Format for JDK 6/7 changeset comments?

[Mark Reinhold]

> Date: Thu, 08 Nov 2007 17:12:08 -0800
> From: andreas.sterbenz at sun.com

> Mark Reinhold wrote:
>> No.  Authorship information should go with the code.
> 
> Exactly because that information is important it should be stored in the
> bug database, where all other information about the bug is stored [*], the
> location of which is well known to everyone and not just the developers
> working on the JDK,

Why does anyone who doesn't know how to navigate the Mercurial web
interface need to find this information?

In what sense is the location of the bug database better known than that
of the Mercurial repositories?

>                     which can be queried using a web interface, etc.

We have a web interface for Mercurial, and it's extensible.

> Capturing everything in the bug database except code reviewers does not
> seem like a good idea to me.

The logical conclusion of this line of reasoning is that a changeset
shouldn't even carry an author name or a bug number or a comment; we
should just put the changeset hash into the relevant bug-database
records.

That seems wrong.

> It is probably a bit more difficult to ensure that the correct information
> is in the bug database than to validate the changeset comments, but it
> should be possible for the hooks to query it and verify that the relevant
> information is present (e.g. must have a comment that includes the text
> "Reviewed-by: X").

Possible, yes, but hardly convenient.  This approach would make it
impossible to validate (prospective) changesets in a local repository
when you're not online.

- Mark