scripts/family.py and tests: family.py init --dry-run's help says its folder answer is os.access's and that it leaves nothing that stays, and a test pins the seed NOTE's GitHub clause (#923, #921)

docs/cli.md

@@ -956,8 +956,14 @@ options:
                         --work-dir, the holder in it, --local-remote-dir) and
                         whether <org>/<Family> already exists (`gh repo view`
                         and its commit listing; with --local-remote-dir, the
-                        bare repository), refusing what it refuses; make no
-                        folder, remote or repository
+                        bare repository), refusing what it refuses. Whether
+                        each folder can be written in is asked of `os.access`,
+                        which reads the mode bits and on Windows says yes for
+                        every folder, so a quota, a Windows ACL or a read-only
+                        drive on Windows can pass here and be refused by the
+                        real run. It leaves nothing that stays: it makes and
+                        removes temporary folders, creates no remote or
+                        repository and pushes nothing
 ```
 
 #### `scripts/family.py add`

scripts/family.py

@@ -90,7 +90,7 @@
 question the real run asks before it creates the holder (#537): `gh repo
 view` and the commit listing, or the bare repository with
 `--local-remote-dir`. It refuses `family-remote-exists` as the real run does,
-and makes no folder, remote or repository. Both runs ask it after whether
+and creates no remote or repository. Both runs ask it after whether
 each folder `init` makes can be made and the holder written (#609), and
 before the real run makes the family folder, so a refusal leaves none: a file
 where a folder must go, or a folder this user cannot write in, is
@@ -99,6 +99,23 @@
 `gh repo view` failure (a bad login, a rate limit, no network) is refused as
 `remote-unreadable`.
 
+`init --dry-run` LEAVES NOTHING THAT STAYS, which is not the same as making
+no folder (#921). It makes and removes temporary folders, as the real run
+does: `openreposhape-branch-*`, where git is asked whether the tracking
+branch is a name it will create, and `openreposhape-ident-*`, one for the
+author when GIT_AUTHOR_NAME, _EMAIL or _DATE is set and not empty, and one
+for the committer when a GIT_COMMITTER_ one is. Each is empty, and a removal
+that fails is not reported. It creates no remote or repository and pushes
+nothing: the only `gh` calls it makes read (`gh repo view`, the commit
+listing).
+
+WHETHER EACH FOLDER CAN BE WRITTEN IN IS THE MODE BITS' ANSWER (#923). The
+family folder or `--work-dir`, the holder in it and `--local-remote-dir` are
+asked of `os.access`, which reads the mode bits and on Windows says yes for
+every folder, and a dry run makes none of them to ask. So a quota, a Windows
+ACL or a read-only drive on Windows can pass the dry run and be refused by the
+real run.
+
 `init` ASKS GITHUB'S LIMITS FIRST, WHERE GITHUB IS (#863), as the scaffold
 and `adopt-project.py` ask theirs (#838), before `_creator_name` and before
 any `gh` call, so `--dry-run` refuses the same. The holder's topic,
@@ -10452,8 +10469,15 @@ def build_parser() -> argparse.ArgumentParser:
                            "--local-remote-dir) and whether <org>/<Family> "
                            "already exists (`gh repo view` and its commit "
                            "listing; with --local-remote-dir, the bare "
-                           "repository), refusing what it refuses; make no "
-                           "folder, remote or repository")
+                           "repository), refusing what it refuses. Whether "
+                           "each folder can be written in is asked of "
+                           "`os.access`, which reads the mode bits and on "
+                           "Windows says yes for every folder, so a quota, "
+                           "a Windows ACL or a read-only drive on Windows "
+                           "can pass here and be refused by the real run. "
+                           "It leaves nothing that stays: it makes and "
+                           "removes temporary folders, creates no remote or "
+                           "repository and pushes nothing")
     init.set_defaults(func=cmd_init)
 
     add = subparsers.add_parser(

scripts/shape_materialize.py

@@ -1618,7 +1618,8 @@ def folder_in_the_way(path: Path) -> tuple[Path, str] | None:
     `path` and the folders above it that exists is the one that answers. A
     file there, or above it, is named, and so is a folder this user cannot
     write in (the system's own answer, `os.access`, where it has one; it
-    says yes for root, and for any folder on Windows, as `mkdir` would). A
+    says yes for root, and for any folder on Windows, where the real `mkdir`
+    can still be refused: an ACL, a quota). A
     `path` nothing stands at is asked of the folder above it.
 
     `folder_cannot_be_made` says it as one sentence, naming the folder as

tests/test_github_limits_init_check_org.py

@@ -409,6 +409,30 @@ def test_init_help_says_each_bound_where_the_value_is_typed():
         assert sentence in entry, (flag, entry)
 
 
+def test_init_dry_run_help_says_the_mode_bits_and_nothing_that_stays():
+    """#923 U063 and #921 U088, the family's half, as the scaffold's
+    `--dry-run` says it: the write question is asked of `os.access`, which on
+    Windows says yes for every folder, so a quota, a Windows ACL or a
+    read-only drive there can pass it, and the run leaves "nothing that
+    stays" where it once said "make no folder": it makes and removes
+    `openreposhape-branch-*` (and `openreposhape-ident-*` where an identity
+    variable is set), creates no remote or repository and pushes nothing. The
+    wording is `docs/cli.md`'s too, which `test_cli_reference.py` holds."""
+    result = run_script(FAMILY, "init", "--help")
+    assert result.returncode == 0, result.stderr
+    said = help_entry(result.stdout, "--dry-run")
+    assert ("Whether each folder can be written in is asked of `os.access`, "
+            "which reads the mode bits and on Windows says yes for every "
+            "folder, so a quota, a Windows ACL or a read-only drive on "
+            "Windows can pass here and be refused by the real run. It leaves "
+            "nothing that stays: it makes and removes temporary folders, "
+            "creates no remote or repository and pushes nothing") in said, said
+    assert "make no folder" not in said
+    assert "its mode bits'" not in said, (
+        "only the write question is `os.access`'s: a file in the way and a "
+        "folder's contents are not the mode bits'")
+
+
 # --- family.py init end to end: GitHub's refusal, by its documentation ------
 
 #: A `gh` with no network that answers as GitHub's documentation says: a

tests/test_scaffold_reuse_seeded_legs.py

@@ -378,6 +378,49 @@ def test_the_seed_note_says_github_only_on_github(scaffold_module, local,
         "T", branches, "main", local) == said
 
 
+@pytest.mark.parametrize("local,ending", [
+    pytest.param(False, "so that branch remains, and on GitHub it may be "
+                 "the leg's default branch.", id="github-leg"),
+    pytest.param(True, "so that branch remains.", id="bare-repository-leg"),
+])
+def test_the_note_a_re_run_prints_says_github_only_for_a_github_leg(
+        scaffold_module, monkeypatch, capsys, local, ending):
+    """RV1052d's NIT-A: `_reusable_remotes` hands `_seed_branch_note` the
+    `local` it was called with. Every run in this file that finds a seed only
+    on `seed/scaffold` is a `--local-remote-dir` run, and the unit test above
+    calls `_seed_branch_note` directly, so a call that always passed `True`
+    for `local` (RV1052d's R2c) printed no GitHub clause for a GitHub leg and
+    no test failed. Here `remote_is_empty` says the leg exists with commits
+    and `_held_seed` says its one commit is the seed, found on
+    `seed/scaffold` alone, and the NOTE that `_reusable_remotes` itself
+    prints is what is read: GitHub's ends with the default-branch clause, and
+    the bare repository's, which has no GitHub, does not. In-process, so it
+    runs on every platform."""
+    seed = "0123456789abcdef0123456789abcdef01234567"
+    monkeypatch.setattr(scaffold_module, "remote_is_empty",
+                        lambda repository: False)
+    monkeypatch.setattr(scaffold_module, "local_remote_is_empty",
+                        lambda target: False)
+    monkeypatch.setattr(scaffold_module, "_held_seed",
+                        lambda role, url, values, tracking, scratch: (
+                            seed, None, ("seed/scaffold",)))
+    reused = scaffold_module._reusable_remotes(
+        {"spec": "Atlas-spec"}, {"spec": "/remotes/Atlas-spec.git"},
+        {"spec": "o/Atlas-spec"}, local=local, reuse_flag=True, values={},
+        tracking="main")
+    assert reused == {"spec": seed}
+    notes = [line for line in capsys.readouterr().out.splitlines()
+             if line.startswith("NOTE the seed in")]
+    target = str(Path("/remotes/Atlas-spec.git")) if local else "o/Atlas-spec"
+    assert notes == [
+        f"NOTE the seed in {target} was found only on its branch "
+        "`seed/scaffold`: this run creates `main` at it and does not touch "
+        f"`seed/scaffold`, {ending}"], notes
+    assert notes[0].endswith(", and on GitHub it may be the leg's default "
+                             "branch.") is (not local)
+    assert ("GitHub" in notes[0]) is (not local)
+
+
 def test_dry_run_help_says_the_mode_bits_and_nothing_that_stays(tmp_path):
     """#923 U063 and #921 U088, the scaffold's half: `--dry-run`'s help says
     the write question is asked of `os.access`, which on Windows says yes for