bug#23785: Emacs 25: "Undo" overdoes things.

[Eli Zaretskii <eliz@gnu.org>]

> From: phillip.lord@russet.org.uk (Phillip Lord)
> Cc: Alan Mackenzie <acm@muc.de>,  23785@debbugs.gnu.org
> Date: Fri, 17 Jun 2016 22:47:20 +0100
> 
> The following patch addresses the issue in my hands. I'd welcome
> confirmation.
> 
> Stefan what do you think? You suggested something like this for
> Bug#22295 also. I think that the problem is the same; if I read
> insert-file-contents correctly, it disables collection of undo
> information. I may well be reading it wrong though.
> 
> Given that this has turned up in two places now, it may be a wider
> problem, and the more general solution might be the better one. It might
> also invalidate the necessity for 12e009e52, but I haven't tested that
> yet.
> 
> Alternatively, revert-buffer could just force undo-boundary.
> 
> >From 2ab1f314ad6fe0e68420cc510445495467d82b8f Mon Sep 17 00:00:00 2001
> From: Phillip Lord <phillip.lord@russet.org.uk>
> Date: Fri, 17 Jun 2016 22:34:50 +0100
> Subject: [PATCH] Fix missing undo-boundary after revert-buffer
> 
> * lisp/simple.el (undo-auto--boundaries): Ensure an undo-boundary after
>   every command whether it (apparently) changes the buffer or not.
> 
> Addresses Bug#23785
> ---
>  lisp/simple.el | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/lisp/simple.el b/lisp/simple.el
> index b66827d..3110430 100644
> --- a/lisp/simple.el
> +++ b/lisp/simple.el
> @@ -2875,6 +2875,10 @@ undo-auto--boundaries
>    "Check recently changed buffers and add a boundary if necessary.
>  REASON describes the reason that the boundary is being added; see
>  `undo-last-boundary' for more information."
> +  ;; (Bug #23785) All commands should ensure that there is an undo
> +  ;; boundary whether they have changed the current buffer or not.
> +  (when (eq cause 'command)
> +    (add-to-list 'undo-auto--undoably-changed-buffers (current-buffer)))
>    (dolist (b undo-auto--undoably-changed-buffers)
>            (when (buffer-live-p b)
>              (with-current-buffer b

I'd indeed suggest to have on emacs-25 a change that only affects
revert-buffer, and have the above on master.  It sounds too scary to
make a change that affects all commands this late in the pretest.

Thanks.