[JarryShaw]

Please follow the guide below

• You will be asked some questions, please read them carefully and answer honestly

• Put an x into all the boxes [ ] relevant to your pull request (like that [x])

• Use Preview tab to see how your pull request will actually look like

• 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. Narrow runs only, listed below.

• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible: N/A — centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

What is the purpose of your pull request?

Tick the commit type your subject line carries.

• fix — corrects a defect
• feat — adds a feature
• perf — changes performance, not behaviour
• refactor — changes neither behaviour nor performance
• test — tests only
• docs — documentation only
• ci — workflows or build tooling
• release — bumps the version or rolls up a distribution
• chore — anything else

---

Counted numbers, per the ruling on #1103:

• Drop counts of members, rows, registries, call sites, modules,
  crawlers, legs, warnings and test cases from test docstrings and
  comments, rewording them count-free or naming the constant that
  holds the set.
• Keep spec quantities, measurements that justify a design choice, and
  figures that derive a constant an assertion pins.

Accuracy:

• The AppType isinstance claim matches pcapkit.protocols, and the
  _make_port literals are placed in the TCP, UDP and SCTP modules.
• The bs4-importing crawler figure was wrong; the prose is count-free.
• test_const_enum_get's sweep arithmetic no longer matched the pinned
  size; it names the sets that compose the population instead.
• tests/_tiers.py names the helpers conftest actually drives.

Timed context: drop "now", "today" and "any more" in touched sentences.

Non-prose check: ast.dump is unchanged once docstrings are blanked; any workflow, TOML or script edits are comment-only (parsed and compared). Narrow test runs are green (tests/project and the tests that read edited files).

Refs #1103

[JarryShaw]

Verdict: NEEDS CHANGES at 02ed50c54 (merged as d846714dd). A Sonnet cross-review (author: Opus) confirmed every load-bearing claim:

• the single AppType isinstance site in pcapkit.protocols (transport.py:233);
• that the _make_port literals live in TCP, UDP and SCTP;
• the bs4 crawler set;
• the 117 + 2 + 0 + 1 = 120 sweep arithmetic, which it ran: 8 passed;
• the conftest helpers named in _tiers.py;
• the derivation that keeps 766.

The ASTs are identical with docstrings blanked. It found three blocking prose defects, all re-checked on main:

• a stranded possessive in test_const_enum_no_mint.py:2384;
• a singular "the last of them" paired with a plural referent in test_base_class_contract.py:35;
• a newly added "dozens each" in test_option_roundtrip_unit.py:116.

A worker is preparing a test-only follow-up for main.

[JarryShaw]

Fixed on main in 4ac931eed (test prose only, ASTs unchanged):

• the three defects are corrected;
• byte figures measured when perf(reg): drop the per-member annotation and TransportProtocol.get() calls #769 landed are now dated to it rather than to the current tree;
• a stale EXPECTED_FAILURES header is made accurate (the dict is empty);
• a default.py line citation is corrected.

A Sonnet re-review flagged the dating, and I re-checked each change before pushing. Every edited test file passes on its own.