[JarryShaw]

• Searched for similar pull requests

• Followed the coding style — ran pylint/mypy/isort on the two source files; isort clean, and pylint/mypy report only findings on unchanged lines (not make targets)

• make test passes, and a test case covers the change — the modules listed below only, not the full suite

• Added a changelog entry — N/A, added centrally after the wave

• fix — corrects a defect

---

Closes #1204
Closes #1205
Closes #1206

• fix(link): InARP and DRARP fall back to the Raw schema and cannot parse any input #1204: InARP/DRARP now pass schema=Schema_ARP, data=Data_ARP.
• fix(link): ARP with an Ethernet hardware type and hlen other than 6 cannot rebuild #1205: parsing any hlen is correct, and from_data rebuilds it through the parsed hlen.
• fix(link): ARP picks the protocol-address IP version from its length instead of ptype #1206: a protocol address is decoded as IP only when ptype and its length agree (IPv4/4, IPv6/16). Otherwise it is kept as hex. make() packs it raw when plen disagrees with ptype.
• Contract change:
  • make() still requires exactly 6 octets when the hardware type is Ethernet (htype 1) and no explicit hlen was given.
  • When the caller passes an explicit hlen, or the htype is not Ethernet, it accepts any octet count.
  • A parsed spa/tpa under a mismatched ptype/plen is now a hex str, where before it was an IPv6Address or a bare ValueError.

Probe (parse, then from_data). Before: InARP/DRARP AttributeError: 'Raw' object has no attribute 'htype'; hlen 8/0 ProtocolError: Invalid MAC address; v4+plen16 FieldValueError ... 6 != 4; v4+plen5 bare ValueError. After: all nine cases (ARP/RARP/InARP/DRARP) rebuild byte-exactly.

Tests:

• New test_arp_roundtrip_unit.py gives 8 passed, 85 subtests passed.
• With the fix reverted it failed in every test method.
• With only the Ethernet 6-octet default reverted, it gives 12 failed.
• test_link_unit, test_arp_proto_addr_length_unit, test_arp_pcapng_hwaddr_octets_unit, test_rarp_unit, test_link_runtime and test_option_roundtrip_unit all pass.
• tests/project gave 389 passed, 1 skipped before this amend.
• run_unittest_leg.py protocols --exclude protocols/internet gives 858 test(s), 0 failure(s), 0 error(s).

[JarryShaw]

Cross-review verdict on b289532fb: NEEDS CHANGES (Sonnet; author Opus).

• Must fix: the loosened PAT_MAC_ADDR (arp.py:409) lets ARP(htype=1, sha='00:11:22:33:44', …) build a frame with hlen=5. sha='' builds hlen=0. On main both raised. Parsing any hlen is right per RFC 826. Building an Ethernet-hardware ARP with no hlen given should still require 6 octets, or the PR body must state and test the change. Ethernet itself is unaffected (ethernet.py:56 keeps the strict copy).
• Confirmed: the fix(link): InARP and DRARP fall back to the Raw schema and cannot parse any input #1204, fix(link): ARP with an Ethernet hardware type and hlen other than 6 cannot rebuild #1205 and fix(link): ARP picks the protocol-address IP version from its length instead of ptype #1206 repros rebuild byte-exactly for ARP, RARP, InARP and DRARP. An explicit hlen/plen is respected. The new tests fail 41/45 on main.
• Nit: the Raises: text should say the empty string is now accepted.

[github-actions]

Coverage: 88.91% (unit tier, Python 3.14, cce483801, 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	15779	190	4038	165	98.18%
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]

Cross-review verdict on cce48380: GOOD TO GO (Sonnet, round 2; author Opus).

• Fixed: both round-1 cases raise ProtocolError again, for ARP, InARP, RARP and DRARP. These are htype=1 with a 5-octet MAC (I re-ran this one) and with an empty MAC.
• Explicit hlen and non-Ethernet htype: explicit hlen 5, 0 and 8 build, and Fibre_Channel accepts any address length.
• Round trips: 28/28 round-trip byte-exactly: seven repros across the four classes, covering fix(link): InARP and DRARP fall back to the Raw schema and cannot parse any input #1204, fix(link): ARP with an Ethernet hardware type and hlen other than 6 cannot rebuild #1205 hlen 8 and 0, and fix(link): ARP picks the protocol-address IP version from its length instead of ptype #1206 v4/16, v4/5 and v6/4. I re-ran the hlen-8 case myself.
• Default make() unchanged: output matches main byte for byte, including IPv6-over-Ethernet ARP and Fibre_Channel.
• Tests: the new tests fail without the amend (12 failed). run_unittest_leg.py protocols --exclude protocols/internet: 858 tests, 0 failures.