bug#81193: 31.0.90; Wrong offsets in 'set-frame-size-and-position'

 > There is a distinction between child frames and "normal" frames. Child
 > frames can be positioned on PGTK Wayland.
 >
 > Given the problems with different window managers and windowing systems in
 > general, I think the long term solution should be to not make child frames
 > top level surfaces at all.

This is not entirely trivial: For example, I have my minibuffer in a
child frame that serves as minibuffer-only frame for all normal frames
with the help of re-parenting.  When I delete the current parent of the
child frame, I do make the child frame top-level because there might be
no other normal frame around to serve as new parent.

 > By top level surfaces I mean windowing system
 > windows/surfaces that involve the window manager. Toolkits like GTK do not
 > implement their widgets (buttons, etc) as top level surfaces, and I think
 > this could be a good choice for Emacs child frames too. This way the child
 > frames would be confined within the well defined application window area,
 > and not at the whim of various window managers.

martin