[compiz] Misc WM buglets

[David Reveman]

On Wed, 2006-04-05 at 15:19 +0100, Mike Hearn wrote:
> > The server keeps the correct stacking order and compiz will track all
> > changes reported by the server and update its linked list of top level
> > windows accordingly. Compiz should never lose track of the stacking
> > order on the server side but I guess there's some special case where
> > compiz gets it wrong.
> > 
> > A well described use-case that makes the visible stacking order wrong
> > would be very useful.
> 
> Hmm, you don't see it at all? That's odd, it's pretty easy to trigger 
> and I think Quinn also sees this.
> 
> Basically, if you get 3 overlapping windows on one face of the cube, 2 
> on another, switch from 3 to 2 and back again, then try clicking on a 
> window and watch your click go "through" the window to the one 
> underneath which will then be raised.

No, this doesn't happen for me at all. It probably goes wrong before you
start switching viewport. You might be running some application that
creates a window that you can't see and it's doing something strange
with this window so compiz's stacking order gets one window wrong. If
one window is wrong you'll get all kinds of weird stacking bugs and
where you first see it go wrong might not be related to where the bug
is.

A small test application that stress tests window stacking could be very
useful to track down this problem.

> 
> So the _visible_ stacking order is correct, though I've seen windows 
> jump to the front on rotate too, but the input events don't seem to 
> correspond to what's on screen.
> 
> I'll have a poke at this myself at some point in the next few days 
> hopefully.

-David