Re: Is it time to drop ChangeLogs?

[Ingo Lohmar]

On Wed, Mar 09 2016 22:06 (+0200), Eli Zaretskii wrote:
>> That's pretty much what I meant.  It may not contain the information you
>> (and I) would like to see in it, but it is highly reliable in that it is
>> the message that was used to describe the commit.
>
> Yeah, it's a reliable lie.  How helpful is that?

This discussion is about the content, not the technical merits of
immutability.  As I said, I am happy to drop the nitpicking.

>> I make a lot of mistakes in commit messages, and I do not agree that it
>> is difficult to fix them before pushing.
>
> Indeed, it isn't difficult.  But since mistakes do get pushed, I guess
> people are not careful enough to review their commits before pushing,
> or else we wouldn't be having this conversation, and there would be no
> need for the procedures to correct pushed commits with mistaken log
> messages.

It happens to the most careful people, correct.

> Telling the wrongdoer to clean up their mess surely does teach them a
> lesson.  It could well be the lesson we both agree is the best one:
> review your commits before pushing.
>
> I hope you agree that NOT fixing the mistakes teaches them an entirely
> different lesson, the one we think is undesirable: that mistakes don't
> matter, and eventually that the log messages don't matter, as long as
> they say _something_.

We completely agree on the best lesson and the importance of fixing
mistakes.  We probably disagree on what is a good way to actually teach
them the lesson.  We also seem to disagree on good ways of "fixing", or
ways to avoid the need for after-the-fact fixing.

>> I cannot speak for anybody but myself.  When I am a user of Emacs,
>> trying to get an idea of what has changed on the level of detail that
>> the Changelog provides, I would rather go to the authoritative source,
>> if online (without any repository) at
>> http://git.savannah.gnu.org/cgit/emacs.git.
>
> I guess you never had to work in segregated networks, where access to
> the outside world is not available.

In any case, I was speaking mostly hypothetically.  I am either
interested in a high-level description of some changes that might affect
me, or cool new things to try, in which case I look at NEWS.  Or I am
interested in details of what a function does differently now, or a
changed call signature or whatever --- in that case, I have never found
the Changelog to be sufficiently detailed.  Instead I will look (on the
next possible occasion) at the (commits that changed the) source code.