[JarryShaw]

• Searched for similar pull requests
• Followed the coding style — mypy on the 3 MH modules: no new errors vs main (311 both); pylint adds only unused-argument on the new TYPE_CHECKING stub parameters (as every existing stub does); isort not run
• make test passes, and a test case covers the change — ordering leg protocols/internet 588 OK; test_bitfield_coverage_unit 3 passed; tests/project 419 passed, 1 skipped
• Changelog entry — N/A — centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

What is the purpose of your pull request?

• fix — corrects a defect

---

Description of your pull request and other information

Refs #1437 (part A: MH). Part B (IPv4, L2TP, PCAP-NG) is a separate PR.

Every MH reserved PaddingField that #1436 did not thread, all 38, zero-filled on rebuild. Measured on main: Home Test Init …0101000000aabb… became …010100000000…, and GRE Key 2102aabb became 21020000. The 38 are 8 message bodies, 22 options, 3 Flow Identification sub-options, 1 ANI sub-option and 4 QoS attributes. Each reader now keeps the octets (msg_reserved on a message, apart from the header's reserved; reserved elsewhere). Each maker takes them as a keyword and writes them back through a tolerant getattr. A fresh build still writes zeros.

Breaking: those 38 data models gain a required bytes field. test_mh_unit.py had five hand-built fixtures, which now carry it.

New tests/protocols/internet/test_mh_reserved_padding_unit.py covers parse→rebuild byte-exact and fresh-build zeros for all 38 sites, and checks that the site list matches the schema. On main it fails 40 (38 sub-tests + 2). After make_samples.py, options-internet.pcap info gains only zero-valued msg_reserved/reserved (8 + 22 frames), with no other change.

[JarryShaw]

Verdict: GOOD TO GO at c29c50a5a (Sonnet cross-review; the author ran on Opus).

• Completeness. Introspecting the schema finds 40 PaddingField sites in schema/internet/mh.py. Taking away fix(corekit)!: non-zero padding octets are zeroed on rebuild (#1223) #1436's PadOption and PadFlowIdentificationSuboption leaves 38, which match the PR's list by name.
• Byte-exactness. All 38 cases were rewritten with different octets, and every one rebuilds exactly through both info and info.to_dict(). Fresh builds write zeros, and wrong-width keywords follow fix(corekit)!: non-zero padding octets are zeroed on rebuild (#1223) #1436's PaddingField.pack.
• The cases are real tests. Each asserts its raw hex differs from the zero form, so none can pass vacuously.
• ! / breaking is justified. The 38 data models gain a required field. Across the captures only options-internet.pcap changes: 30 of 148 frames gain a zero-valued key, and nothing else differs.
• Tests, which I re-ran: the new module gives 5 passed with 76 subtests. With the three mh.py files from main it gives 40 failed, 3 passed.
• Merge. It merges into 7c6203892 with no conflicts. The internet leg is 588 OK, test_mh_unit 52 passed, and tests/project 419 passed.

Worth a decision while the change is breaking anyway:

• msg_reserved is already an int on BindingRefreshRequestMessage (pcapkit/protocols/data/internet/mh.py:171). After this PR it is bytes on the seven other messages, so one name carries two types.
• Renaming the new field now is cheaper than doing it later. Otherwise the PR body should say why the name is shared.

[github-actions]

Coverage: 89.49% (unit tier, Python 3.14, c29c50a5a, 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	2043	89	636	25	94.77%
pcapkit/dumpkit	234	2	84	2	98.74%
pcapkit/foundation	2648	98	930	48	94.80%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	16719	218	4492	182	98.08%
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.