[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (docs only)
• make test passes, and a test case covers the change (tests/project: 274 passed, 2 skipped)
• Changelog entry: 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?

• docs — documentation only

---

Description

Part of #719 (does not close it): the docs/source/pcapkit/protocols/ autodoc pages, 17 of 46 touched; the other 29 had nothing to fix or cut.

• Header tables corrected against the code: Ethernet, VLAN and OSPF octet/bit columns, the IPv4/IPv6/Ethernet/AH field prefixes, IHL's expansion.
• NGAP: an unknown IE or procedure code resolves to a non-registering Unassigned member (the page said a self-extending Unassigned_<n>).
• Unresolved references fixed (AH, PCAP Header, NGAP next-layer hook) and the mermaid click anchors for IPv4, IPv6, TCP, UDP and the four layer bases.
• Prose tightened on L2TP, VLAN, C-Tag, S-Tag, IPv6-Ext, ESP and PCAP-NG, keeping the rationale.

Sphinx -n warnings for these pages, fresh build each side: 4 on main, 1 after (the remaining AppType reference is not in scope), none new.

[JarryShaw]

Cross-review verdict on 4352cce1b: NEEDS CHANGES (ran on Opus).

One blocking item:

• docs/source/pcapkit/protocols/link/l2tp.rst:59-60 says _guess_version recognises the HTTP/2 preface, "else trial-parsing each candidate". That skips a step. pcapkit/protocols/application/http.py:356 checks for an HTTP/1 request-line or status-line (_test_start_line) before any trial parse, and commits to HTTPv1 on a match. Suggested: "recognising the HTTP/2 preface or an HTTP/1 start line, else trial-parsing each candidate".

Confirmed: the octet and bit columns (Ethernet, VLAN, OSPF, by parse), the ethernet./ah./frag./hopopt. prefixes, the NGAP unknown-code wording (by probe), the Mermaid anchors (8 of 45 missing on main, 0 on the branch), and Sphinx -n (−3 warnings, none new). tests/project: 271 passed, 5 skipped.

Non-blocking, outside this PR's files: internet/ipv6_opts.rst uses opt. where the parsed key is opts, and the ngap.py module docstring still has the old Unassigned_<n> wording.

[JarryShaw]

Cross-review verdict on af35550cf: GOOD TO GO (ran on Sonnet).

• l2tp.rst now lists HTTP._guess_version's steps in their real order: the HTTP/2 preface (http.py:288-329), then the HTTP/1 start line (:356-357), then the trial parse (:430-435).
• ipv6_opts.rst: opts. is the parsed key. It comes from ipv6.py:397, alias.lstrip('IPv6-').lower(), applied to 'IPv6-Opts'. A parse of IPv6 with next header 60 gives info['opts'] = IPv6_Opts(next=17, length=8, …) and no opt key. The octet and bit columns match the schema.
• Sphinx -n: no new warnings; main has 1303 and the branch 1300. tests/project: 275 passed, 1 skipped.

The ngap.py module docstring still has the old Unassigned_<n> wording. It goes to the #719 docstring pass.