Don't parse an HTTP error page's body as feed content - #597

Summary

Fixes #460.

>>> feedparser.parse("http://httpstat.us/500")
{'bozo': 1, 'entries': [], 'feed': {}, ...,
 'bozo_exception': SAXParseException('syntax error'), 'status': 500, ...}

http.get() returned the HTTP response body regardless of status code, so a server's 500 error page (HTML, plain text, whatever the error handler produces) got fed straight into the XML parser. That produces a bozo_exception that looks like a malformed feed, when the real problem — the HTTP failure — is already recorded cleanly in result["status"] a few lines above.

Fix: in feedparser/http.py, get() now returns b"" for any response with status_code >= 400, the same empty-body return it already uses when the request itself raises RequestException. status/headers/href are still set from the real response before that check, so callers can still see exactly what happened (status == 500), they just don't also get a misleading parse error bozo_exception.

Test plan

• Added tests/test_http.py with two tests using the responses mock already used elsewhere in the suite (tests/helpers.py):
  • test_error_status_body_is_not_parsed: a mocked 500 response with a plain-text body → status == 500, bozo is falsy, no bozo_exception key, entries == [].
  • test_success_status_body_is_still_parsed: a mocked 200 RSS response, to confirm normal parsing is unaffected.
• Full suite: pytest tests/ → 4299 passed, 8 skipped (was 4297 passed, 8 skipped on main before this change — the +2 are the new tests, zero regressions).
• mypy feedparser/http.py → no issues. flake8 on both changed files → clean.

Co-Authored-By: Claude Sonnet 5 [email protected]
🤖 Generated with Claude Code