[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort)
• make test passes, and a test case covers the change
• 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 #1337. Each comment was re-checked with a probe on a8cc02594.

• ILNP nonce (:561): nonce=0 is one octet, and widths 1–16 all build and round-trip. Comment now says so.
• _httpv2_rebuild: since fix(httpv2): HTTP/2 frames lose octets on rebuild (#1242-#1248, #1251) #1341 make ORs flags in, so the generator now passes frame.flags.__value__ and reserved, like HTTP._make_data. Before, an undefined flag bit (DATA 0x02/0x80, PING 0x40, WINDOW_UPDATE 0x02) or the reserved bit was lost on rebuild.
• SKIP: _make_tcp_options keeps EOOL/NOP since fix(tcp): from_data reorders and re-pads TCP options #1163, so both now have cases (headers end 00000000 / 01010100, both OK). The rationale now covers only HOPOPT/IPv6-Opts PadN, which list input still collapses onto Pad1.
• IPv4 extract comment: narrowed to a fresh option list.

Captures: only options-tcp.pcap changes, 27 → 29 frames (2158 → 2306 bytes), the two new TCP padding cases. All others are md5-identical.

Tests: new tests/protocols/test_option_generator_1337_unit.py, 9 subtests fail without the fix. Every module importing the generator: 465 passed, 9 skipped, 5186 subtests, 0 failed.

[JarryShaw]

Verdict: GOOD TO GO at 2e84ed2a6 (Sonnet cross-review; the author ran on Opus).

• Tests bite, re-derived by me: the new test_option_generator_1337_unit passes 3/3 on this head. With only options.py put back to main, 9 subtests fail.
• HTTP/2 rebuild: the reviewer built 216 synthetic frames (types 0–11, flags 0/1/2/4/8/0x20/0x40/0x80/0xFF, reserved 0/1).
  • On main, undefined flag bits and the reserved bit are dropped; DATA 0x02 rebuilds as 0x00.
  • On this head there are 0 byte mismatches among the frames that parse.
  • The flags OR is idempotent for defined bits.
  • Unknown types (0x20, 0xFE) carry flags.__value__.
• TCP EOOL/NOP: now generated, with distinct headers (00000000 / 01010100) that round-trip OK. HOPOPT/IPv6-Opts PadN correctly stays skipped. Cases go 266 → 268, and no test pins the count.
• Captures: only options-tcp.pcap changes (27 → 29 frames, 2158 → 2306 bytes). The other four options-* md5s match main. examples/captures/* is gitignored, so nothing needs committing.
• Comments: the ILNP nonce (1–16 octets round-trip; 17 also works, so the wording is conservative but true) and the IPv4 extract comment were re-probed true.
• Scope: all three docs(generators): two false claims in options.py — ILNP nonce widths, and the SKIP padding rationale #1337 items are addressed, and only options.py plus the new test change.
• Neighbours: 16 generator/option/dispatch modules pass, including test_option_roundtrip_unit (364 subtests) and test_capture_tracking.
• test(generators): the 5-second per-case DEADLINE in options.py fails CI on a stalled runner #1428 (the 5 s DEADLINE): unaffected; the two new TCP cases are cheap.

[github-actions]

Coverage: 89.46% (unit tier, Python 3.14, 2e84ed2a6, 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	16615	220	4474	184	98.05%
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.