[JarryShaw]

Please follow the guide below

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort) — run on the three changed modules with the Makefile flags; no new message kinds
• make test passes, and a test case covers the change — the modules listed below only, not the full suite
• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible — 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 #1212
Closes #1213
Closes #1214
Closes #1215
Closes #1216
Closes #1217
Closes #1220

Every MPTCP subtype now rebuilds byte-exactly. Reserved bits are mapped and kept. MP_FAIL reads its 12 reserved bits, and unknown subtypes read length - 3. DSS widths follow a/m, and its checksum follows the length. ADD_ADDR takes its family, port and HMAC from the length (RFC 8684), and RFC 6824 options still round-trip. MP_CAPABLE keeps C and D–G.

Breaking: these are new required data-model fields:

• reserved on MPTCPJoinSYN/SYNACK/ACK, MPTCPDSS, MPTCPRemoveAddress, MPTCPPriority, MPTCPFallback and MPTCPFastclose
• ack_wide and dsn_wide on MPTCPDSS
• reserved, echo and hmac on MPTCPAddAddress
• deny_join and reserved on MPTCPCapableFlag

MPTCPAddAddress.version now comes from the address family. The schema test dicts gained reserved/echo keys, and the ADD_ADDR one lost version. mptcp_add_address_selector reads length. make() output is unchanged by default. The ADD_ADDR maker now rejects an hmac that is not 8 octets, or one given with echo=True.

Probe (24 option repros, TCP.from_data(TCP(raw,len(raw)).info).data): 19 DIFF/ERR before (the 3 controls already OK) → 24 OK after. Fixtures: 1259 TCP frames round-trip, 0 bad.

Tests:

• New test_mptcp_roundtrip_unit.py: 12 passed, 36 subtests. With the three source files reverted to main, 38 failed.
• The test_tcp_udp_unit, test_tcp_mptcp_addaddr_unit and test_tcp_mptcp_error_message_unit tests now build the new fields. One rejection case moved from version 5 to length 9, and the OptNo count went 33→35. The tests/corekit/test_fields_misc.py ADD_ADDR rows now pass lengths 8 and 20.
• Other MPTCP/TCP modules, test_option_roundtrip_unit and tests/project (389 passed) all pass. The protocols --exclude protocols/internet unittest leg ran 862 tests OK. test_fields_misc 10 passed, test_fields_ipaddress 23 passed.

[github-actions]

Coverage: 88.92% (unit tier, Python 3.14, 7eb1b2110, Unit Tests run failure)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18757	1034	2342	845	90.24%
pcapkit/corekit	1878	89	580	22	94.43%
pcapkit/dumpkit	142	0	42	0	100.00%
pcapkit/foundation	2442	102	852	39	94.38%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	15806	190	4048	165	98.18%
pcapkit/toolkit	487	65	144	1	85.42%
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.

[JarryShaw]

Cross-review verdict on 634ac75cf: NEEDS CHANGES (Sonnet; author Opus).

• Must fix: tests/corekit/test_fields_misc.py::SwitchFieldBoolDispatchTests::test_address_typed_switch_branches_still_take_a_real_address now fails: 5 subtests with KeyError: 'length'. Its hand-built table at :57-60 still passes {'test': {'version': 4/6}}, but mptcp_add_address_selector now reads pkt['length']. The module passes on main (10 passed). The leg excluding protocols/internet does not cover tests/corekit, which is why the PR's run missed it. The docstring at tests/corekit/test_fields_ipaddress.py:522 is stale too.
• Confirmed: every issue repro rebuilds byte-exactly, and a 114-case MPTCP probe gives 114 OK. Every layout matches the RFC 8684 text. ADD_ADDR lengths 8/10/16/18 (IPv4) and 20/22/28/30 (IPv6) are unambiguous, and RFC 6824 options still round-trip. 1256 TCP fixture frames rebuild identically.
• Nits:
  • The default ADD_ADDR make() emits E=0 without an HMAC (the RFC 6824 layout).
  • echo=True is accepted together with an HMAC.
  • A short hmac is zero-padded silently.

[JarryShaw]

Cross-review verdict on 7eb1b2110: GOOD TO GO (Sonnet, round 2; author Opus).

• Round-1 blocker fixed: tests/corekit/test_fields_misc.py gives 10 passed, and test_fields_ipaddress.py gives 23 passed. No other MPTCP schema in tests/ uses version. All 13 MPTCP-related modules pass.
• Repros: all 13 issue repros rebuild byte-exactly, and fail on main. I re-ran the MP_FAIL repro myself.
• Fixtures: every TCP fixture layer rebuilds exactly.
• ADD_ADDR checks: these match RFC 8684 §3.4.1. An HMAC must be 8 octets, an echo carries no HMAC, and parsed options rebuild as read.
• Error-message pin: the 34→35 change matches exactly the one new invalid HMAC message.
• Legs:
  • protocols --exclude protocols/internet: 862 tests, 0 failures.
  • The corekit leg's 5 failures in test_sentinel_exports_unit.py also occur on clean main. They are pre-existing and will be filed separately.
• Nit: with no HMAC and echo=False, ADD_ADDR still builds the legacy RFC 6824 layout.