bug#12170: save-excursion fails boundary case with recenter

[martin rudalics]

> Maybe.  If it's really important,

It's not.

> I could try looking at this in the
> debugger, although doing so to investigate such situations is
> notoriously hard, because redisplay is entered several times.
>
> IMO, save-excursion is simply not designed to make sure display isn't
> changed in such cases.

Agreed.  But `save-window-excursion' doesn't handle this case either.
So the difference should be in one of the four assignments

  w->start_at_line_beg = 0;
  w->update_mode_line = 1;
  w->last_modified = 0;
  w->last_overlay_modified = 0;

and I'd obviously vote for the first one.  Interesting side-effect.

martin