From: Jonathan Nieder <jrnieder@gmail.com>
Date: Tue, 22 May 2012 23:38:21 GMT
Subject: Re: [PATCH/RFC] rebase -p: do not redo the merge, but cherry-pick first-parent changes
Message-ID: <20120522233821.GA31667@burratino>
In-Reply-To: <4FBAA33D.1080703@kdbg.org>

Johannes Sixt wrote:

> - The 'git merge' invocation passed the commit message of the old merge
>   commit, but it still obeyed the merge.log option. If it was set, the log
>   ended up twice in the commit message.

Likewise for the "Conflicts:" paragraph, which should also be pretty
simple to fix directly.  I don't think that points directly to a
fundamental flaw in the current implementation, but it does make it
seem like no one has been using "rebase -p" for anything nontrivial. :)

[...]
> --- a/git-rebase--interactive.sh
> +++ b/git-rebase--interactive.sh
> @@ -312,12 +312,14 @@ pick_one_preserving_merges () {
>  			msg_content="$(commit_message $sha1)"
>  			# No point in merging the first parent, that's HEAD
>  			new_parents=${new_parents# $first_parent}
> -			if ! do_with_author output \
> -				git merge --no-ff ${strategy:+-s $strategy} -m \
> -					"$msg_content" $new_parents
> +			printf "%s\n" $new_parents >"$GIT_DIR"/MERGE_HEAD
> +			printf "%s\n" "$msg_content" >"$GIT_DIR"/MERGE_MSG
> +			if output git cherry-pick -m 1 -n "$sha1"
>  			then
> +				do_with_author output git commit --no-verify -F "$GIT_DIR"/MERGE_MSG ||
> +					die_with_patch $sha1 "Could not replay merge $sha1"
> +			else
> -				printf "%s\n" "$msg_content" > "$GIT_DIR"/MERGE_MSG
> -				die_with_patch $sha1 "Error redoing merge $sha1"
> +				die_with_patch $sha1 "Could not pick merge $sha1"
>  			fi

Cute.  The approach looks very sensible to me.

Thanks,
Jonathan