From: Pushkar Singh <pushkarkumarsingh1970@gmail.com>
Date: Sat, 03 Jan 2026 11:34:54 GMT
Subject: Re: bug report: git status -z doesn't respect status.relativePaths=true
Message-ID: <CALE2CrQ9OmCprNckT5qMwx87NZZfKTe0C6bfD76d4tyUFDgJgQ@mail.gmail.com>
In-Reply-To: <20260103112642.GA2706421@coredump.intra.peff.net>

Hi Jeff,

Thanks for the detailed explanation. That clarifies why `-z` implies
`--porcelain=v1` and why `status.relativePaths` gets ignored in that case.

I agree that the `--short -z` behavior is surprising. From the
documentation, I would also expect `status.relativePaths=true` to be
honored there, so ignoring the prefix entirely does look like a real
issue rather than just a quirk.

The `--porcelain=v2 -z` case is interesting as well. If v2 is intended
to respect the configuration, then the inconsistency when using `-z`
does seem like a related bug.

If it would be useful, I am happy to help investigate further or work
on tests once there is consensus on the intended behavior.

Thanks,
Pushkar

On Sat, Jan 3, 2026 at 4:56 PM Jeff King <peff@peff.net> wrote:
>
> On Sat, Jan 03, 2026 at 02:47:36AM -0800, Artur Pyrogovskyi wrote:
>
> > According to the man page of git-status: "-z Terminate entries with
> > NUL, instead of LF."
> >
> > However, it ignores status.relativePaths=true and always shows absolute paths.
>
> This matches the documented behavior. In --porcelain=v1 mode, we ignore
> most configuration options (like status.relativePaths). And -z puts us
> into v1 porcelain mode by default:
>
>        -z
>            Terminate entries with NUL, instead of LF. This implies the
>            --porcelain=v1 output format if no other format is given.
>
> I do think there is at least one bug here, though. I'd expect from that
> documentation to be able to do:
>
>   git -c status.relativePaths=true status --short -z
>
> and get relative paths. But it doesn't work. The --short output code
> that handles "-z" ignores the prefix entirely.
>
> Something like this would fix it:
>
> diff --git a/wt-status.c b/wt-status.c
> index e12adb26b9..22797371a6 100644
> --- a/wt-status.c
> +++ b/wt-status.c
> @@ -2005,6 +2005,13 @@ static void wt_shortstatus_unmerged(struct string_list_item *it,
>         }
>  }
>
> +static void print_with_nul(struct wt_status *s, const char *fn)
> +{
> +       struct strbuf scratch = STRBUF_INIT;
> +       fprintf(s->fp, "%s%c", relative_path(fn, s->prefix, &scratch), 0);
> +       strbuf_release(&scratch);
> +}
> +
>  static void wt_shortstatus_status(struct string_list_item *it,
>                          struct wt_status *s)
>  {
> @@ -2020,9 +2027,9 @@ static void wt_shortstatus_status(struct string_list_item *it,
>                 fputc(' ', s->fp);
>         fputc(' ', s->fp);
>         if (s->null_termination) {
> -               fprintf(s->fp, "%s%c", it->string, 0);
> +               print_with_nul(s, it->string);
>                 if (d->rename_source)
> -                       fprintf(s->fp, "%s%c", d->rename_source, 0);
> +                       print_with_nul(s, d->rename_source);
>         } else {
>                 struct strbuf onebuf = STRBUF_INIT;
>                 const char *one;
>
> but:
>
>   1. it would need similar adjustments to a few other printing
>      functions; and
>
>   2. it's not clear to me if we want to change this behavior or keep
>      it as a historical quirk and document it as such.
>
>      I guess nobody noticed because using "-z" with "--short" is a bit
>      of an odd thing to want to do.
>
>
> There's another oddity, which is this:
>
> > Repro steps:
> > $ mkdir test-repo && cd test-repo && git init .
> > $ mkdir subdir && touch subdir/test-file.txt && cd subdir && git add
> > test-file.txt
> > $ git -c status.relativePaths=true status --porcelain=2
> > 1 A. N... 000000 100644 100644
> > 0000000000000000000000000000000000000000
> > e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 test-file.txt
>
> I am surprised to see that the v2 porcelain respects that config option
> at all. I don't know if that's a bug, or a subtle change from v1. I
> don't see it mentioned in the documentation.
>
> If it isn't a bug, and we expect v2 porcelain to respect the config,
> then this:
>
> > $ git -c status.relativePaths=true status --porcelain=2 -z
> > 1 A. N... 000000 100644 100644
> > 0000000000000000000000000000000000000000
> > e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 subdir/test-file.txt%
>
> is another bug similar to the --short one.
>
> -Peff
>