[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (did not run make pylint/mypy/isort)
• make test passes — ran only tests/corekit/test_sentinel_exports_unit.py (pytest) and util/run_unittest_leg.py corekit
• Changelog entry — N/A, added centrally

What is the purpose of your pull request?

• test — tests only

---

Closes #1332

test_sentinel_exports_unit.py bound the four sentinels at import, so once an earlier module re-imported pcapkit, assertIs compared a stale object with the canonical one. Fix: classes call reimport_once_per_class in setUp, and the sentinels and BytesField are read inside each test. No assertion is weakened; assertIs stays an identity check.

Before / after (util/run_unittest_leg.py corekit, file-size capped): 593 tests, 5 failures -> 622 tests, 0 failures, 0 errors. Module alone under pytest: 19 passed, 28 subtests passed. The original module fails 5/5 when run after test_fields_collections alone, so the new one is the one that holds.

Polluter. Any earlier module that calls reimport_once_per_class re-imports pcapkit and leaves it in place under plain unittest. The first in leg order is tests/corekit/test_fields_collections.py:25 (setUp), which reaches purge_modules(['pcapkit']) at tests/_support.py:970. Bisected one module at a time: all of test_fields_* (except test_fields_misc_schema_length_1148_unit), test_from_dict_roundtrip_unit, test_infoclass, test_io, test_module, test_multidict trigger it; the test_enum_*, test_protochain* modules do not. This is the helper working as designed (#981), so the stale binding was the sentinel module's fault.

[JarryShaw]

Cross-review verdict on d02df663e: NEEDS CHANGES — one line in the same file (Opus reviewed; author Sonnet).

test_star_import_hands_back_the_canonical_object cannot fail. Both sides of its assertIs come from the same shim module: from <shim> import * against <shim>.NULL. A star-import is getattr(shim, name), so the two can never differ. This predates the PR, since main compared against the same shim. The reviewer proved it with a mutation: giving corekit/module.py and corekit/enum.py their own NULL and NO_DEFAULT objects still passes this module 19/19, on both the PR and main. test_sentinels_housing_unit.py does catch it, with 10 failures.

Fix: take expected from importlib.import_module('pcapkit.corekit.sentinels') (the canonical module) rather than from the shim. Against the same mutation that reports module False, enum False, field True, i.e. it catches the broken shims and passes the clean one.

Held up:

• No weakening: every assertIs, assertIn and assertEqual is unchanged and none was weakened. The NO_VALUE/BytesField identity tests are not vacuous.
• Results: corekit leg 622 tests, 0 failures (5 on main); the module alone gives 19 passed; tests/project 389 passed.
• Pollution source confirmed: test_fields_collections followed by main's sentinel module reproduces exactly the 5 failures, and it is the first purger in leg order.

[JarryShaw]

Round-2 verdict on fc7584e1a: GOOD TO GO. I read the delta from d02df663e myself. It is one file, +5/−7. The identity test now runs self.assertIs(namespace[obj], getattr(canonical, obj)) with canonical = importlib.import_module(CANONICAL_MODULE), so the expected object comes from the defining module instead of the shim being star-imported.

The author reran the mutation proof. Giving corekit/module.py and corekit/enum.py their own NULL and NO_DEFAULT now fails exactly those two subtests, and the unchanged field shim still passes. With the mutation reverted, the module alone gives 19 passed and the corekit leg 622 tests, 0 failures.

[github-actions]

Coverage: 88.97% (unit tier, Python 3.14, fc7584e1a, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18777	1032	2348	843	90.27%
pcapkit/corekit	1949	84	606	20	94.91%
pcapkit/dumpkit	189	1	68	1	99.22%
pcapkit/foundation	2443	102	852	39	94.39%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	16013	206	4178	177	98.06%
pcapkit/toolkit	539	74	168	3	84.58%
pcapkit/utilities	429	4	122	4	98.55%
pcapkit/vendor	4409	2359	1006	158	42.84%

Per-file detail: the coverage-html artifact of this run.