[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort) — run on the 9 changed source files with the Makefile flags; no new pylint/mypy findings against base, isort clean
• make test passes, and a test case covers the change — per-module only, listed below
• Added a changelog entry — N/A, added centrally after the wave

What is the purpose of your pull request?

• fix — corrects a defect

---

Description

Closes #1230
Closes #1231
Closes #1236
Refs #1235 (IPv6-Frag Res bits and RPL routing reserved bits only; HOPOPT/IPv6-Opts sites are another worker's)

• fix(ipv6-ext): IPv6_Ext rebuild drops the header body, and an overrun writes Hdr Ext Len 0xff #1230: IPv6_Ext info now records len, declared_next and data (body), and _make_data re-emits them. Before → after: 3b00010203040506 → 3b00 / exact; 3b00 (overrun) → 00ff / 3b00; IPv6 next=140 48 → 42 / 48 octets.
• fix(ipv6): next-header codes 253 and 254 are parsed and rebuilt twice #1231: a layer with no next is no longer recorded in __exthdr__. IPv6 next=253/254: 48 → 56 / 48 octets; chain IPv6:Raw:Raw → IPv6:Raw.
• fix(ipv6-route): IPv6_Route.make(data=<bytes>) raises for every routing type #1236: bytes data is unpacked into the type's schema (or ProtocolError). Types 0/2/3/4/5: AttributeError → builds 3b02<t>00+20×00.
• fix(ipv6): reserved bits not named in a BitField are written back as zero (Frag, RPL routing, MPL, IP_DFF, Quick-Start) #1235 (part): Frag 3b00000712345678 → …0112… / exact; RPL route reserved 0xabcde → 0 / kept.

Contract changes: new required data fields IPv6_Ext.{declared_next,len,data}, IPv6_Frag.reserved, RPL.reserved; schema dicts IPv6_Frag.flags / RPL.pad now need a reserved key; new make() keywords IPv6_Ext(data=), IPv6_Frag(reserved=), RPL reserved=. For 253/254, info.protocol is the code (was None), and the chain's last layer is named after it.

Tests: new tests/protocols/internet/test_ipv6_core_roundtrip_unit.py, 7 passed; all 7 fail with the source fix reverted. Updated tests that pinned the old behaviour: test_ipv6_ext_unit.py (253 walk) and test_ipv6_extension_unit.py (schema dicts and data models gain reserved; bytes route data). Also passing: test_ipv6_extension_unit 61, test_ipv6_ext_unit 30, test_option_roundtrip_unit, the ipv6/ah/mh/esp/reassembly/toolkit modules, and tests/project (all green once isort was fixed). util/run_unittest_leg.py protocols/internet: 510 tests, 0 failures. All 247 IPv6 frames in examples/captures/* rebuild byte-exactly.

[JarryShaw]

Cross-review verdict on 1c4621b50: GOOD TO GO (Sonnet reviewed; the author ran on Opus).

• Repros: the fix(ipv6-ext): IPv6_Ext rebuild drops the header body, and an overrun writes Hdr Ext Len 0xff #1230, fix(ipv6): next-header codes 253 and 254 are parsed and rebuilt twice #1231 and fix(ipv6-route): IPv6_Route.make(data=<bytes>) raises for every routing type #1236 repros fail on main and pass here. I re-ran the fix(ipv6-ext): IPv6_Ext rebuild drops the header body, and an overrun writes Hdr Ext Len 0xff #1230 body and overrun cases myself.
• Fixtures: I measured 242 IPv6 frames (the PR body says 247). They parse to the same info and protochain as on main, apart from the new keys, and rebuild byte-exactly.
• Fuzzing: 6000 random IPv6 chains showed no regressions, and 1244 chains that failed on main now rebuild.
• Specs: checked against RFC 8200 §4.5 (Fragment header) and RFC 6554 §3 (CmprI/CmprE/Pad/Reserved).
• 253/254: RFC 3692/4727 define no format for these codes, so decoding them as the upper layer and preserving the octets is defensible.
• Tests: the protocols/internet leg passes (510 tests, 0 failures), along with about 45 other modules. The new tests fail without the fix.
• Not addressed here: the Fragment header's 8-bit Reserved octet and the 4-octet reserved word in routing types 0 and 2 are still zeroed on rebuild. Both are PaddingField sites already recorded on fix(corekit): non-zero padding octets are zeroed on rebuild #1223.

[github-actions]

Coverage: 88.90% (unit tier, Python 3.14, 1c4621b50, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18757	1034	2342	845	90.24%
pcapkit/corekit	1878	91	580	22	94.34%
pcapkit/dumpkit	142	0	42	0	100.00%
pcapkit/foundation	2442	102	852	39	94.38%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	15785	192	4038	166	98.15%
pcapkit/toolkit	487	65	144	1	85.42%
pcapkit/utilities	429	4	122	4	98.55%
pcapkit/vendor	4409	2359	1006	158	42.84%

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

[JarryShaw]

The red mark is the #1306 mirror fault, not this PR. PyPCAP/PyPCAPFile parity Python 3.11 (job 112948578234) failed at step 4, Install libpcap headers, a C toolchain, and tshark; every later step is skipped, so nothing in the diff was ever exercised.

From the log: Need to get 27.9 MB of archives, then azure.archive.ubuntu.com returning Ign: at 18:33:16 and 18:33:46 — apt's default 30s Acquire::http::Timeout cycling — and the 90s apt-get install bound firing at 18:34:16, on the second of two attempts. The 9011 kB apt-get update beforehand managed 530 kB/s, so the mirror was reachable and then stalled on the archive fetch. #1309 raises those bounds to 60s/180s for exactly this.

Re-running the job.