From: Chen Linxuan <me@black-desk.cn>
Date: Mon, 25 May 2026 09:00:23 GMT
Subject: Re: [PATCH v5 0/2] includeIf: add "worktree" condition for matching working tree path
Message-ID: <CAC1kPDPbyxs-aTrAOi_PNTZF7EApG31iLYwm+Eddpeh2hT8a-w@mail.gmail.com>
In-Reply-To: <xmqqjysseyid.fsf@gitster.g>

On Mon, May 25, 2026 at 3:31 PM Junio C Hamano <gitster@pobox.com> wrote:
>
> Chen Linxuan via B4 Relay <devnull+me.black-desk.cn@kernel.org>
> writes:
>
> > Changes in v5:
> > - Fix Windows CI failure: use `**` glob pattern instead of `/` in the
> >   "worktree without repository" tests, since `/` as a path pattern is
> >   Unix-specific and does not match Windows paths.
>
> Would it have worked if you used something like "[/\\].path",
> instead of "/.path", to cover directory delimiters for both systems?
>
> I am not asking to make further changes.  I am trying to understand
> what the extent of the problem was.

The root cause is that on Windows,
strbuf_realpath() returns paths with a drive letter prefix (e.g.
D:/a/git/...), which does not start with /.

Here is the trace output from the Windows CI [1]:

include_by_path: text='D:/a/git/git/t/trash
directory.t1305-config-include', pattern='/**', prefix=0

The pattern worktree:/ becomes /** after add_trailing_starstar_for_dir().
Then wildmatch("/**", "D:/a/git/...", WM_PATHNAME) fails because the text
does not start with /.

[1] https://github.com/black-desk/git/actions/runs/26391768962/job/77683708185