[JarryShaw]

• Searched for similar pull requests

• Followed the coding style — ran isort and pylint on the touched files (no new findings beyond existing too-many-arguments/redefined-builtin); mypy reports no errors in them. Did not run the make targets.

• make test passes, and a test case covers the change — ran only the modules listed below

• Added a changelog entry — N/A, added centrally

• fix — corrects a defect

---

Closes #1315
Closes #1327

• IPX: _make_data now packs the parsed Address back into 12 octets (network, node, socket); make gains total_length=None and writes a given value as is, so the parsed len survives.
• AH: reserved is a UInt16Field, carried as AH.reserved in the data model and make(reserved=0) (RFC 4302 §2.3).
• Breaking: new required data-class field Data_AH.reserved; new IPX.make(total_length=) keyword.
• test_ipx_unit.py / test_ah_unit.py: _make_data tests pinned the old pass-through / missing key; updated.

Probe (from_data(parsed.info).data): before, IPX 0000001e+26×00 → struct.error; AH 0604dead… → 06040000…. After: both byte-exact.

Tests: new test_ipx_ah_roundtrip_unit.py 5 passed (7 subtests); without the fix 7 failed; with only the address half, the two length subtests fail. test_ipx_unit 3 passed, test_ah_unit 4 passed, test_option_roundtrip_unit 6 passed, tests/project 389 passed 1 skipped; run_unittest_leg.py protocols/internet 508 tests, 0 failures.

[JarryShaw]

Cross-review verdict on b185ad43a: GOOD TO GO (Sonnet reviewed; the author ran on Opus).

• Repros: on main, IPX from_data raises struct.error for every packet tried, and AH 0604dead… rebuilds as 06040000…. Here both rebuild byte-exactly across all-zero, len=0, len=0xffff, len above the data and len below 30, and the AH dead/zero/ffff cases.
• No transposition: an asymmetric header (dst 0a0b0c0d/010203040506/4004, src deadbeef/112233445566/abcd) rebuilds exactly and parses into the right fields.
• RFC 4302 §2.3, fetched: 16 bits, MUST be zero on send, SHOULD be ignored on receipt, and included in the ICV. So UInt16Field at offset 2 is right.
• Blast radius: Data_AH( is only built in ah.py; a hand-built Schema_AH without reserved still works (defaults to 0). No caller passes total_length outside the PR. Both edited tests pinned wrong behaviour — the old IPX test fed the string 'dst-node' where 12 octets belong.
• Tests: the new module gives 7 failures on main. protocols/internet leg 508 tests, 0 failures. No fixture contains IPX or AH, so coverage rests on the unit tests and probes.

Nits, none blocking: the docstring says only "zero on send"; it should also say the field is ICV-covered, which is the real reason to carry it through. An out-of-range reserved is silently masked (-1 → ffff). And the SchemaWarning: packet length < 0 predates this PR but comes from IPX payload decoding in _parse_next_layer — bare AH headers do not emit it, contrary to the PR body.

[github-actions]

Coverage: 88.91% (unit tier, Python 3.14, b185ad43a, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18757	1033	2342	844	90.25%
pcapkit/corekit	1878	91	580	22	94.34%
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	15782	190	4038	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.