bug#17405: C-g C-g in GC does not autosave

bug#17405: C-g C-g in GC does not autosave

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

I think Emacs used to be smart enough to autosave from C-g C-g
within a GC.  I guess that broke and someone decided to fix it
by not trying to autosave.

Can't we make it autosave correctly?

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use Ekiga or an ordinary phone call.

bug#17405: C-g C-g in GC does not autosave

[Lars Ingebrigtsen]

Richard Stallman <rms@gnu.org> writes:

> I think Emacs used to be smart enough to autosave from C-g C-g
> within a GC.  I guess that broke and someone decided to fix it
> by not trying to autosave.
>
> Can't we make it autosave correctly?

I think that sounds like somewhat surprising behaviour -- why should
`C-g C-g' (at any time) autosave?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#17405: C-g C-g in GC does not autosave

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Date: Fri, 04 Dec 2020 12:58:51 +0100
> Cc: 17405@debbugs.gnu.org
> 
> Richard Stallman <rms@gnu.org> writes:
> 
> > I think Emacs used to be smart enough to autosave from C-g C-g
> > within a GC.  I guess that broke and someone decided to fix it
> > by not trying to autosave.
> >
> > Can't we make it autosave correctly?
> 
> I think that sounds like somewhat surprising behaviour -- why should
> `C-g C-g' (at any time) autosave?

Because "C-g C-g" (if you type it fast enough) triggers the so-called
"emergency escape", whereby we might abort and dump core, subject to
user's confirmation.  See handle_interrupt in keyboard.c.

The answer to Richard's question is that this "someone who decided to
fix it" was Richard himself, back in Nov 2995 (see commit 118d6ca).
The comment to the "not in GC" condition explains this:

      /* It doesn't work to autosave while GC is in progress;
	 the code used for auto-saving doesn't cope with the mark bit.  */

Maybe we should see whether this is still the case, i.e. whether
auto-saving still cannot cope with the mark bit.

bug#17405: C-g C-g in GC does not autosave

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

>> I think that sounds like somewhat surprising behaviour -- why should
>> `C-g C-g' (at any time) autosave?
>
> Because "C-g C-g" (if you type it fast enough) triggers the so-called
> "emergency escape", whereby we might abort and dump core, subject to
> user's confirmation.  See handle_interrupt in keyboard.c.

Oh, the "abort and dump core (and autosave)" message?  Yes, I've seen
that one, but I assumed that Richard was talking about something else,
since he just mentioned "autosave".

(Speaking of which -- that message has always confused me.  I don't know
what I'm answering yes to "abort", "dump core" or "autosave", or a
combination thereof.)

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#17405: C-g C-g in GC does not autosave

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: rms@gnu.org,  17405@debbugs.gnu.org
> Date: Fri, 04 Dec 2020 13:54:29 +0100
> 
> (Speaking of which -- that message has always confused me.  I don't know
> what I'm answering yes to "abort", "dump core" or "autosave", or a
> combination thereof.)

They are two messages, shown one after the other:

  Auto-save? (y or n)
  Abort (and dump core)? (y or n)