[JarryShaw]

• Searched for similar pull requests
• Followed the coding style — pylint (Makefile flags), mypy and isort run on the touched files: no new messages; make targets not run
• make test passes, and a test case covers the change — only the modules listed below
• 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

Closes #1418
Refs #1288

All 24 sites name a reserved sub-field, read into the data model (flags_reserved on the ten messages, since MH.reserved is the header octet; reserved on the options, Multi-Prefix extension and per-session bit-rate attribute) and written by make(), default 0. Follows #1345.

Breaking: each of the 24 data classes gains a required field.

Before/after: each frame sets every reserved bit of its site. On main, from_data rebuilt all 24 with those bits zeroed (e.g. Heartbeat fffc → 0000, FBU cfff → c000). Now all 24 are byte-exact.

Tests: new tests/protocols/internet/test_mh_flags_reserved_unit.py, table-driven over all 24 sites: parse→from_data, make→parse, default 0, over-wide value rejected. Without the fix: 120 subtests fail. With it: 6 passed / 120 subtests. test_mh_unit.py stubs updated for the new field (52 passed). test_mh_roundtrip_unit 6 passed, test_option_roundtrip_unit 6 passed, leg protocols/internet 572 OK, tests/project 399 passed.

Rebase note: #1416 is not merged yet. Its KNOWN_GAPS lists these 24 MH entries; delete them when rebasing over #1416.

[JarryShaw]

Verdict: GOOD TO GO at 009e5f4e0 (Sonnet cross-review; the author ran on Opus).

• Repro: on main, all 24 sites zero their reserved bits. Two examples: Heartbeat fffc→0000, and FBU cfff→c000. All 24 are byte-exact here, including the widest sites (Traffic Offload Selector and Multi-Prefix, (1,31)) and the odd offsets (IPv4 Address Ack (6,2), Redirect (2,14), Delegated MNP).
• The new module: with the 3 source files reverted, it fails all 120 site subtests.
• RFCs: all 24 sites were checked against the fetched RFCs (5846, 5568, 5949, 5847, 5555, 5844, 5535, 5026, 6058, 6463, 6909, 7028, 7148, 7222, 8278, 6705, 7077, 7109, 7161). The new reserved bits are exactly each RFC's reserved bits, and the neighbouring named flags sit at the right offsets. No defined flag is mislabelled.
• Naming: flags_reserved on the 10 messages and reserved elsewhere, matching fix(mh): keep zero-length PadN, reserved fields and BU/BA flag bits on rebuild #1345.
• Blast radius: no constructor or hand-built dict outside this PR breaks. The other MH, option, schema and make modules pass. The test_mh_unit.py edits only add reserved.
• Fixtures: 22 captures, 1588 frames. They differ only by the new keys. All 99 MH layers in options-internet.pcap rebuild byte-exact.
• Leg: protocols/internet gives 572 tests, 0 failures.
• test(schema): every BitField namespace covers all of its bits #1416: with its test applied on top, the only failure is the 24 MH KNOWN_GAPS entries going stale. Whichever PR merges second deletes them.

[github-actions]

Coverage: 89.43% (unit tier, Python 3.14, 009e5f4e0, 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	2038	89	634	25	94.76%
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	16529	219	4420	183	98.04%
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.