From: ZheNing Hu <adlternative@gmail.com>
Date: Tue, 23 Mar 2021 06:29:01 GMT
Subject: Re: [PATCH v13] [GSOC] commit: add --trailer option
Message-ID: <CAOLTT8S09oTZYjDArMNhQSLAaHtikrw=H4Z8WHF9GvWukQytaQ@mail.gmail.com>
In-Reply-To: <CAP8UFD275WiHC2sUJjsnLd1yonfMO-SVda=BZ6mMkgDxGs959g@mail.gmail.com>

Christian Couder <christian.couder@gmail.com> 于2021年3月23日周二 上午5:55写道：
>
> On Mon, Mar 22, 2021 at 11:23 AM ZheNing Hu <adlternative@gmail.com> wrote:
>
> > 2.
> > `git interpret-trailers --in-place`  seem like work on git top-dir,
> > If I am in a sub-dir `b` and I want to change a file such as `d.c`,
> > then I must use `git interpret-trailers --in-place b/d.c` to add some
> > trailers.
>
> What happens without --in-place? Are the input files read correctly?

It's still wrong.
the git die() in `read_input_file` of "trailer.c".

>
> > I think the original intention of `--in-place` is to modify a file similar to
> > "$COMMIT_MSG_FILE", so make it run at top-dir, but this is not reflected
> > in the git documentation. This at least confuses people who use this
> > option for the first time. Is it worth modifying? Or is there something
> > wrong with the design of `--in-place`?
>
> I haven't checked but there is perhaps a bug in
> create_in_place_tempfile() in "trailer.c".

I haven't check finished. But I think it may do something like `chdir()`.

Thanks.