Buckets:
AnonymousSpottedToad/candidate_bucket / codebase /python /als-apg_osprey /tests /utils /test_logging_guard.py
| """Tests for the ``restore_root_logging`` guard in tests/conftest.py. | |
| ``configure_logging()`` mutates process-global logging state. The guard undoes | |
| that between tests so ``caplog`` behaves identically no matter what ran before. | |
| Several tests here come in ordered pairs: the first deliberately leaks, the | |
| second — which pytest runs immediately after it, in file order, on the same | |
| xdist worker — asserts the leak is gone. | |
| Expected values come from ``_PRISTINE_LOGGING``, the guard's own baseline, | |
| rather than from a level captured in a sibling test. That is deliberate: an | |
| `import litellm` anywhere in the process raises the httpx logger to WARNING as | |
| an import side effect, and six ``osprey`` modules import litellm — three inside | |
| functions, and three (the LiteLLM provider adapter, the in-context | |
| channel-finder query tool, the benchmark harness) at module level, where | |
| collection alone is enough to trigger it. So a value captured in one test is not | |
| a reliable expectation for the next one — the number it produces (WARNING) is | |
| also indistinguishable from what ``configure_logging()`` produces. | |
| ``_PRISTINE_LOGGING`` is captured at conftest import, before any of that, and is | |
| exactly the contract the guard promises. | |
| """ | |
| from __future__ import annotations | |
| import logging | |
| import pytest | |
| from rich.logging import RichHandler | |
| from osprey.utils.logger import QUIET_THIRD_PARTY_LOGGERS, configure_logging | |
| # Private by name, but this is the module that tests it. | |
| from tests.conftest import _PRISTINE_LOGGING | |
| _PROBE = "osprey.tests.logging_guard_probe" | |
| def _rich_handler_count() -> int: | |
| return sum(isinstance(h, RichHandler) for h in logging.getLogger().handlers) | |
| def _assert_third_party_pristine() -> None: | |
| for name in QUIET_THIRD_PARTY_LOGGERS: | |
| assert logging.getLogger(name).level == _PRISTINE_LOGGING["third_party"][name] | |
| # =================================================================== | |
| # configure_logging() leaves no trace for the next test | |
| # =================================================================== | |
| def test_configure_logging_mutates_global_state(): | |
| """Leak on purpose. The pre-conditions also pin that nothing leaked into us.""" | |
| root = logging.getLogger() | |
| assert _rich_handler_count() == 0, "a previous test leaked a RichHandler onto the root logger" | |
| _assert_third_party_pristine() | |
| # Move a third-party level somewhere configure_logging() will not land on, | |
| # so the follow-up test can prove the restore actually ran. | |
| logging.getLogger("httpx").setLevel(logging.DEBUG) | |
| configure_logging(logging.DEBUG) | |
| assert _rich_handler_count() == 1 | |
| assert root.level == logging.DEBUG | |
| for name in QUIET_THIRD_PARTY_LOGGERS: | |
| assert logging.getLogger(name).level == logging.WARNING | |
| def test_configure_logging_left_no_trace(): | |
| """Runs right after the leaking test: root logging is back to the baseline.""" | |
| assert _rich_handler_count() == 0 | |
| assert logging.getLogger().level == _PRISTINE_LOGGING["root_level"] | |
| _assert_third_party_pristine() | |
| # Non-vacuous: the previous test explicitly set this one to DEBUG. | |
| assert logging.getLogger("httpx").level != logging.DEBUG | |
| # =================================================================== | |
| # caplog keeps working | |
| # =================================================================== | |
| def test_caplog_captures_after_configure_logging(caplog): | |
| """The guard must not strip caplog's handler, and configure_logging must not either.""" | |
| configure_logging() | |
| with caplog.at_level(logging.INFO, logger=_PROBE): | |
| logging.getLogger(_PROBE).info("during") | |
| assert "during" in caplog.text | |
| def test_caplog_captures_in_the_test_after_a_configuring_one(caplog): | |
| """Runs right after a test that configured logging; capture is unaffected.""" | |
| with caplog.at_level(logging.INFO, logger=_PROBE): | |
| logging.getLogger(_PROBE).info("after") | |
| assert "after" in caplog.text | |
| assert _rich_handler_count() == 0 | |
| # =================================================================== | |
| # Configuration done by higher-scoped fixtures | |
| # =================================================================== | |
| def configuring_module_fixture(): | |
| """A module-scoped fixture that configures logging, as a CLI-invoking one would. | |
| Mirrors ``tests/cli/test_dockerfile_template.py``, whose module-scoped | |
| fixtures run ``osprey build`` through ``CliRunner``. pytest sets higher-scoped | |
| fixtures up first, so this lands on the root logger *before* the | |
| function-scoped guard runs. | |
| """ | |
| configure_logging() | |
| def test_higher_scoped_configuration_is_undone_for_the_test(configuring_module_fixture): | |
| """Both the handler and the third-party levels the fixture set are already undone.""" | |
| assert _rich_handler_count() == 0 | |
| _assert_third_party_pristine() | |
| def test_higher_scoped_configuration_stays_undone_afterwards(): | |
| """And stays undone for later tests that never request the fixture.""" | |
| assert _rich_handler_count() == 0 | |
| _assert_third_party_pristine() | |
| def test_third_party_capture_is_not_silenced_by_a_previous_test(caplog): | |
| """A prior configure_logging() must not decide whether httpx records are captured. | |
| Two tests above this one raised httpx to WARNING through configure_logging(). | |
| Without the guard's third-party level restore, whether this capture works at | |
| all would depend on which files ran first on this worker. | |
| """ | |
| assert logging.getLogger("httpx").level == _PRISTINE_LOGGING["third_party"]["httpx"] | |
| with caplog.at_level(logging.INFO, logger="httpx"): | |
| logging.getLogger("httpx").info("third-party record") | |
| assert "third-party record" in caplog.text | |
Xet Storage Details
- Size:
- 5.75 kB
- Xet hash:
- 3288b85f30403fc98ffe26bffc22db356aa302826b8fcca8e4f4b67fd6bfa740
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.