Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> There are still the case when point position is not saved on
>> ewoc-invalidate - when the buffer is displayed in the other window:
>
> Plus when the buffer is "displayed in a window configuration saved in
> a register", plus when invalidate is called from within
> `save-excursion`, plus ....
>
> Arguably Emacs should provide some way to better handle those issues
> (e.g. provide a hook that can be used to influence the way markers move
> in response to text removal+insertion), but until it does, I recommend
> you live with the occasional suboptimal behavior.

Ok, I will.  But take a look at the attached patch, please.  Few
additional lines and the point is saved in all windows!  Not sure about
those additional cases you mention though, but if the behavior will be
suboptimal, I'll be ready :)

Thanks!