kwin_wayland xdgv6: crash on ping

[David Edmundson]

> do this theory makes any sense or the cause should be found elsewhere?
> (also,
> Martin, any idea? or i just talked nonsense? :p)
>
> That theory makes a lot of sense.
takeActive previously didn't do any talking to the client we were taking so
it could have been always broken and we wouldn't have known.

ShelllClients *should* be destroyed if the ShellSurface is destroyed or the
underlying Surface is destroyed.

Pings are on the global not the resource. For us to be getting a wayland
protocol error here, the client must have unbound the global, which implies
the app we're talking to has completely closed. We /shouldn't/ still have a
ShellClient.

I can imagine this crash if we have two windows belonging to the same
process, and the client dies. We handle the first one closing (frame 18)
and in doing so activate the second window of that now disconnected app.
Under X that situation would just be a harmless BadWindow  error.

It might be good to print the window titles of the shellclients of both
frame 18 and 4.
I'll try to too.

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/plasma-devel/attachments/20170803/d73161a1/attachment-0001.html>