Format for JDK 6/7 changeset comments?

[John Coomes]

Mark Reinhold (mr at sun.com) wrote:
> > Date: Tue, 04 Dec 2007 13:25:52 -0800
> > From: kelly.ohair at sun.com
> ...
> >    - Will we check the bugid to make sure it's a valid bugid?
> 
> That requires doing a bug-database query which can, with the current
> system anyway, be kind of slow, and impossible if you're not connected.
> (We want the hg script that implements these checks to be usable by any
> developer creating changesets, so that changesets can be validated
> locally before being pushed up to group integration forests.)
> 
> We could do this check just on the hg server side, or just when the
> machine running the check is actually connected.  We could also, in the
> longer term, play various caching tricks to speed up the query.
> 
> If we did do such a check it'd have limited value.  It could only verify
> that the bugid is associated with the appropriate release, which is worth
> something I suppose but not as thorough a check as one might like, given
> the cost.
> 
> So at the moment I'm leaning against checking bugids mechanically.

Since our (currently internal-only) build automation tools go to a lot
of trouble to automatically update the bug database, I think a basic
sanity check is worthwhile.  Even if it just catches typos initially,
getting the bug numbers right will save time for integrators and those
that have to do bug archaeology.

Gracefully handling the non-connected state would be a nice addition.
Personally I could live without it, certainly for a while.

> >    - Can the Summary line be any length?
> 
> Hmm, good question.  I suppose we could limit it to prevent mistakes or
> abuse.  80 characters?  200?  Or allow some number of continuation lines,
> as in e-mail headers, that start with whitespace?  (That might be more
> readable.)

I'd prefer to remove the temptation to create excessively long summary lines, so either a small limit or continuation lines.  Please :-).

-John