Re: Possible git-rev-list bug

From: Junio C Hamano <junkio@cox.net>
Date: Jan 29, 2006, 20:15 UTC
Message-ID: <7vmzhekcz3.fsf@assigned-by-dhcp.cox.net>
In-Reply-To: <43DC8DDF.6080904@yahoo.it>

Marco Costalba <mcostalba@yahoo.it> writes:

Show 9 quoted lines

> $ git-rev-list --max-count=1 --parents addafaf92eeb86033da91323d0d3ad7a496dae83 -- rev-list.c
> addafaf92eeb86033da91323d0d3ad7a496dae83
> d8f6b342ae200b2eb72e2f81afea7fe0d41aec0b
> 93b74bca86f59b8df410b6fd4803b88ee0f304bf
> d8f6b342ae200b2eb72e2f81afea7fe0d41aec0b
> d8f6b342ae200b2eb72e2f81afea7fe0d41aec0b
> 3815f423ae39bf774de3c268c6d3e3b72128a4e5
>
> We have the same parent (d8f6b342ae200b2eb72e2f81afea7fe0d41aec0b) multiple times.

I think it probably is a bug. The commit really has five parents that bundle five independent tracks of development. Two of them touch the specified path since they forked. The rest do not, and for them, the last commit that touched the specified path is behind where they forked:

                  .---o---.
                 /         \
                .---*---o---.
               /   93b74bc   \
  ---*---o---o-----o---o------o addafaf
     d8f6b34   \             /
                .---o---o---.
                 \         /
                  .---*---.
                      3815f42

The "rev-list.c" path limiter is meant to omit "uninteresting" commit, so in this case showing only the three parents after filtering dups would be the right thing to do, I think.

Let me wait for a while to hear Linus contradicts me, though...