[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

---

Description

Part of #1103, the whole-repository prose sweep, for the docs slice. It follows the #719 rules: prose only, accuracy first, no churn and no timed context.

• Directives: versionchanged 1.5.0 on the install section for the narrowed
  all extra and the new dev extra (design: write down what belongs in the all extra, and why #910).
• Toctrees: Utilities fourth in pcapkit/index.rst; OSPF under Application and
  Miscellaneous last in const and vendor; MH inside the IPv6 extension-header
  block. Conventions, foundation, corekit, misc and utilities preambles follow
  their toctree order.
• ext.rst: from_data raises UnsupportedCall for caller keywords and warns only
  for keys from _make_data.
• pep.rst, testing.rst, test-coverage.rst: NULL/LOOP/RAW link types are bound,
  EXPECTED_FAILURES is empty, SKIP holds three, CI runs the fixture tier, the
  dumper round trip is asserted, test: round-trip coverage for every protocol code, dumper and lossless path #1202 is closed.
• protocol-layer-placement, process, registry-protocol and conf.py: figures
  re-measured against the tree.
• Markup: nine bold spans that nested an inline literal split, three headings
  in Title Case, bare issue citations linked, timed wording removed.

tests/project: 930 passed, 1 skipped.

Checks:

• An AST guard holds on every changed .py file.
• tests/project gives 930 passed and 1 skipped, and every test that reads the edited paths passes.
• A merged Sphinx build of all nine slices adds no new warnings.
• A Sonnet cross-review sent this back once. After the fix round, a re-check returned GOOD TO GO.

Not covered here; a round-2 PR follows: the CI-timing and GraalPy figures in workflows.rst, which were not re-derived, and the html-theme block of conf.py.

Refs #1103

[JarryShaw]

Verdict: GOOD TO GO at c4846b423. A Sonnet cross-review checked it (author: Opus), after one fix round.

• First review, one blocking finding: the name count in protocol-layer-placement.rst was off.
• Re-check: re-measured from vars(). Link has 10 names including _shipped_proto, Application has 9, they share 6, and Link overrides 4 that Application does not.
• Toctree reorders: applied as the inventory found, re-verified against main.
• conf.py: the AST is identical; only comments changed.
• Tests: tests/project gives 930 passed.
• Non-blocking: tests/protocols/application/test_layer_placement_unit.py:200 still says seven names; round 2 fixes it.