bug#38993: 26.3; EBDB saves the database on Emacs exit even when told not to

[Eric Abrahamsen <eric@ericabrahamsen.net>]

Jorge P. de Morais Neto <jorge+list@disroot.org> writes:

> Em [2020-01-06 seg 11:51:48-0800], Eric Abrahamsen escreveu:
>
>> Probably you've also got `ebdb-save-on-exit' set to t -- that adds a
>> save hook to `kill-emacs-hook'.
>
> Yes, it was t (the default).  I actually had read about it on the EBDB
> manual, but, since Emacs was asking whether to save, I interpreted that
> variable as whether Emacs would ask at all.  That is, I thought that if
> the variable was t then Emacs would ask whether to save, and if it was
> nil then Emacs would exit without saving and without any question.

I suppose I could have it do that (prompt for saving) rather than saving
silently. I guess I just assumed that in most cases users would want the
database saved, rather than discarding changes. I don't have very strong
feelings about it, though.

>> What you're seeing is the `buffer-offer-save' mechanism, which kicks
>> in both when you run `save-some-buffers' and before you kill Emacs.
>> So there's sort of two potential save mechanisms when you exit Emacs.
>> If that's annoying I could also provide an option for not offering to
>> save the database.

[...]

> Isn't it peculiar to EBDB that Emacs offers to save the
> computer-generated contents of the *EBDB* buffer on exit? IIRC, when I
> exit Emacs it does not offer to save my Org Agenda (only the actual Org
> files from which it generated the agenda), nor Dired buffers, nor any
> other computer-generated buffer.

This is an error, either in Emacs or EBDB, I'm not sure. I originally
implemented the ability for special-mode buffers to offer themselves as
saveable, but the idea was that, if they saved themselves using their
own function, Emacs wouldn't then *also* save the buffer itself in a
file.

Either my code in Emacs or my code in EBDB has a bug, I haven't gone and
figured out which. I'll try to do that soon.

Eric