repos: find submodule git directories the way git does

A submodule git directory which borrows another (commondir) has no
objects of its own, and was skipped instead of refused, because
"has objects/" decided what a git directory was.  The walk now follows
git's layout: under .git/modules every directory holding HEAD is a git
directory (whatever the submodule's name, even `refs`), which must not
borrow, must have its own objects and a regular config, and is
sanitized; nested submodules are found in its own `modules`, and its
refs and logs are never mistaken for git directories.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

modules/repos/lib.py

@@ -443,12 +443,26 @@ def sanitize_git_config(directory: pathlib.Path):
         if path.is_symlink():
             # git would follow it out of the repo.
             raise ValueError(f'{path} is a link')
-    for submodule_git_dir in sorted({i.parent for i in modules.rglob('HEAD')}):
-        if not (submodule_git_dir / 'objects').is_dir():
-            continue  # e.g. refs/remotes/origin/HEAD, a ref, not a git directory.
-        if borrowed := _borrowing_git_file(submodule_git_dir):
-            raise ValueError(f'{submodule_git_dir} uses another git directory ({borrowed})')
-        _sanitize_config_file(submodule_git_dir / 'config', worktree_root=root)
+    _sanitize_submodule_git_dirs(modules, root)
+
+
+def _sanitize_submodule_git_dirs(modules: pathlib.Path, root: pathlib.Path):
+    """Check and sanitize every git directory in a `modules` directory, as git finds them: a directory holding HEAD
+    is a submodule's git directory (a submodule's name may contain `/`), and its nested submodules are in its own
+    `modules`.  Its refs and logs (which also hold files named HEAD) are not git directories."""
+    for child in sorted(modules.iterdir()):
+        if not child.is_dir():
+            continue
+        if not (child / 'HEAD').exists():
+            _sanitize_submodule_git_dirs(child, root)
+            continue
+        if borrowed := _borrowing_git_file(child):
+            raise ValueError(f'{child} uses another git directory ({borrowed})')
+        if not (child / 'objects').is_dir():
+            raise ValueError(f'{child} has no objects of its own')
+        _sanitize_config_file(child / 'config', worktree_root=root)
+        if (child / 'modules').is_dir():
+            _sanitize_submodule_git_dirs(child / 'modules', root)
 
 
 def find_readme(directory: pathlib.Path) -> Optional[str]:

modules/repos/test/test_ownership.py

@@ -308,3 +308,43 @@ def test_submodule_git_directories_pass(tmp_path):
     (git_dir / 'config').write_text('[core]\n\tbare = false\n')
     lib.sanitize_git_config(repo)
     assert 'proxy' not in (module / 'config').read_text()
+
+
+@pytest.mark.parametrize('name', ['library', 'refs', 'vendor/library'])
+def test_submodule_git_directory_without_objects_is_refused(tmp_path, name):
+    """A git directory which borrows another (commondir) has no objects of its own; it is refused, not skipped,
+    whatever the submodule is named."""
+    repo = tmp_path / 'repo'
+    git_dir = repo / '.git'
+    module = git_dir / 'modules' / name
+    module.mkdir(parents=True)
+    (git_dir / 'config').write_text('[core]\n\tbare = false\n')
+    (module / 'HEAD').write_text('ref: refs/heads/main\n')
+    (module / 'commondir').write_text(f'{tmp_path}\n')
+    with pytest.raises(ValueError):
+        lib.sanitize_git_config(repo)
+
+
+def test_nested_submodule_git_directories(tmp_path):
+    """A nested submodule's git directory is checked too; refs and logs named HEAD are not git directories."""
+    repo = tmp_path / 'repo'
+    git_dir = repo / '.git'
+    (git_dir).mkdir(parents=True)
+    (git_dir / 'config').write_text('[core]\n\tbare = false\n')
+    outer = git_dir / 'modules/library'
+    inner = outer / 'modules/nested'
+    for gitdir in (outer, inner):
+        (gitdir / 'objects').mkdir(parents=True)
+        (gitdir / 'HEAD').write_text('ref: refs/heads/main\n')
+        (gitdir / 'config').write_text('[core]\n\tbare = false\n[http]\n\tproxy = evil\n')
+        (gitdir / 'refs/remotes/origin').mkdir(parents=True)
+        (gitdir / 'refs/remotes/origin/HEAD').write_text('ref: refs/remotes/origin/main\n')
+        (gitdir / 'logs').mkdir()
+        (gitdir / 'logs/HEAD').write_text('0000 1111 t <t@t> 0 +0000\tclone\n')
+    lib.sanitize_git_config(repo)
+    assert 'proxy' not in (outer / 'config').read_text()
+    assert 'proxy' not in (inner / 'config').read_text()
+
+    (inner / 'commondir').write_text(f'{tmp_path}\n')
+    with pytest.raises(ValueError):
+        lib.sanitize_git_config(repo)