Re: [PATCH] Add committer and author names to top of COMMIT_EDITMSG.

From: Stephen Sinclair <radarsat1@gmail.com>
Date: Jan 12, 2008, 00:09 UTC
Message-ID: <9b3e2dc20801111609t3103af1frc23519cab43ae8be@mail.gmail.com>
In-Reply-To: <7v3at42avd.fsf@gitster.siamese.dyndns.org>

Show 5 quoted lines

> I'd almost agree with this patch if if added AUTHOR but not
> COMMITTER, and only when AUTHOR is different from me.  That
> would help reassure anybody while amending other's changes.
> COMMITTER is always me and I should not reminded with extra
> lines that waste precious screen real estate.

The purpose of my patch was to remind _myself_ what my name is, in case I hadn't configured it correctly. But I can see this use case also being useful.

Show 5 quoted lines

> And no, I did not check if your change correctly supports the
> use case of amending other's changes.  But if I recall the code
> correctly, I suspect that your change doesn't.  The recorded
> author is determined after the log message is prepared, way
> later.

Sure, that's possible.

> I strongly agree with Dscho that this change needs to be
> defended with a good description on the reason why this is good.

The patch was really to go along with my RFC about the idea. I guess it was too early to post a possible implementation. (I have only just begun to look at the git code after all..)

> If the reason is "newbie protection", I do not think this is a
> good change at all.  Newbie protection is never a good reason to
> make people who graduated that state to pay extra price
> unconditionally.

I agree. I wouldn't necessarily say it is "newbie protection", so much as a friendly reminder of what username you are using while doing a commit, which, as I said, might not be as expected if you have just sat down at a new machine. Especially if you have been using git for a long time on a single machine, it is something you might easily forget to configure. (As I have, several times now.) I agree, however, that this it is not necessarily worth having this on the screen every time you do a commit, for the exceptional instance where it might be wrong. Perhaps more usefully it could appear only if you haven't yet created a user.email and user.name config entry.

Actually in my honest opinion, the default of using the computer's host name and login is pretty much _never_ right, but I thought this patch might be less intrusive than introducing a new error message.

I do have a slightly better patch now that has a more informative message and uses git_committer_info() and git_author_info(), however I'll wait for any more opinions before posting it.

In retrospect, I guess I could just as easily solve my problem by introducing a post-receive hook for my personal repo that issues a warning for commits not configured to my email address.

Steve