[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
• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible — N/A, centralised in chore(docs): centralize 1.6.0 changelog updates in one place, not per-feature-PR #2449

What is the purpose of your pull request?

• fix — corrects a defect

---

Description of your pull request and other information

Per RFC 8200 §4, a header whose dedicated parser raises (HOPOPT, Routing, Opts, MH, HIP, Frag, AH) is now Raw via beholder and ends the walk; IPv6._import_next_layer's IPv6_Ext fallback and __generic_ext_codes__ are removed. IPv6_Ext stays the parser for Shim6, so only Shim6 matches every member name. Docs (ipv6.py, ipv6_ext.py, ipv6-ext-entry-paths, the subclassing convention) and the tests pinning the fallback are updated; new tests cover malformed and truncated headers, Shim6, and byte-exact rebuilds. Stacked on #2464.

Closes #2477

[JarryShaw]

NEEDS CHANGES (Sonnet, c40071afb). Behaviour is right. Across 12,317 hand-built packets, every difference from the base is a former fallback now giving Raw. Samples match frame for frame, Shim6 is unchanged, and docs warnings equal the base's.

• Regression: a malformed header is now logged twice. The walk breaks on hasattr(info, 'next') but keeps proto, so _decode_next_layer re-parses the same octets. Reuse the walk's Raw instead.
• Breaking-change markers: add a versionremoved:: 1.6.0 for __generic_ext_codes__ and ! in the subject. Document that the IPv6_Ext.error field and read(error=) are no longer filled, and point readers to payload.info.error.
• Stale prose: mh.py still says IPv6 substitutes IPv6_Ext for a failed MH (4 docstrings). Also stale: test_ipv6_ext_hint_unit.py, _has_fallback in test_short_read_roundtrip_unit.py, and "last two subtests" in test_mh_unit.py.
• Missing test: none pins AH → Raw.

[JarryShaw]

Verified (ad7cd10a3): double-log fixed (2 records, down from 4; the remaining WARNING is beholder's own, flagged for decorators.py); breaking markers added and the subject carries !; stale MH prose rewritten; AH pinned. Sonnet review dispatched.

[JarryShaw]

Base retargeted to main now that #2464 has merged. Still review: running; no other change.

[JarryShaw]

NEEDS CHANGES (Sonnet, ad7cd10a3). The double-log fix holds: one malformed HOPOPT now logs once plus beholder's WARNING, and the corpus is identical to base apart from one case.

• Unexpected change: 45 of 26,205 cases now give NoPayload where base gives an empty Raw. Repro: a Routing header whose declared length overruns the Payload Length, with nothing after it. Cause: the reused next_ also captures a NoPayload built with length=None. Fix: reuse only when type(next_) is Raw. Or keep NoPayload on purpose and pin it.
• Stale prose: test_mh_unit.py ~1466 still says the MH is "replaced by IPv6_Ext".

[JarryShaw]

Fixed (9dd653e74): overrun-Routing-with-nothing-after is Raw again (pinned by a new test), and the MH prose is corrected. Sonnet review dispatched.

[github-actions]

Coverage: 94.50% (unit tier, Python 3.15, 69d6728d7, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19677	133	2628	120	98.85%
pcapkit/corekit	2605	74	818	37	96.52%
pcapkit/dumpkit	281	3	100	3	98.43%
pcapkit/foundation	3070	24	1072	26	98.79%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	18050	194	4786	180	98.35%
pcapkit/toolkit	924	36	334	10	95.55%
pcapkit/utilities	502	7	144	5	98.14%
pcapkit/vendor	4449	1854	1020	167	55.11%

Per-file detail: the coverage-html artifact of this run.

[JarryShaw]

NEEDS CHANGES (Sonnet, 9dd653e74). Behaviour verified: the overrun-Routing case is Raw, and 173k replayed cases differ only as the intended fallback-to-Raw change.

• ipv6_ext.py read alias bullet (~L410-413) and the HOPOPT examples at ~L306-309 and ~L659-662 say IPv6_Ext supplies a failed header; only Shim6 does now.
• unidiomatic-typecheck pragma: add a reason clause (exact-type is deliberate because Raw subclasses exist).
• E305: one blank line missing before if __name__ in test_ipv6_exthdr_cut_rebuild_unit.py.

[JarryShaw]

NEEDS CHANGES (Sonnet, 9dd5d5fcf). Prose fixes, pragma reason, E305, and the Raw-ends-walk behaviour are verified; 3 tests pass.

• ipv6.py _decode_next_layer: the # block above if not hasattr(info, 'next'): is 20 lines; cap 8. Keep three points: no next ends the walk (RFC 8200 section 4, A malformed known IPv6 extension header should parse as Raw, not IPv6_Ext #2477), it is not recorded as a header, and a Raw is reused as the upper layer so its error logs once.

[JarryShaw]

GOOD TO GO (Sonnet, 69d6728d7). The round is comment-only (AST-identical); the new block is 4 lines; 3 tests pass.

• Non-blocking: the body still says "Stacked on fix(protocols): compose IPv6_Ext.id() from its direct concrete members #2464", which merged as 4ca3f30d7.