bug#43412: [FEATURE] autorevert-only-if-visible [PATCH]

[Boruch Baum <boruch_baum@gmx.com>]

On 2020-09-17 16:13, Eli Zaretskii wrote:
> > Date: Wed, 16 Sep 2020 16:11:04 -0400
> > From: Boruch Baum <boruch_baum@gmx.com>
> > Cc: 43412@debbugs.gnu.org
> >
> > It looks to me like that's not the case: my testing seems to show that
> > it does catch cases of changing a buffer displayed in a window but does
> > not catch changes of frames due to functions like `other-frame' or
> > `select-frame'.
> >
> > So, the good news is that I've written the code that makes the
> > improvement for the caught cases, and I can submit that.
> >
> > As for the cases of changing frames, a less-desirable option would be to
> > preempt bug-reports by documenting the limitation. Auto-revert already
> > has other curious limitations (eg. for dired buffers it doesn't operate
> > _at__all_ on many types of file changes), and this limitation only
> > introduces a delay, so by comparison its a pretty mild limitation.
>
> If we have no better way, we could document this as a limitation,
> yes.  But let's make one more attempt to solve this.
>
> > A better option would be able to catch frame-change events. I haven't
> > found a straightforward way to trap that. Does such a method exist?
>
> Can you describe the problematic case in more detail?  With that in
> hand, perhaps Martin (CC'ed) could suggest a method.

I'm needing now to focus on Rosh HaShanna preparations, and can let
emacs distract me for the next few days, but from memory, on tty emacs:

1) touch ~/foo

2) find it in emacs

3) bury the frame so it's not human-visible

   3.1) OT: I find it useful to globally bind C-x 5 b to function
        select-frame-by-name, because I do this routinely as part of my
        workflow. Another option might be C-x 5 2.

4) M-! echo "bar" >> ~/foo

5) switch to the buried frame

6) Depending upon your timing, you'll need to wait up to
   auto-revert-interval seconds for the buffer to be updated.

> > An inelegant solution that would cover most of the remaining events
> > would be to advise :after ~4 frame functions, and to add an element to
> > variable `move-frame-functions'.
>
> Yes, I'd prefer to avoid such solutions.

It maybe could all be solved if there were a hook for frame change of
state (eg. getting/losing focus, being raised/buried). It's surprising
to me that such events were never thought significant enough in emacs to
merit a hook.

--
hkp://keys.gnupg.net
CA45 09B5 5351 7C11 A9D1  7286 0036 9E45 1595 8BC0