From: Junio C Hamano <gitster@pobox.com>
Date: Thu, 04 Sep 2008 19:12:40 GMT
Subject: Re: [PATCH] Mention the fact that 'git annotate' is only for backward compatibility.
Message-ID: <7vd4jj3etz.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <20080904123046.GX10544@machine.or.cz>

Petr Baudis <pasky@suse.cz> writes:

> I'm also curious about
>
>                 if (suspect->commit->object.flags & UNINTERESTING) {
>                         if (blank_boundary)
>                                 memset(hex, ' ', length);
>                         else if (!cmd_is_annotate) {
>                                 length--;
>                                 putchar('^');
>                         }
>                 }
>
> in builtin-blame.c. Junio, you introduced this in e68989a739d - why
> do you use a separate flag instead of OUTPUT_ANNOTATE_COMPAT? The fact
> that git annotate == git blame -c does not hold true because of this
> (admittedly obscure case).

I do not recall the context of this change, but I do not think this is a
deliberate omission.  The breakage the quoted commit fixed was about
cvsserver that does run "git-annotate" not "git-blame -c" and that must be
the reason this has been unnoticed for a long time.

Oh, by the way, somebody should update cvsserver to run "git annotate" and
send in a tested patch, please?