multi-screen fixathon

[Marco Martin]

On Tuesday 16 December 2008, Aaron J. Seigo wrote:
> On Tuesday 16 December 2008, Marco Martin wrote:
> > did set up a really precary twinview, and i can quite reproduce a problem
> > like that, the problem seems when you try to select the same activity in
> > two views, a thing that can likely break :p
>
> the only way i can imagine that happening is with a multi-input (not multi-
> touch!) system with one plasma thread per input queue. iow, doubtful.
>
> what is supposed to happen is this (S for screen, C for containment, V for
> view):
>
> * V1 is on S1, V2 is on S2
> * S1 is paired with C1, S2 is paired with C2
> * user selects C2 on S1
> * C2 takes S1 from C1, C1 temporarily get assigned no screen
> * C2 then gives it's old screen, S2, to C1
> * we end up with S1 paired with C2 and S2 paired with C1
> * screenOwnerChanged signals should be emitted at every step along the way,
> causing the V1 and V2 views to follow along with what's happening.
> * V1 is now viewing C2, V2 is now viewing C1
>
> the question is where along this path does it break down.
it seems that at the end of operations the sistuation is the following:
c2 is correctly on s1 c1 on s2
v1 is tied to s2, but v2 doesn't have containment, so the position of the view 
is quite random