface spans more 
than one output.

This should replace all the proposed automatic-positioning of child 
surfaces in the shell api. That was never going to work except for 
trivial popup modal questions. Real overlays want to draw graphics in 
both the overlay and main window pointing at each other, and popup in 
positions that use complex rules (like a submenu on the opposite side of 
a parent, or the parent menu shifted over so that a potential submenu 
can appear) that cannot be conveyed from client to server without the 
server becoming a program interpreter.