Re: Should we restore manually maintained ChangeLogs

[Stefan Monnier]

> You (and some others) say the format and the content in the log
> messages are important, and I agree.  But if we do care about them,
> how can we NOT clean them up?  Having them in their current state
> means they cannot be trusted, which is worse than not having them at
> all.

For people like me who solely rely on the VCS commit messages when doing
forensics (because I find it tremendously more powerful, e.g. via
vc-region-history or vc-annotate), it's already the case that mistakes
can't be corrected, no matter what we do with ChangeLog files.

So for my own personal self, the issue is not "should we keep ChangeLog"
but "should we try and fix mistakes in commit messages":
- either we do care about fixing the commit messages.  In this case
  either we need to switch to a system that lets us review changes
  before they get to master, or we figure out a way to update commit
  messages after the fact (this is clearly a missing feature in Git and
  there's no reason it can't be implemented).
- or we don't care enough, so we keep doing as we've been doing so far,
  i.e. relying on "best effort and education" to try and keep the
  mistakes to a tolerable level.

So far we've been following a halfway path where we tolerate mistakes in
the commit messages but not in the ChangeLog file, probably because
there's a fairly strong correlation between those who care about
mistakes in those messages and those who cares about the
ChangeLog files.

But really, fixing the ChangeLog while leaving the VCS commit messages
unfixed is just not a good solution (to me, it's kind of like installing
security fixes in the tarball rather than in the VCS repository).

For those reasons, I think ChangeLog files should be 100%
auto-generated, and never committed into the VCS.

> Yes, a few of us don't need any corrections.  But enough of us do, and
> that's where the problem lies.  It is that problem that we need to
> fix.  Leaving it unfixed makes your accurate work unreliable as well.

Right.  But if we need/want to fix it, we should do it *at the source*
(i.e. in the VCS metadata) rather than only in some of the derived data
(the ChangeLog files).


        Stefan