Finishing Composite to handle transformed windows

[Amir Bukhari]

> On Sat, 2006-01-07 at 12:20 +0100, Amir Bukhari wrote:
> 
> > That may solve the cliping issue, but what about event processing. 
> > Normal window (bellow topmost window) will never receive 
> input events.
> 
> Oh, that's already a separate issue -- we have an external 
> mechanism to remap how global mouse position is translated to 
> a window-relative position. We need this as any window 
> transformation changes the mapping in a fashion unknown to 
> the X server.
> 
> > For reason (as I think) we reparent all windows (also 
> redirected like 
> > menus
> > ...) to one Pseudo root window (PRW),
> 
> We can't reparent OverrideRedirect windows, so we'd have to 
> use something like the PseudoRoot extension that I developed 
> years ago so that non-CompositeManager applications would see 
> this new window as the 'real' root and create their override 
> redirect windows within that.
> 
Yes and thus we have a hack in xserver to reparent OverrideRedirect windows
to our PRW. 
If event processing has a solution as you mention above then "topmost"
approach will be a good choice.
We should be carefull not to break existing Window Manager, which may be
confused by those new class of windows, this can be solved by declare them 
As OverrideRedirect, so WM will not interface with them.

-Amir