[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 chore(docs): centralise changelog entries for unreleased versions #2483

What is the purpose of your pull request?

• feat — adds a feature

---

Description of your pull request and other information

DCCP (RFC 4340, plus DCCP-Listen from RFC 5596) gets a full dissector in pcapkit.protocols.transport.dccp, replacing the NotImplemented stub. IP protocol 33 dispatches to it.

• Header and fields: the generic header with 48- or 24-bit sequence numbers (the X bit chooses), each packet type's extra fields, and every RFC 4340 option. Unknown options and types are kept raw, and reserved bits are kept as captured.
• Malformed input: a fuzz of every option type against every length octet leaves no input that both parses and fails to rebuild byte-exact. Every input either rebuilds byte-exact or is refused, which under IP means Raw.
• Checksum: it follows pep.rst's checksum design. Verification is opt-in with a recorded ChecksumStatus, and the coverage is CsCov's (§9.2). make() writes CsCov 0 when the coverage reaches the end of the packet, as Linux's dccp_csum_outgoing did.
• Registries: DCCP.register_option plus the register_dccp and register_dccp_option wrappers. There are six IANA registries, as crawlers plus generated const modules that regenerate byte-identically.

Settled on the worker's recommendations, open to the owner:

1. A CsCov that runs past the packet is INVALID and raises a ChecksumWarning.
2. DCCP has no NOT_COMPUTED exemption, because it has no "no checksum" value.
3. Option lengths behave as in TCP: an impossible length refuses the layer.

Follow-ups:

• register_apptype fan-out to DCCP's port map.
• Retargeting the checksum imports once refactor(corekit): move the Internet checksum into corekit.checksum.internet #2586 merges.

Closes #2489

[JarryShaw]

GOOD TO GO at 0bfbe212f. Two Sonnet reviews covered it, the dissector first and then its fix round, and I checked the rebase myself.

• Malformed options: fuzzing every option type against every length octet gives 0 parse→rebuild mismatches. 448 inputs are refused at parse and 56,928 rebuild byte-exact.
• Checksum: CsCov coverage follows RFC 4340 §9.2. make()'s CsCov 0 matches Linux's dccp_csum_outgoing.
• Lint and tests: mypy is clean on the DCCP modules, and the protocols/transport leg ran 566 tests, OK.
• Rebase: onto f4aa57742, patch-identical outside two registry test tables, where refactor(application)!: dispatch HTTP to HTTPv1/HTTPv2 under the RFC names (#2524) #2544's HTTPv2 rename and the DCCP rows both stay.

[JarryShaw]

Question for the owner: make() takes type as PacketType | int, resolved with PacketType.get. It does not use the type/type_default/type_namespace/type_reversed group that ICMP uses, which pep.rst's "Deterministic Construction Arguments" calls fragile.

1. Keep it as PacketType | int (recommended). New dissectors should not take on the group pep.rst flags.
2. Use the ICMP group, for consistency with existing dissectors.
   The PR implements 1 and is otherwise complete.

[JarryShaw]

Would prefer 1. But if that is so, we need to rework all other make paths to follow this convention.

[JarryShaw]

Owner chose 1, paraphrased: make(type=...) keeps taking PacketType | int, and every other make() should follow the same convention. So this PR stands as written. The rollout across the other make() methods (dropping the X_default/X_namespace/X_reversed groups) is filed as #2594, with one sub-question for the owner: remove the shipped keywords in 1.6.0, or deprecate them first.

[JarryShaw]

The red "Python pypy3.11" leg is a lost runner, not a test failure. The log ends at 9 m 23 s with "The runner has received a shutdown signal", in the middle of the tests/protocols shard. That is #2600: since the checksum wave, the PyPy protocols shard has run close to the runner's memory, and this PR adds DCCP's tests to the same shard. The shard split in #2600 is the fix. Re-run this leg once #2600 has merged. The review verdict stands.

[github-actions]

Coverage: 94.06% (unit tier, Python 3.15, 30e3b2904, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	20284	154	2760	135	98.70%
pcapkit/corekit	2664	74	836	37	96.60%
pcapkit/dumpkit	281	3	100	3	98.43%
pcapkit/foundation	3382	27	1196	32	98.71%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	20779	220	5468	204	98.37%
pcapkit/toolkit	961	28	342	10	96.32%
pcapkit/utilities	506	7	144	5	98.15%
pcapkit/vendor	5065	2228	1160	188	52.64%

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

[JarryShaw]

With #2586 merged, this PR's checksum imports point at a path that no longer exists on main. A fix round is rebasing it onto main and retargeting them to pcapkit.corekit.checksum.internet and misc.checksum.ip_pseudo. That push will also re-run the PyPy leg that was lost to the #2600 stall.

[JarryShaw]

Rebased onto main (a25eb0e39): 0bfbe212f → 30e3b2904, with no conflicts.

• Imports: every DCCP import and reference to the checksum helpers now points at pcapkit.corekit.checksum.internet or misc.checksum.ip_pseudo. That covers make, _verify_pseudo_checksum, the data module, the test setUp and the pep.rst DCCP bullet. The only remaining old-path string is main's own negative example in the boundary test.
• Patch-id (--stable): unchanged outside those lines, at 79ff5093639d on both heads.
• Tests: 95 modules pass, test_dccp_unit passes 43 tests, and the protocols/transport leg runs 566 tests OK.
• Review: GOOD TO GO stands, since only the import paths changed. This push also re-runs the PyPy leg that was lost to the ci: the PyPy protocols shard stalls past the 30s case timeout since the checksum wave #2600 stall.