git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 2/2] branch: allow recursion with no tracking name

From
D. Ben Knoble <ben.knoble@gmail.com>
Date
Nov 6, 2026, 14:10 UTC
Message-ID
<CALnO6CDqtVuUqm0xW7Cd=MVbDSaa0Q+T6D9rx_CL7jjUbeb=DQ@mail.gmail.com>
In-Reply-To
<20260822-vv-branch-recurse-no-start-ref-v1-2-46dc140acaa8@zitro.id>
On Fri, Aug 21, 2026 at 7:04 PM Volodymyr Vriukalo <0@zitro.id> wrote:
Show 32 quoted lines
>
> Creating a branch across submodules from a commit that no ref points
>   at fails, with the helper's usage text reprinted as an error:
>
>     submodule 'sub': usage: git submodule--helper create-branch [...]
>     fatal: submodule 'sub': cannot create branch 'branch-a'
>
> `submodule_create_branch()` runs the helper in a child process because
>   `install_branch_config_multiple_remotes()` cannot write config into a
>   submodule, and passes the branch name, the start oid and the tracking
>   name as three positionals.
> `dwim_branch_start()` leaves the tracking name NULL where the start
>   point named no ref, and `strvec_pushl()` stops at the first NULL, so
>   the child receives two positionals.
> `module_create_branch()` requires exactly three and prints its usage.
>
> Make the third positional optional, since a start point that named no
>   ref has no tracking name to give and the recursion has nothing to
>   track in the submodule either.
> Push it separately in the caller too: relying on `strvec_pushl()` to
>   stop early leaves the argument dropped by accident rather than by
>   intent, and a reader has to know where the terminator falls to see
>   that it can go missing at all.
>
> 961b130d20 (branch: add --recurse-submodules option for branch
>   creation, 2022-01-28) introduced both sides.
>
> This is the same NULL tracking name as the previous patch, reached one
>   step earlier: the dry-run pass over the submodules runs before the
>   superproject's own `setup_tracking()` call, so with a submodule
>   present this failure hides the abort that patch removes.
> The new test therefore needs that patch under it.
Right, with the previous patch, we get a little further and then still fail.

I think we should also simplify this message quite a bit: creating a submodule branch may not always have tracking information, and we should omit it when there is none. An exercise for the reader to rewrite the message ;)

(Nit, we don't typically use hanging indents in our messages.)
Show 23 quoted lines
> Assisted-by: An LLM.
> Signed-off-by: Volodymyr Vriukalo <0@zitro.id>
> ---
>  branch.c                    | 10 +++++++++-
>  builtin/submodule--helper.c |  7 ++++---
>  t/t3207-branch-submodule.sh | 13 +++++++++++++
>  3 files changed, 26 insertions(+), 4 deletions(-)
>
> diff --git a/branch.c b/branch.c
> index 182fc4a3dd..2dab1f1e35 100644
> --- a/branch.c
> +++ b/branch.c
> @@ -726,7 +726,15 @@ static int submodule_create_branch(struct repository *r,
>                 break;
>         }
>
> -       strvec_pushl(&child.args, name, start_oid, tracking_name, NULL);
> +       /*
> +        * The tracking name is absent when the start point named no ref.
> +        * Push it separately: strvec_pushl() stops at the first NULL, so
> +        * passing it inline would drop the argument by accident rather
> +        * than by intent.
> +        */

Strictly speaking, this comment is covered by the log message. I agree it's better to be clear about what we mean here rather than punning on tracking_name==NULL, but I don't think future readers of the code will much care how it could have / used to look (and if they do, that's what blame is for).

Show 34 quoted lines
> +       strvec_pushl(&child.args, name, start_oid, NULL);
> +       if (tracking_name)
> +               strvec_push(&child.args, tracking_name);
>
>         if ((ret = start_command(&child)))
>                 return ret;
> diff --git a/builtin/submodule--helper.c b/builtin/submodule--helper.c
> index 1cc82a134d..6895216712 100644
> --- a/builtin/submodule--helper.c
> +++ b/builtin/submodule--helper.c
> @@ -3335,7 +3335,7 @@ static int module_create_branch(int argc, const char **argv, const char *prefix,
>                 OPT_END()
>         };
>         const char *const usage[] = {
> -               N_("git submodule--helper create-branch [-f|--force] [--create-reflog] [-q|--quiet] [-t|--track] [-n|--dry-run] <name> <start-oid> <start-name>"),
> +               N_("git submodule--helper create-branch [-f|--force] [--create-reflog] [-q|--quiet] [-t|--track] [-n|--dry-run] <name> <start-oid> [<start-name>]"),
>                 NULL
>         };
>         struct repo_config_values *cfg = repo_config_values(the_repository);
> @@ -3344,13 +3344,14 @@ static int module_create_branch(int argc, const char **argv, const char *prefix,
>         track = cfg->branch_track;
>         argc = parse_options(argc, argv, prefix, options, usage, 0);
>
> -       if (argc != 3)
> +       if (argc < 2 || argc > 3)
>                 usage_with_options(usage, options);
>
>         if (!quiet && !dry_run)
>                 printf_ln(_("creating branch '%s'"), argv[0]);
>
> -       create_branches_recursively(the_repository, argv[0], argv[1], argv[2],
> +       create_branches_recursively(the_repository, argv[0], argv[1],
> +                                   argc > 2 ? argv[2] : NULL,
>                                     force, reflog, quiet, track, dry_run);

This is of course the function whose tracking_name being NULL we just made more sense out of, good.

Show 21 quoted lines
>         return 0;
>  }
> diff --git a/t/t3207-branch-submodule.sh b/t/t3207-branch-submodule.sh
> index 54f7caeb2f..c56cea31cb 100755
> --- a/t/t3207-branch-submodule.sh
> +++ b/t/t3207-branch-submodule.sh
> @@ -115,6 +115,19 @@ test_expect_success 'should move a branch to a start point that names no ref' '
>         )
>  '
>
> +test_expect_success 'should recurse into submodules from a start point that names no ref' '
> +       test_when_finished "reset_test" &&
> +       (
> +               cd super &&
> +               oid=$(git rev-parse HEAD) &&
> +               git branch --recurse-submodules branch-a "$oid" &&
> +               git rev-parse branch-a &&
> +               git -C sub rev-parse branch-a &&
> +               git -C sub/sub-sub rev-parse branch-a &&
> +               git -C second/sub rev-parse branch-a
> +       )

This test contains no assertions, although the rev-parse's will fail if the branch wasn't created… so I suppose that works. Was a bit jarring at first!

Show 10 quoted lines
> +'
> +
>  test_expect_success 'should ignore submodule.recurse when not creating branches' '
>         test_when_finished "reset_test" &&
>         (
>
> --
> 2.55.0.2.g927b4b9963
>
>
-- 
D. Ben Knoble
Previous: Volodymyr VriukaloNext: Junio C Hamano
Message 5 of 6 in “branch: fix --recurse-submodules with a nameless start point”
  1. 0/2 branch: fix --recurse-submodules with a nameless start pointVolodymyr Vriukalo, Aug 21, 2026
  2. 1/2 branch: do not track a start point with no refVolodymyr Vriukalo, Aug 21, 2026
  3. D. Ben KnobleNov 6, 2026
  4. 2/2 branch: allow recursion with no tracking nameVolodymyr Vriukalo, Aug 21, 2026
  5. D. Ben KnobleNov 6, 2026
  6. Junio C HamanoNov 2, 2026

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.