Re: [PATCH] Do not record unstaged deleted file upon recursive merge if file was moved outside of working tree with enabled sparse-checkout.

From: Junio C Hamano <gitster@pobox.com>
Date: Sep 11, 2016, 23:24 UTC
Message-ID: <xmqqeg4q6mw4.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <1473625505-7641-1-git-send-email-mikhail@filippov.me>

Mikhail Filippov <mikhail@filippov.me> writes:

> ---

You'd need a lot more explanation on why this is needed (i.e. without it what behaviour you would get, and why that behaviour is wrong).

Show 21 quoted lines

>  merge-recursive.c                    |  9 +++++---
>  t/t6042-merge-rename-corner-cases.sh | 42 ++++++++++++++++++++++++++++++++++++
>  2 files changed, 48 insertions(+), 3 deletions(-)
>
> diff --git a/merge-recursive.c b/merge-recursive.c
> index e349126..25dc701 100644
> --- a/merge-recursive.c
> +++ b/merge-recursive.c
> @@ -1724,9 +1724,12 @@ static int merge_content(struct merge_options *o,
>  		 */
>  		path_renamed_outside_HEAD = !path2 || !strcmp(path, path2);
>  		if (!path_renamed_outside_HEAD) {
> -			add_cacheinfo(o, mfi.mode, &mfi.oid, path,
> -				      0, (!o->call_depth), 0);
> -			return mfi.clean;
> +			struct stat st;
> +			if (lstat(path, &st) == 0) {
> +				add_cacheinfo(o, mfi.mode, &mfi.oid, path,
> +					      0, (!o->call_depth), 0);
> +				return mfi.clean;
> +			}

I cannot guess what you are trying to achieve without explanation in the proposed log message, but I can say that this unconditional checking of a working tree file cannot be correct (there may or may not be other things that are wrong with this change, which cannot be judged without more information).

Imagine your o->call_depth is not zero, i.e. we are making a virtual common ancestor with this merge, in which case any of the three trees involved may have nothing to do with the current working tree files.

Show 5 quoted lines

> +test_expect_success 'move file/sparse-checkout/merge should not delete moved file' '
> +	git rm -rf . &&
> +	git clean -fdqx &&
> +	rm -rf .git &&
> +	git init &&

Yuck. This is inherited from existing tests but I think they need to be cleaned up. It is not your fault, and it is not in the scope of this change.

> +	git status >output &&
> +	cp output /tmp/a &&

Huh?

> +	test_i18ngrep "nothing to commit" output
> +'
> +
>  test_done