Re: How to populate index/worktree when recursive merge merges multiple common ancestors?

From: Shawn Pearce <spearce@spearce.org>
Date: Jun 30, 2014, 21:30 UTC
Message-ID: <CAJo=hJsVSqvEB4wnSp_RtfF3rMXyLWjm8kUXHxmOWmzYn6MM2Q@mail.gmail.com>
In-Reply-To: <CAENte7i3f_zvPmnPZe2s7guPv03o1d_Ln+nVaj5PYmP8uoK=rw@mail.gmail.com>

On Mon, Jun 30, 2014 at 5:48 AM, Christian Halstrick <christian.halstrick@gmail.com> wrote:

Show 9 quoted lines

>> They don't. The conflicts are preserved into the virtual ancestor. The
>> user only sees the final conflicts during merging of A and B with
>> virtual X3 as the common ancestor.
>
> Ah, now I understand. When I merge X1 and X2 into the virtual X3
> I should not stop if this is not doable without conflict resolution. I
> should store in memory the X3 content, including all the conflict
> markers. If I finally merge A and B I will specify a common base
> content which may contain conflict markers. Right?

Yes.

If X3 content is large, it could flow to loose objects on disk. These will be unreachable and cleaned up automatically in a future garbage collection.

> Are git config param's like merge.conflictstyle=diff3 are also
> effective when creating the virtual X3 content? Couldn't that lead to
> complicated conflict marker situations? In the area where you expect
> common base content you again see conflict markers in diff3 style.

Yes, but I think this is the correct behavior. The machine can't reconcile the two branches any better than this, so now a human has to step in and fix all of the conflicts.

IIRC, this is uncommon. Usually you use A's common content as A and B do not differ relative to X3 in the regions where X3 has a conflict, so those conflicts aren't considered relevant when A and B merge.