US01-01: Inventory and Characterize Donor Code

work_item/.work-item.yml

@@ -10,6 +10,7 @@ workflow:
   require_ci: false
   required_tests:
     - work_item/scripts/python -m unittest discover -s work_item/tests -v
+    - work_item/scripts/python -m unittest discover -s tests/characterization -v
 
 safety:
   max_file_bytes: 5000000

work_item/src/work_item/core.py

@@ -317,7 +317,13 @@ class GitRepo:
         return branch
 
     def changed_paths(self) -> list[Path]:
-        output = self.git("status", "--porcelain=v1", "--untracked-files=all", "-z")
+        # Porcelain's leading status column is significant (an unstaged modification
+        # starts with a space). ``git()`` strips surrounding whitespace for ordinary
+        # scalar commands, so read this machine format directly and preserve it.
+        output = self.runner.run(
+            ("git", "status", "--porcelain=v1", "--untracked-files=all", "-z"),
+            cwd=self.root,
+        ).stdout
         if not output:
             return []
         entries = output.split("\0")

work_item/tests/test_core.py

@@ -354,6 +354,12 @@ class GitSafetyTests(RepositoryFixture):
         run("git", "mv", "README.md", "RENAMED.md", cwd=self.repo)
         self.assertEqual(git.changed_paths(), [Path("RENAMED.md")])
 
+    def test_changed_paths_preserves_modified_and_untracked_filenames(self) -> None:
+        git = GitRepo(self.repo, self.config, Runner())
+        (self.repo / "README.md").write_text("changed\n", encoding="utf-8")
+        (self.repo / "Another file.md").write_text("new\n", encoding="utf-8")
+        self.assertEqual(set(git.changed_paths()), {Path("README.md"), Path("Another file.md")})
+
     def test_diff_check_rejects_whitespace_errors(self) -> None:
         git = GitRepo(self.repo, self.config, Runner())
         (self.repo / "README.md").write_text("seed\ntrailing whitespace   \n", encoding="utf-8")