On showing windows where the users want them, on wayland

[Aleix Pol]

On Tue, Apr 14, 2020 at 11:54 AM David Edmundson
<david at davidedmundson.co.uk> wrote:
>
> It depends on whether someone needs to recreate xeyes!
>
> >- make Qt report the screen where the current cursor is the primary
> one,
>
> QScreen constantly changing would make plasma panels move, we don't want to expose cursor pos that way.
>
> >- include in the KWayland::Client::PlasmaShell a way to report the
> screen (e.g. the edid or another identifier) with the current cursor.
>
> We would probably use a wl_output ID as an arg.
>
> Being in PlasmaShell is also potentially problematic.
>
> You need this information before we show.
> We don't have a wl_surface after the first hide() before the second show(), which means we can't have a plasmashell.
>
> It might be doable still, but that order needs thinking about before making protocol mods.

Was talking about org_kde_plasma_shell rather than org_kde_plasma_shell_surface.

> -----
>
> From a "pure wayland" POV, this approach is backwards - the client should be specifying the position semantically then kwin would do its thing.
> We have the client doing things in absolute positions to have a nice client migration path from X. I hope to be changing that for KF6.

Would you suggest having a property clients can enable so they get
assigned to the cursor's screen upon show()? Works for me too.

I don't see how KF6 will change anything there.

Aleix