[JarryShaw]

• Searched for similar pull requests
• Followed the coding style — isort clean; pylint/mypy (Makefile flags) on the 9 sources: no new mypy errors, pylint adds 3 R0801 duplicate-code (the shared guard). make targets not run.
• make test passes, and a test case covers the change — the legs below only
• Changelog entry — N/A, added centrally

What is the purpose of your pull request?

• fix — corrects a defect

---

Description of your pull request and other information

Closes #1362. Refs #1363.

The 14 parser registrars (IPv4, HOPOPT, IPv6-Opts, IPv6-Route, HIP, MH ×3, TCP ×2, SCTP ×3, HTTP/2) now warn only when the incumbent differs from the replacement; an equal entry is a silent no-op and keeps the stored object. PCAP-NG's four (register_block/_option/_record/_secrets) are left for the PCAP-NG work in #1398. The traceflow.py dumper guard already compares identity after #1376 and is unchanged.

Note on #1363: these registrars refuse nothing, so the shipped entry is already accepted back; the ruling here means restoring it after an override warns once (a real displacement) and re-restoring is then silent. No snapshot is needed, unlike the class registrars in #1376.

Probe: re-registering every shipped entry warned 14/14 sites (271 entries) on origin/main, 0 now.

Tests: new tests/protocols/test_parser_registrar_identity_unit.py, 5 tests × 14 sites. With origin/main sources: 42 subtests fail; the foreign-warns and unregistered-silent guards pass both ways. Eleven existing "warns on overwrite" tests re-registered the shipped value and now overwrite with a different one. run_unittest_leg.py protocols/internet: 552, 0 failures. protocols --exclude protocols/internet: 1009, 0 failures. pytest tests/foundation/traceflow/: 25 passed. pytest tests/project: 399 passed, 1 skipped.

[JarryShaw]

Cross-review verdict on 5cbed8ebf: GOOD TO GO (Sonnet reviewed; author Opus).

• == is safe.
  • Stored values are only str and tuples of callables. A swapped (g, f) against a stored (f, g) still warns, and distinct lambdas never compare equal.
  • A falsy non-None incumbent such as '' still warns when replaced.
  • No stored type makes == raise.
• .get changes nothing. All 14 containers are defaultdicts, and neither .get nor the old in inserts a missing key.
• No test was weakened. All 11 modified tests still assert one warning and the new stored value on a real displacement. The IPv4 and HIP tests also gained a check that a repeat stays silent.
• No site was missed. Of 18 parser registrars in pcapkit/protocols, 14 are guarded here. The other 4 are PCAP-NG's, now tracked in fix(pcapng): the four PCAP-NG parser registrars warn when handed their own entry #1401.
• Revert proof: 42 failed, as claimed. The modified test modules pass, and protocols/internet gives 552/0. I re-ran the new module myself and it passes (5/0).

Nit: sibling class-storing registrars compare with is not, so == here is a deliberate widening for strings and does no harm.

[github-actions]

Coverage: 89.44% (unit tier, Python 3.14, 5cbed8ebf, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18791	968	2352	869	90.90%
pcapkit/corekit	2018	88	628	24	94.78%
pcapkit/dumpkit	234	2	84	2	98.74%
pcapkit/foundation	2637	98	928	48	94.78%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	16369	205	4352	167	98.17%
pcapkit/toolkit	539	74	168	3	84.58%
pcapkit/utilities	429	4	122	4	98.55%
pcapkit/vendor	4409	2342	1006	157	43.25%

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