Re: Regression in v2.26.0-rc0 and Magit

[Junio C Hamano <gitster@pobox.com>]

Junio C Hamano <gitster@pobox.com> writes:

> Junio C Hamano <gitster@pobox.com> writes:
> ...
> If nobody complains in the coming 4 hours or so, I'll squash this in
> to e6c57b49 ("prefix_path: show gitdir if worktree unavailable",
> 2020-03-02) and mark the topic as "ready for 'next'".
>
> Thanks.
>
>  t/t6136-pathspec-in-bare.sh | 30 ++++++++++++++++++++++++++++++
>  1 file changed, 30 insertions(+)
> ...
> +test_expect_success 'log and ls-files in .git directory' '
> +	(
> +		cd .git &&
> +		test_must_fail git log -- .. &&
> +		test_must_fail git ls-files -- ..
> +	) >out 2>err &&
> +	test_i18ngrep "outside repository" err
> +'
> +
> +test_done

This is outside the scope of fixing the regression e0020b2f
("prefix_path: show gitdir when arg is outside repo", 2020-02-14)
brought in, but I wonder if this last piece should even fail in the
first place.

If you give "." instead of ".." to these commands, they behave as if
we did so from the top-level of the working tree, i.e. these are
equivalent:

    git -C .git ls-files -- .
    git -C .git/info/ ls-files -- .
    git ls-files -- .

which somehow does not sound quite right, but that is how tools
written in the past 15 years expect and is hard to change?

That does not still explain why Magit (which is sufficiently mature)
is expecting "cd .git && ls-files .." to show the entire working
tree, though.