fix: respect LC_ALL and preserve Unicode stop words in Bash

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Author: mnriem

docs/reference/core.md

@@ -80,6 +80,10 @@ The Python helper also accepts `--short-name`; the PowerShell helper uses
 `-ShortName`. A supplied short name is cleaned by the same rules, so it must
 contain at least one letter or digit. For non-ASCII names, the Bash helper needs
 an installed UTF-8 locale and a Python 3 interpreter for Unicode classification.
+If `LC_ALL` is non-empty, Bash uses that locale rather than selecting another:
+Unicode names fail with an error if the selected locale is not usable for UTF-8
+names. With `LC_ALL` unset or empty, Bash selects an installed UTF-8 locale even
+when `LANG` or `LC_CTYPE` names a non-UTF-8 locale.
 ASCII capitals are lowercased; non-ASCII letter casing is preserved across the
 script variants.
 

scripts/bash/create-new-feature.sh

@@ -152,7 +152,11 @@ spec_prefix_exists() {
 #     '+', leaving repeated separators uncollapsed on macOS.
 #   * printf instead of echo, so a name of "-n"/"-e"/"-E" is text, not options.
 UNICODE_LOCALE=""
-for candidate in C.UTF-8 C.utf8 en_US.UTF-8 en_US.utf8 "${LC_ALL:-${LC_CTYPE:-${LANG:-}}}"; do
+locale_candidates=(C.UTF-8 C.utf8 en_US.UTF-8 en_US.utf8 "${LC_CTYPE:-${LANG:-}}")
+if [ -n "${LC_ALL:-}" ]; then
+    locale_candidates=("$LC_ALL")
+fi
+for candidate in "${locale_candidates[@]}"; do
     if [ -n "$candidate" ] && [ "$(printf 'é。' | LC_ALL="$candidate" sed 's/[^[:alnum:]]/-/g' 2>/dev/null)" = 'é-' ]; then
         UNICODE_LOCALE="$candidate"
         break
@@ -162,7 +166,11 @@ done
 if [ -z "$UNICODE_LOCALE" ]; then
     UNICODE_LOCALE=C
     if printf '%s' "${SHORT_NAME:-$FEATURE_DESCRIPTION}" | LC_ALL=C grep -q '[^ -~]'; then
-        echo "Error: A UTF-8 locale is required to create a Unicode feature name" >&2
+        if [ -n "${LC_ALL:-}" ]; then
+            echo "Error: A UTF-8 locale is required to create a Unicode feature name; LC_ALL=$LC_ALL is not usable" >&2
+        else
+            echo "Error: A UTF-8 locale is required to create a Unicode feature name" >&2
+        fi
         exit 1
     fi
 fi
@@ -279,7 +287,7 @@ generate_branch_name() {
         [ -z "$word" ] && continue
 
         # Retain non-ASCII words even when shorter than three characters.
-        if ! echo "$word" | grep -qiE "$stop_words"; then
+        if ! printf '%s\n' "$word" | LC_ALL=C grep -qE "$stop_words"; then
             if [ ${#word} -ge 3 ] || printf '%s' "$word" | LC_ALL=C grep -q '[^ -~]'; then
                 meaningful_words+=("$word")
             # Keep short words that appear as an uppercase acronym in the original.

tests/test_create_new_feature_python_parity.py

@@ -217,6 +217,32 @@ def test_bash_reports_missing_utf8_locale_for_unicode_only(
         assert not (repo / "specs").exists()
 
 
+@requires_bash
+@pytest.mark.parametrize("locale_name", ["C", "POSIX"])
+def test_bash_respects_explicit_non_utf8_lc_all(repo: Path, locale_name: str) -> None:
+    env = clean_env()
+    env["LC_ALL"] = locale_name
+    env["LANG"] = "C.UTF-8"
+
+    ascii_result = run(
+        bash_cmd(repo, SCRIPT, "--json", "--dry-run", "Add user authentication"),
+        repo,
+        env,
+    )
+    assert ascii_result.returncode == 0, ascii_result.stderr
+    assert json_stdout(ascii_result)["BRANCH_NAME"] == "001-user-authentication"
+
+    for args in (("添加用户",), ("--short-name", "用户", "Add users")):
+        unicode_result = run(
+            bash_cmd(repo, SCRIPT, "--json", "--dry-run", *args), repo, env
+        )
+        assert unicode_result.returncode == 1
+        assert unicode_result.stdout == ""
+        assert "A UTF-8 locale is required" in unicode_result.stderr
+        assert "LC_ALL" in unicode_result.stderr
+        assert not (repo / "specs").exists()
+
+
 @requires_bash
 def test_bash_requires_python_only_for_unicode_names(
     repo: Path, tmp_path: Path
@@ -381,6 +407,54 @@ def test_python_branch_name_generation_matches_bash(
     assert json_stdout(py) == json_stdout(bash)
 
 
+@requires_bash
+@pytest.mark.parametrize(
+    ("description", "expected"),
+    [("ſet account", "001-ſet-account"), ("Set account", "001-account")],
+)
+def test_bash_stop_words_match_only_ascii_words(
+    repo: Path, description: str, expected: str
+) -> None:
+    bash = run(bash_cmd(repo, SCRIPT, "--json", "--dry-run", description), repo)
+    py = run(py_cmd(repo, SCRIPT, "--json", "--dry-run", description), repo)
+
+    assert bash.returncode == py.returncode == 0
+    assert json_stdout(bash) == json_stdout(py)
+    assert json_stdout(bash)["BRANCH_NAME"] == expected
+
+
+@requires_bash
+def test_bash_stop_words_ignore_locale_case_folding(repo: Path, tmp_path: Path) -> None:
+    real_grep = shutil.which("grep")
+    assert real_grep is not None
+    shim_dir = tmp_path / "bin"
+    shim_dir.mkdir()
+    grep_shim = shim_dir / "grep"
+    grep_cmd = shlex.quote(_bash_posix_path(Path(real_grep)))
+    grep_shim.write_text(
+        "#!/bin/sh\n"
+        'if [ "$1" = "-qiE" ]; then\n'
+        "  IFS= read -r word\n"
+        '  [ "$word" = "ſet" ] && exit 0\n'
+        f'  printf "%s\\n" "$word" | {grep_cmd} "$@"\n'
+        "  exit $?\n"
+        "fi\n"
+        f'exec {grep_cmd} "$@"\n',
+        encoding="utf-8",
+        newline="\n",
+    )
+    grep_shim.chmod(0o755)
+    env = clean_env()
+    env["PATH"] = f"{shim_dir}{os.pathsep}{env['PATH']}"
+
+    bash = run(bash_cmd(repo, SCRIPT, "--json", "--dry-run", "ſet account"), repo, env)
+    py = run(py_cmd(repo, SCRIPT, "--json", "--dry-run", "ſet account"), repo, env)
+
+    assert bash.returncode == py.returncode == 0
+    assert json_stdout(bash) == json_stdout(py)
+    assert json_stdout(bash)["BRANCH_NAME"] == "001-ſet-account"
+
+
 @requires_bash
 @pytest.mark.skipif(not HAS_POWERSHELL, reason="no PowerShell available")
 @pytest.mark.parametrize(
@@ -1489,7 +1563,8 @@ def test_no_ascii_word_description_matches_across_twins(
     ps_repo = _setup_repo(tmp_path, "s")
 
     env = clean_env()
-    env["LC_ALL"] = env["LANG"] = "C"
+    env.pop("LC_ALL", None)
+    env["LANG"] = "C"
     bash = run(
         bash_cmd(bash_repo, SCRIPT, "--json", "--dry-run", description), bash_repo, env
     )