diff --git a/scripts/devstream/license_detect.py b/scripts/devstream/license_detect.py index 7275b69..bc6c3ab 100644 --- a/scripts/devstream/license_detect.py +++ b/scripts/devstream/license_detect.py @@ -276,7 +276,8 @@ def file_bytes(gh, r, data: dict, what: str) -> bytes: if data.get("encoding") != "base64": raise Undecided(0, f"{what} of {r} has no readable content") try: - return base64.b64decode("".join((data.get("content") or "").split()), validate=False) + # validate=True: otherwise stray characters are dropped and garbage decodes as a short file. + return base64.b64decode("".join((data.get("content") or "").split()), validate=True) except (binascii.Error, ValueError) as e: raise Undecided(0, f"{what} of {r} could not be decoded: {e}") from e diff --git a/scripts/devstream/quality.py b/scripts/devstream/quality.py index d8dbf25..4423f72 100644 --- a/scripts/devstream/quality.py +++ b/scripts/devstream/quality.py @@ -62,7 +62,11 @@ class RepoResult: def _readme_text_length(text: str) -> int: text = text[:README_SCAN] # also bounds the regex work below text = re.sub(r"```.*?(```|\Z)", " ", text, flags=re.S) - text = re.sub(r"^\s{0,3}#.*$", " ", text, flags=re.M) + # ATX headings need a space (or nothing) after the #s, so a #hashtag line is prose. + text = re.sub(r"^ {0,3}#{1,6}(?:[ \t].*)?$", " ", text, flags=re.M) + # Setext headings: a line underlined with === or ---. + # The lookahead scans each line once; [^\n]*\S[^\n]* would backtrack on a long blank line. + text = re.sub(r"^(?=[^\n]*\S)[^\n]*\n {0,3}(?:=+|-+)[ \t]*$", " ", text, flags=re.M) text = re.sub(r"<[^>]{0,2000}>", " ", text) text = re.sub(r"!\[[^\]]{0,500}\]\([^)]{0,2000}\)", " ", text) text = re.sub(r"\[([^\]]{0,500})\]\([^)]{0,2000}\)", r"\1", text) @@ -102,6 +106,25 @@ def _fork_diverged(repo, gh) -> bool: return cmp_ is not None and cmp_.get("ahead_by", 0) > 0 +def _community_files(gh, r) -> dict[str, bool]: + """Which community files exist. The profile 404s for forks, so then look where GitHub would.""" + profile = gh.get(f"/repos/{r}/community/profile") + if profile is not None: + return {k: v is not None for k, v in (profile.get("files") or {}).items()} + names = set() + for d in ("", "/.github", "/docs"): + listing = gh.get(f"/repos/{r}/contents{d}") + if isinstance(listing, list): + names |= {(f.get("name") or "").lower() for f in listing} + + def has(stem): + return any(n == stem or n.startswith(stem + ".") for n in names) + return {k: has(k) for k in ("contributing", "code_of_conduct", "issue_template", "pull_request_template")} + + +EXTRAS = 7 + + def _extras_missing(repo, gh, r) -> list[str]: """Empty list when 2+ extras exist. Cheap checks first; stops at 2.""" found = 0 @@ -117,11 +140,11 @@ def tally(ok, label): if tally(len(repo.get("topics") or []) >= 1, "a topic or two"): return [] - files = (gh.get(f"/repos/{r}/community/profile") or {}).get("files") or {} + files = _community_files(gh, r) checks = [ - (files.get("contributing") is not None, "a CONTRIBUTING guide"), - (files.get("code_of_conduct") is not None, "a code of conduct"), - (files.get("issue_template") is not None or files.get("pull_request_template") is not None, + (files.get("contributing", False), "a CONTRIBUTING guide"), + (files.get("code_of_conduct", False), "a code of conduct"), + (files.get("issue_template", False) or files.get("pull_request_template", False), "an issue or pull request template"), ] for ok, label in checks: @@ -214,5 +237,7 @@ def check_repo(full: str, gh, now: datetime, mode: Literal["submit", "rot"], failures.append(MSG_DEPRECATED.format(deprecated)) missing = _extras_missing(repo, gh, r) if missing: - failures.append("Add at least two of these to the repo: " + ", ".join(missing) + ".") + # Only a full miss reaches here, so every extra was checked and the rest were found. + need = "one more" if EXTRAS - len(missing) == 1 else "at least two" + failures.append(f"Add {need} of these to the repo: " + ", ".join(missing) + ".") return RepoResult(True, r, failures, "ok", meta, repo.get("id")) diff --git a/scripts/tests/test_quality.py b/scripts/tests/test_quality.py index c3b5a91..0f925fb 100644 --- a/scripts/tests/test_quality.py +++ b/scripts/tests/test_quality.py @@ -174,6 +174,16 @@ def test_readme_noise_not_counted(self): self.assertEqual(len(run(gh(readme=noise)).failures), 1) self.assertEqual(run(gh(readme=noise + "x" * 300)).failures, []) + def test_readme_hashtags_are_prose_and_setext_headings_are_not(self): + tags = "\n".join(["#rustlang #cli #tools #terminal #productivity"] * 7) + self.assertGreater(len(tags), 300) + self.assertEqual(run(gh(readme=tags)).failures, []) + setext = "\n".join([f"A heading that is long enough number {i}\n=====" for i in range(8)] + + [f"Another long subheading number {i}\n---" for i in range(4)]) + self.assertGreater(len(setext), 300) + self.assertEqual(len(run(gh(readme=setext)).failures), 1) + self.assertEqual(len(run(gh(readme="#\n## Usage\n### " + "x" * 400)).failures), 1) + def test_readme_exactly_300(self): self.assertEqual(run(gh(readme="a" * 300)).failures, []) self.assertEqual(len(run(gh(readme="a" * 299)).failures), 1) @@ -189,8 +199,11 @@ def test_extras_counts_two_needed(self): r = run(none) self.assertEqual(len(r.failures), 1) self.assertIn("CONTRIBUTING", r.failures[0]) - # one extra is not enough - self.assertEqual(len(run(gh(topics=(), coc=False)).failures), 1) + self.assertIn("Add at least two of these", r.failures[0]) + # one extra is not enough, and the message says one more is + r = run(gh(topics=(), coc=False)) + self.assertEqual(len(r.failures), 1) + self.assertTrue(r.failures[0].startswith("Add one more of these to the repo: a topic or two, a code of conduct")) # two is self.assertEqual(run(gh(topics=(), coc=False, contributing=True, extra={f"GET {R}/releases?per_page=1": [{}]})).failures, []) @@ -208,6 +221,19 @@ def test_extras_each_kind(self): r = run(gh(**base, extra=extra)) self.assertEqual(len(r.failures), 1, name) + def test_extras_without_a_community_profile(self): + # Forks have no community profile (404), so the files are looked up directly. + base = dict(topics=(), extra={f"GET {R}/community/profile": None}) + self.assertEqual(len(run(gh(**base)).failures), 1) + listings = {f"GET {R}/contents": [{"name": "README.md"}, {"name": "CONTRIBUTING.md"}], + f"GET {R}/contents/.github": [{"name": "ISSUE_TEMPLATE", "type": "dir"}]} + g = gh(topics=(), extra={f"GET {R}/community/profile": None, **listings}) + self.assertEqual(run(g).failures, []) + listings = {f"GET {R}/contents/docs": [{"name": "code_of_conduct.md"}, {"name": "contributing-notes.md"}]} + r = run(gh(topics=(), extra={f"GET {R}/community/profile": None, **listings})) + self.assertEqual(len(r.failures), 1) + self.assertIn("one more", r.failures[0]) + def test_extras_short_circuit(self): g = gh() # topics + contributing already = 2 run(g) @@ -349,10 +375,12 @@ def test_readme_encoding_none_without_sha_is_undecided(self): run(gh(extra={f"GET {R}/readme": bad})) def test_bad_base64_is_undecided(self): - bad = {"encoding": "base64", "content": "abcde"} - for mode in ("submit", "rot"): - with self.assertRaises(Undecided): - run(gh(extra={f"GET {R}/readme": bad}), mode) + # "!!!!" used to be dropped silently and read as an empty, short README. + for content in ("abcde", "!!!!", b64(README)["content"] + "!!!!"): + bad = {"encoding": "base64", "content": content} + for mode in ("submit", "rot"): + with self.assertRaises(Undecided, msg=content): + run(gh(extra={f"GET {R}/readme": bad}), mode) def test_null_pushed_at(self): r = run(gh({"pushed_at": None}))