[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort) — ran pylint/mypy/isort on internet.py only; no new messages on the added lines
• make test passes, and a test case covers the change — protocols/internet leg, every test_ipv6*/test_mh* module, test_option_roundtrip_unit, tests/project
• Added a 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
• feat · [ ] perf · [ ] refactor · [ ] test · [ ] docs · [ ] ci · [ ] release · [ ] chore

---

Description of your pull request and other information

Closes #1446

An extension header parsed with extension=True is handed the rest of the datagram, but parses only itself: info has no payload and .payload raises UnsupportedCall. Internet.__post_init__ now cuts .data to the header's length in that mode, so type(ex).from_data(ex.info).data == ex.data. This keeps IPv6._make_data as it is, since it already joins from_data(info).data per header. Carrying the remainder in info would have emitted those octets twice.

Covers HOPOPT, IPv6-Route, IPv6-Frag, IPv6-Opts, MH, AH, HIP and IPv6_Ext. With HOPOPT + UDP, data/rebuilt goes from 28/8 to 8/8. On ipv6.pcap frame 16 (Fragment), 442/8 becomes 8/8. IPv6 parent rebuilds stay exact, 247/247 across examples/captures.

Breaking (!): data, bytes() and len() of an extension-mode header now cover the header only. A standalone parse is unchanged.

[github-actions]

Coverage: 89.51% (unit tier, Python 3.14, 0b9fc265c, 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	16805	219	4522	183	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.

[JarryShaw]

Verdict: GOOD TO GO at 0b9fc265c (Sonnet cross-review; the author ran on Opus).

• No other reader breaks. ipv6.py:448 is the only extension=True site, and toolkit/{pcap,pcapng}.py read only .info and .length.
• Reassembly is unchanged. IPv6 fragment reassembly of ipv6.pcap frames 13–16 is byte-identical to main (4778 bytes, same sha1).
• Only the extension header's .data changes. Across 1469 frames, per-frame info and the parent IPv6.data match main.
• Every class goes through the hook. HOPOPT, Route, Frag, Opts, MH, AH, HIP, ESP and Shim6 all call super().__post_init__. On 13 synthetic chains, len(data) == length holds everywhere.
• Rebuilds are exact. Every extension header in the 221 IPv6 frames rebuilds byte-exact from its own info.
• test: capture-driven parse→rebuild and dumper round trips (#1202) #1445 interaction. Merged with test: capture-driven parse→rebuild and dumper round trips (#1202) #1445, the only failures are the stale-entry flags on its fix(ipv6): an extension header parsed with extension=True cannot rebuild from its own info #1446 LAYER_GAPS rows, as expected. Whichever lands second deletes them.
• Tests, which I re-ran: the new module gives 4 passed with 14 subtests. With internet.py from main it gives 8 failed, 3 passed.
• Neighbours. The internet leg runs 601 OK, and reassembly, traceflow and toolkit are clean.

Nits, no change needed:

• The test's class list omits Shim6 (IPv6_Ext, 140), although the probe shows it fixed.
• A header truncated inside itself still rebuilds padded. That is unchanged from main, so I filed it separately as fix(ipv6): a truncated IPv6 extension header rebuilds padded to its declared length #1451.