Re: git-visualize(1) plumbing equivalent

From: Alejandro Colomar <alx@kernel.org>
Date: Oct 3, 2026, 15:13 UTC
Message-ID: <asEa_Lp01DVJ2ThZ@debian>
In-Reply-To: <CALnO6CCUY2KF8rEotihdVNy+D10mmzuW0RXbH8nqhSS9jsgQUg@mail.gmail.com>

Hi Ben,

Show 20 quoted lines

> Date: 2026-10-03 09:45:17-0400
> From: "D. Ben Knoble" <ben.knoble@gmail.com>
>
> On Sat, Oct 3, 2026 at 6:13 AM Alejandro Colomar <alx@kernel.org> wrote:
> >
> > Hi!
> >
> > I have this code, which I use to loop in a script using git-bisect(1):
> >
> >         while
> >                 git bisect visualize --oneline \
> >                 | wc -l \
> >                 | xargs -I{} test {} -gt 1;
> >         do
> >                 if
> >                         git rebase $gropts BISECT_HEAD >/dev/null 2>/dev/null;
> >                         test $? -eq 0;
> 
> Aside: this "test $? -eq 0" is a bit redundant, no? "if cmd" in shell
> works by checking whether "cmd" exits 0 or not.

Ouch! Indeed. IIRC, I originally had the test reversed (-ne 0), and later probably flipped without thinking it could be removed. :)

Show 32 quoted lines

> 
> >                 then
> >                         echo 'Rebase: success';
> >                         git bisect good BISECT_HEAD;
> >                 else
> >                         echo 'Rebase: conflict';
> >                         git rebase --abort >/dev/null;
> >                         git bisect bad BISECT_HEAD;
> >                 fi;
> >         done;
> >
> > (
> > I know this resembles "git rebase run", but I'm avoiding it, because
> > passing all of that as a command is non-trivial (and I'd like to avoid
> > having to write a separate script to pass its name to "git bisect run").
> > )
> >
> > Having read the documentation for git-bisect(1), visualize reads several
> > environment variables, and thus this code doesn't seem robust.  What
> > would be the plumbing version of the while-loop condition?
> >
> >                 git bisect visualize --oneline \
> >                 | wc -l \
> >                 | xargs -I{} test {} -gt 1;
> >
> > The goal is to know whether git-bisect(1) has found a commit yet or not,
> > to stop looping.
> 
> I think you are probably looking for the (size of the) set of commits
> between bisect/bad and all the bisect/good-* refs. So you might need
> to "git refs list" the good ones, and feed those as negated refs
> alongside bisect/bad to rev-list?

Yup, this seems to work:

git refs list | grep refs/bisect/ | sed '/good/s/^/^/' | cut -f1 -d' ' | xargs git rev-list

> 
> In the general case, that wouldn't account for skipped commits as I
> understand it, where multiple commits are left at the end of the
> bisect, but in your script it doesn't look like you skip any.

Hmmmm. I'm now working on adding the ability to skip commits, so this would be a problem. Do you have any idea on how to deal with that?

Have a lovely day! Alex

> 
> -- 
> D. Ben Knoble

-- 
<https://www.alejandro-colomar.es>