bug#81406: 31.0.90; rotate-windows loses window balance when window-combination-resize is t

Eli:

 > Well, I guess this should also be mentioned, because dedicated windows
 > are in many cases used so as to keep them in the same place on the
 > frame.

Sean:

 > I don't think that should be allowed.  Lisp code assumes dedicated
 > buffers really are dedicated, we are opening up too many annoying ways
 > in which things could break.

 > I'm not sure we should be allowing this, tbh.  It might be the same
 > window object but if it ends up with a different size and position then
 > I think we have done something wrong.

Since I avoid dedicated windows like the plague I have no opinion on
this.  But I have been told (by Stefan Monnier or Drew Andrews probably)
that the main purpose of dedicated windows is to avoid that
'display-buffer' uses them for showing another buffer.  IIRC that
happened when I proposed to not split a dedicated window for showing
another buffer in the new window.  And as a matter of fact one can move
around a dedicated window within a frame by deleting and splitting
windows in various ways.

So to meet your concerns we would always have to signal an error when
'rotate-windows' encounters a dedicated window or add an option to
handle that case.  Keeping dedicated windows at their original position
and rotating the rest is not a feasible option at the moment IMHO.

martin