[JarryShaw]

• Searched for similar pull requests

• Followed the coding style (make pylint, make mypy, make isort) — not run

• make test passes, and a test case covers the change — new module, protocols (non-internet) and const legs, 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

• fix — corrects a defect

---

Closes #1443

Before: FTP(b'999 hi\r\n', 8), FTP(b'099 x\r\n', 7), HTTP(b'HTTP/1.1 999 X\r\n\r\n', 20) and ReturnCode.get(777) raise EnumValueError; 440/1000 FTP and 500/1000 HTTP codes fail to parse. After: all 1000 of each parse and rebuild byte-exact via from_data; values outside the registry range resolve to an Unknown pseudo-member (not registered), -1 and 1000 still raise. make() now writes the code zero-padded (099, not 99).

Vendor templates (pcapkit/vendor/{ftp/return_code,http/status_code}.py) carry the same change; rendering both crawlers offline from today's Wikipedia/IANA snapshots reproduces the committed const/ files byte-identically (and did so on main before the change).

Tests: tests/protocols/application/test_ftp_http_unassigned_code_1443_unit.py — 9 passed / 2017 subtests; with the source reverted, 7 of 9 tests fail (949 subtests). protocols --exclude protocols/internet leg: 1119 tests, 0 failures. tests/project: 419 passed, 1 skipped.

Existing test: tests/const/test_const_enum_lookup.py drops ReturnCode and StatusCode from EXPECTED_TO_REJECT_ZERO (#492), which pinned 0 raising. Both wire fields are three digits, so 000 now parses to an Unknown pseudo-member. const leg: 502 tests, 0 failures.

[github-actions]

Coverage: 89.51% (unit tier, Python 3.14, e4a9c71a4, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18795	968	2356	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	16798	219	4520	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 e4a9c71a4 (Sonnet cross-review; the author ran on Opus).

• Round trips, all 1000 codes. Parse→from_data and make(code=N)→parse are byte-exact for every code 000–999, across FTP single-line, FTP multi-line and HTTP. On main, 440 FTP and 500 HTTP codes fail.
• No change for codes that worked. Every member, and every code that parsed on main, gives info identical to main (1181 entries). The 237 FTP/HTTP layers in the captures are also unchanged.
• Pseudo-members are never registered: __members__ stays the same size after 1000 lookups. They pickle, hash equal to the int, and use the name Unknown, which 4 other registries already use.
• FTP kind/group below 100 follows RFC 959's 3-digit form, and nothing outside the const module reads them.
• Regeneration. Re-running both crawlers on the saved snapshots gives byte-identical files.
• Tests, which I re-ran: the new module gives 9 passed and 2017 subtests. With the 4 runtime files from main, it gives 951 failed subtests.
• Neighbours. The const leg runs 502 tests with 0 failures. tests/project gives 419 passed.

For the owner: ReturnCode(0) and StatusCode(0) used to raise, and #492 recorded that as correct. They now resolve to Unknown, which follows the byte-exact rule. The PR body says so. No ! in my view, but it is your call whether that counts as public contract.