Merge pull request #2654 from ShaoLOWu/fix/bootutils-files-global-clobber

fix(core): stop generate_files from clobbering the required_paths global

src/langbot/pkg/core/bootutils/files.py

@@ -18,13 +18,11 @@ required_paths = [
 
 
 async def generate_files() -> list[str]:
-    global required_files, required_paths
-
     from ...utils import paths as path_utils
 
-    for required_paths in required_paths:
-        if not os.path.exists(required_paths):
-            os.mkdir(required_paths)
+    for path in required_paths:
+        if not os.path.exists(path):
+            os.mkdir(path)
 
     generated_files = []
     for file in required_files:

tests/unit_tests/core/test_bootutils_files.py

@@ -0,0 +1,48 @@
+"""Regression tests for bootutils.files.generate_files.
+
+Historically the directory-creation loop iterated as
+``for required_paths in required_paths:`` while the function declared
+``global required_paths``, which rebound the module-level list to the last
+iterated string (``'data/labels'``) after the first call. A second call then
+iterated over that string character by character and attempted to create
+single-character directories. These tests pin the module-level state and the
+repeat-call behaviour.
+"""
+
+from __future__ import annotations
+
+import asyncio
+from importlib import import_module
+
+
+def get_files_module():
+    """Lazy import to avoid circular import issues."""
+    return import_module('langbot.pkg.core.bootutils.files')
+
+
+def test_required_paths_stays_a_list_after_generate_files(tmp_path, monkeypatch):
+    files = get_files_module()
+    monkeypatch.chdir(tmp_path)
+    original = list(files.required_paths)
+
+    asyncio.run(files.generate_files())
+
+    assert isinstance(files.required_paths, list)
+    assert files.required_paths == original
+
+
+def test_generate_files_is_idempotent(tmp_path, monkeypatch):
+    files = get_files_module()
+    monkeypatch.chdir(tmp_path)
+
+    first = asyncio.run(files.generate_files())
+    second = asyncio.run(files.generate_files())
+
+    for path in files.required_paths:
+        assert (tmp_path / path).is_dir()
+    # The config template is copied on the first run only.
+    assert first == ['data/config.yaml']
+    assert second == []
+    assert (tmp_path / 'data' / 'config.yaml').is_file()
+    # No junk single-character directories left behind by a second run.
+    assert sorted(p.name for p in tmp_path.iterdir()) == ['data', 'temp']