transient surfaces for xwayland and tooltip love

[Bill Spitzak]

Tiago Vignatti wrote:
> Hi,
> 
> I stumbled on the problem of transient surfaces for xwayland. So I first
> checked tooltip type of windows and decided to create these for toytoolkit as
> well, so now we have a way to test the transient's on it as well.
> 
> Probably the most important on this set is how we'll decide the keyboard
> focus behavior on newly created surfaces. So that's why I'm suggesting this
> new transient_method on the protocol. The tooltip implementation then is just
> a funny way to check if the proposed solution works here. I think it's cute :)
> 
> Now I want your opinion now to see if this is reasonable and whether we can
> cross that TODO item as done or not. Protocol and implementation reside on
> "tooltip" branches on my trees.

The server cannot be in charge of focus, because you will end up with 
the X mess of clients guessing what type of window gets the correct focus.

Instead the client should be able to set the focus for any input device 
to any of it's windows. The server can refuse the request (for instance 
if there is a newer one, or if the input device's related pointer is not 
pointing at one of the client's windows) but never changes focus except 
in response to such requests.

Then whether the tooltip gets focus is simply a decision by the client, 
and it can even vary depending on the user's actions (clicking on the 
text in the tooltip may want to put the focus inside it, for instance).