From: Protesilaos Stavrou <info@protesilaos.com>
To: Philip Kaludercic <philipk@posteo.net>
Cc: Noboru Ota <me@nobiot.com>, ~protesilaos/denote@lists.sr.ht
Subject: Re: [PATCH] Display context of identifier in backlinks buffer with
 xref
In-Reply-To: <87sfhf0xp8.fsf@posteo.net>
References: <86ilk6uzf4.fsf@nobiot.com> <87zgdi6idt.fsf@protesilaos.com>
 <87sfhf0xp8.fsf@posteo.net>
Date: Sat, 17 Dec 2022 05:58:54 +0200
Message-ID: <87tu1ud6kh.fsf@protesilaos.com>
MIME-Version: 1.0
Content-Type: text/plain

> From: Philip Kaludercic <philipk@posteo.net>
> Date: Fri, 16 Dec 2022 22:51:15 +0000

Good day Philip,

> Protesilaos Stavrou <info@protesilaos.com> writes:
>
>> About Emacs 27, I think it is better we avoid all those workarounds and
>> just depend on Emacs 28.  No-one reported the error which indicates that
>> this new requirement should not pose any problems.
>
> I just saw your announcement on your blog, so I am not sure if I am
> missing something.  That being said, have you considered using xref from
> GNU ELPA?  It is a core package but maintains compatibility back until
> Emacs 26.  Of course a line must be drawn, but it would be a pity to
> drop compatibility if it weren't necessary to do so, especially
> considering that Debian Stable still distributes 27.1.

The best case scenario is to support Emacs 27.  The problem was that
Denote was broken for a while on Emacs 27 and we never heard anything
about it.  This led me to believe that no-one is using it there.
Providing a broken package is not nice for users and it also reflects
poorly on us, as we try to deliver something usable.

Xref was not the only reason.  I have had other cases in the recent past
where I was caught off guard.  Compare, for instance, the arguments of
'mapconcat' between Emacs 27 and 29, respectively:

    (mapconcat FUNCTION SEQUENCE SEPARATOR)

    (mapconcat FUNCTION SEQUENCE &optional SEPARATOR)

So I write it without the optional separator, get no warnings, and then
my code is broken on Emacs 27...

The problem I have is that I do not know how to catch those cases, other
than testing every little thing on the older versions.  A fully fledged
test suite would probably be the real solution, though I don't know how
to do this yet.

As for Debian specifically, note that we were requiring 27.2 because it
provided some Org-related feature that is not in 27.1 (this is what the
linter was saying).

If we can address those, then I am happy to do it.  There is no
principled opposition to this cause.

All the best,
Prot

-- 
Protesilaos Stavrou
https://protesilaos.com