Re: Is it time to drop ChangeLogs?

[John Wiegley]

>>>>> Stefan Monnier <address@hidden> writes:

> In theory it's possible, yes. But Someone™ needs to figure out how to do it
> efficiently enough, and then Someone™ needs to actually do it.

I'm willing work on tooling for a satisfactory mechanism to allow us to
eliminate the need for a separate ChangeLog file. I'm volunteering to be that
Someone™, and to afterwards maintain it and field bugs in its use. I would
also like to work directly with Eli and others to ensure it satisfies their
needs and concerns.

We know we need the following:

  1. That commit messages be written in ChangeLog form.  Regardless.

  2. A file-based mechanism to amend commit messages, by allowing substitute
     commit messages. I prefer a single file, or set of files, rather than one
     file per commit.

  3. That the mechanism in #2 work with merges, cherry-picking, rebasing, etc.

     (One way this could be achieved: base amendments on an SHA256 of the
     commit message+date being replaced, rather than the Git hash).

  4. A way to generate a proper ChangeLog from #1 and #2, without taking too
     long, and without requiring Emacs to be built before-hand.

  5. That "proper" history presented in #4 must also be available to other
     tools, such as `vc-region-history', etc.

  6. An easy way for contributors to craft ChangeLog entries for commit
     messages. I mention this because I saw issues mentioned with the current
     C-x 4 a.

I'll also take a look at what coreutils is already doing, but I share Eli's
reservation about a Perl-based process for correcting messages. I prefer
something dead simple, as a plain text file that anyone can edit and commit.

-- 
John Wiegley                  GPG fingerprint = 4710 CF98 AF9B 327B B80F
http://newartisans.com                          60E1 46C4 BD1A 7AC1 4BA2

signature.asc
Description: PGP signature