[JarryShaw]

Please follow the guide below

• Searched for similar pull requests
• Followed the coding style
• make test passes, and a test case covers the change (tests/project/test_pr_template_commit_types_1102_unit.py and tests/project/test_python_versions_update_1660.py run)
• Described the user-visible change below and left its changelog entry to the maintainer (N/A)

What is the purpose of your pull request?

• chore — anything else

---

Description of your pull request and other information

Closes #2449

Feature PRs no longer add entries under docs/source/changelog/ or to CHANGELOG.md; the maintainer writes them centrally. CONTRIBUTING.md says so, and this PR also rewords the template's changelog box to match (not deferred). util/update_python_versions.py and its test follow the new box wording. No runtime code or release content changed.

[JarryShaw]

NEEDS CHANGES (Sonnet, 30953ed00). CONTRIBUTING.md change is correct and contained; no code, release content or counts changed.

• .github/PULL_REQUEST_TEMPLATE.md still tells contributors to add an entry and regenerate CHANGELOG.md, and links to the section this PR rewrites. Reword it in this PR, with util/update_python_versions.py (~L1301) and tests/project/test_python_versions_update_1660.py (~L879), which pin the current wording.
• If the template is instead deferred, the body must say so and Closes #2449 becomes Refs #2449.
• Non-blocking: "Feature pull requests" (L61) is narrower than L28; "Open" (L61) is undefined; L28 is 125 columns.

[JarryShaw]

NEEDS CHANGES (Sonnet, 4f2af68ee). Regex and pinned test verified; no code changed.

• CONTRIBUTING.md ~L69 is 102 columns. Break before "the" so line 70 starts "the .rst entry, then regenerates:".
• PULL_REQUEST_TEMPLATE.md ~L10: "described … above" should say "below" (the description is below the checklist); update test_python_versions_update_1660.py ~L879 to match.
• Non-blocking: "Feature branches" (CONTRIBUTING ~L64), "docs-only" in the body (the diff touches util/ and tests/), "open versions" in the title.

[github-actions]

Coverage: 94.45% (unit tier, Python 3.15, 5732de283, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	106	4	20	2	95.24%
pcapkit/const	20918	175	2908	156	98.57%
pcapkit/corekit	2918	77	942	38	96.81%
pcapkit/dumpkit	286	3	102	3	98.45%
pcapkit/foundation	3512	27	1218	32	98.75%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	22839	157	5950	127	99.00%
pcapkit/toolkit	961	26	342	8	96.62%
pcapkit/utilities	565	7	152	5	98.33%
pcapkit/vendor	5590	2286	1280	214	55.75%

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

[JarryShaw]

Local-only commit 30953ed forks from 4ca3f30 and is redundant with origin a9db8a4 (rebase onto origin is empty). Nothing to push. Local commit to be dropped.

[JarryShaw]

Add to convention and session/sweep memo: upon each release, amend inline or post updates in thread to the PEP's Discussion post. PEP doc is source of truth but aim to keep the Discussion also up-to-date.

[JarryShaw]

Ruling reconciled: upon each release, amend pep.rst inline and update Discussion #106 ("PyPCAPKit Enhancement Proposals") to match -- the page stays authoritative. Documented in process.rst (20f39eb73) and in the sweep memo.

[JarryShaw]

Owner ruling (2026-11-01), paraphrased: changelog entries are consolidated in this PR and stripped from the open feature PRs, so they stop conflicting. The ruling named #2383, which is a merged HIP fix, so I've taken it to mean this PR (it implements #2449's practice). Say so if that's wrong.

In flight on this branch: merge current main, add #2509's Radiotap entry, and reconcile the 1.6.0 entries this merge wave left stale. When that is pushed, the label resets to review: pending and a fresh review runs.

[JarryShaw]

CI note on 1818bb3a2: one job failed, Python 3.15 (LATEST tier, half 2). The failure is test_register_restores_shipped_descriptor_1504_unit::test_restoring_imports_nothing, which raises KeyError: 'pcapkit.protocols.link.arp.protocol'.

This PR changes only the changelog. Main passes the same job, and the module passes when run on its own. The cause is in the test: its _unimported helper pops the shipped module from sys.modules on the assumption it is already imported, and since #2757's lazy imports that depends on which tests ran first. A test-only fix goes straight to main. I'll re-run this job after that.

[JarryShaw]

GOOD TO GO at 0cd0d2044, for the entries folded so far (cross-review of the #2770 correction, and my own read of folds 26–28).

• The refactor(protocols)!: replace the pcapkit.protocols.__proto__ dict with ProtocolBase.__registry__ (#2763) #2770 entry is corrected against v1.5.1. A Protocol subclass already registered itself on definition there. 1.6.0 skips abstract subclasses, and on a name clash keeps the existing entry, where 1.5.x replaced it.
• The fix(protocols): rebuild a cut standalone header with the values its parse read (#2761) #2767 sentence no longer claims the IPv6_Ext length fix, which landed separately in fix(protocols): rebuild a cut header at the width its parse read it (#2761) #2777.
• Fold 28 adds fix(internet): rebuild a whole extension header in the extension mode its parse used (#2768) #2774 and fix(tcp): rebuild a TCP header with the End of Option List padding its parse read (#2769) #2775, checked against a v1.5.1 export.
• Main's 1504 test fix is merged in. Project tests, Lint and deploy-pages pass, and changelog_md.py --check is clean.
• Please hold the merge until the last fold. fix(protocols): rebuild a cut header at the width its parse read it (#2761) #2777, fix(pcapng): rebuild a block's info with the options and trailer its parse read (#2776) #2778 and the still-open docs(conventions): settle the member-order page wording and align the protocol registry with the other maps (#2772, #2773) #2779 and fix(utilities): concurrent first access to lazy protocol exports deadlocks the import system #2771 get one more fold once they land, so this PR stays last.

[JarryShaw]

GOOD TO GO at 26a038916, after the final fold for #2777, #2778, #2779 and #2781 on top of main 812ed0fe2. A Sonnet cross-review of the fold delta checked these claims against code, and each one held:

• __short_width__ is carried by to_dict() and read back by from_data.
• On v1.5.1, rebuilding a PCAP-NG block after opt_endofopt lost padding_opts and its options. The block now round-trips.
• The clash warning's namespaces are sorted now; v1.5.1's order was unsorted.
• A name restore now warns "overwriting"; v1.5.1 restored it silently.
• import_lock is public, and the never-released registry lock is not named.

I checked the SCTP, layer-grouping and page-order sentences against #2779's commit. changelog_md.py --check passes. The fold touches only the changelog files.

[JarryShaw]

GOOD TO GO at 5732de283. Since 26a038916 this adds main's #2780 guard (3d45352c1) and one clause in the member-order entry: a project test now keeps the protocol pages in source order (#2780). I checked the delta myself; it touches only that clause and the regenerated CHANGELOG.md, and changelog_md.py --check passes. Project tests, Lint and deploy-pages are green.

[JarryShaw]

Note on 5732de283: the non-blocking GraalPy experimental leg failed one test, test_an_abstract_subclass_is_skipped. That comes from main (#2770), not from this PR, which changes only the changelog. It is filed as #2784 and a fix is in progress. This PR's content verdict stands. Since this PR merges last, please hold it until the #2784 fix lands.