Preparing 21.4-1

[Jérôme Marant]

Quoting Rob Browning <rlb@defaultvalue.org>:

> > I think we can find a consensus about handling autofiles. If you
> > agree that we can restrict this to manual updates, I'll be OK
> > for such a solution. I don't have any ideological thought on
> > this.
>
> > I hope you're not fed up with our recent discussion :-P
>
> Nope.  I realize the autofiles.diff handling is one of the "trickier"
> bits in the emacs package, so it's worth discussing.  It's a messy
> problem in the first place because the use of dpatch and the use of
> the autotools means that you may have some patches which affect the
> content of another in somewhat unpredictable ways.

I ask the dpatch author about how he would handle such cases. I'll
forward the reply as soon as get it ; it might be good to have an
external sight.

> > Also, it would be nice to apply this AMD64 patch (which requires
> > changes to configure.in).
>
> Which one?

The last one from the BTS, which was recently discussed. I can
manage this.

> > I have small fixes around, and I'll try to see if I can backport
> > perl-mode from the CVS trunk, which would fix many bug.
>
> Hmm.  How big a change is that likely to be?  If it's likely to change
> the public interface of perl-mode (var names, functions, etc.), then
> I'd be hesitant.  It would probably be better to backport individual
> fixes if possible, or just mark those bugs as fixed-upstream.
> Although I'd love to just fix things in 21.4, I think it's also
> important for people to be able to depend on the fact that a given
> version of emacs behaves a certain way.

As far as I can tell there have only been bug fixes to perl-mode
and no new features (I'll recheck before doing anything).
In that case, it is better to patch the whole mode instead of
releasing small patches.

> For example, I wouldn't be comfortable backporting the current gnus
> into emacs21.  IMO that should be handled via an add-on package you
> have to intentionally install.

Sure.

> Of course it's a very fuzzy line -- *anything* you fix might be a bug
> someone was expecting (and accomodating) in a given version of emacs.
> i.e.
>
>   (if (equal? (emacs-version) "21.4)
>     ... do something that depends on bug/feature in 21.4...)
>
> So it'll always be a judgement call, and in any cases where we're
> really not sure, I'd consult emacs-devel to see what they'd prefer.

I wouldn't backport features that depend on upper versions.

--
Jérôme Marant