[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

• 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 of your pull request and other information

Refs #1538: P4 option (a), P6 and P8. #1538 stays open for P1, P2 and P7.

• P4(a). unittest-ordering gets needs: [test]. Its if: is the old one with !cancelled() && in front, so the cells still run when test fails or is skipped, and a red cell still fails the push run that Pages, Conda, Vendor and Release wait on. No concurrency group is added (ci: stop dropping a commit's validation when the next one lands #394).
• P6. Run the 3.15 cell (experimental, non-blocking) gets if: ${{ github.event_name != 'pull_request' }}. The name and continue-on-error: true are unchanged. This narrows ci: engines not exercised across their claimed Python ranges — PyShark has no real-capture coverage, PyPCAPFile dark on 3 of 5 legs #845's ruling to main pushes.
• P8. DPKT is removed from engine-tests' matrix and from both case arms. The premise holds: the cell's only TEST_PATHS entry was tests/toolkit/test_dpkt_unit.py. test installs .[test,DPKT,crypto,NGAP] on 3.10–3.14 and runs that file, since none of its --ignores match it. The DPKT job failed in 0 of the 120 cached runs from Oct 5–9. What is lost: that file run in a venv with only .[test] and DPKT, and run on 3.15.
• Required checks. Nothing is renamed, required-checks' needs: is unchanged, and no other workflow needs: these jobs. No permission is added. The net line change above changelog: is zero, so every unit-tests.yml:N citation still resolves.

Saving. Figures are estimates unless measured.

• Measured on 3 PR runs: each used 76.9–77.8 runner-min. The DPKT job took 2.9–3.6 min (median 3.07 over 120 runs). The other five engines' 3.15 steps took 133–161 s per run.
• That is about 5.6 runner-min per PR run (~7%) and ~3.1 per push run. At Oct 5–9's rate of 77 PR and 48 push runs a day, that comes to ≈580 runner-min/day (estimate).
• P4(a) saves 0 runner-min. In push runs 38008314081 and 38008792570, the cells started 13–15 min before that run's own Python legs, and those legs queued for 13–19 min.
• The cost of P4(a): the push run now ends at least ~5.8 min later. That figure is the slowest cell's run time and excludes queue wait (estimate). Any gain in PR latency is unmeasured.

Tests. Three new TestJobGating cases each fail on main's workflow. The tier-guard doctoring now edits the Scapy/PyShark arms instead of the retired DPKT ones. Results: tests/project 469 passed, tests/test_tier_guard.py 109 passed.

[JarryShaw]

Verdict: GOOD TO GO at e184be0ec. Cross-reviewed on Sonnet; the author ran Opus.

• P4(a): needs: [test] together with !cancelled() means the cells still run when test fails, and a red cell still fails the push run. The push run is what deploy-pages.yml:66-70, cron-conda.yml:64-68 and cron-vendor.yml:68-72 key on, so publishing is still gated. !cancelled() matters here: in 7 of the last 8 failed push runs, test and protocols (rest) were red together.
• P6: only the 3.15 step gains if: event_name != 'pull_request'. Its name and continue-on-error are kept.
• P8: the DPKT cell's only path, test_dpkt_unit.py, already runs in test on 3.10–3.14. What is lost is the clean .[test]+DPKT venv and DPKT on 3.15. Nothing else referenced the cell.
• Required contexts, permissions and doc citations: all unchanged. TestDocCitationsResolve gives 249 subtests passed. tests/project gives 469 passed, and the three new TestJobGating tests fail on main's workflow.

For the owner, P4(a) measured: this comes from 60 push runs and 127 PR runs, replayed against the 20-slot pool.

• PR runs: about 0.7 min faster on average (mean wall time 20.1 → 19.4 min, p90 47.2 → 43.6).
• Push runs: they finish about 2.6 min later at the median, so Pages, Conda and Vendor publish that much later.
• Verdict: it is a small, roughly even trade, not a clear win. The replay omits CodeQL, Compat and Lint, so the real contention is higher.

Not blocking:

• In workflows.rst, "ahead of that run's own Python legs" reads as if it still holds after the change.
• "About 6 min" is the p90; the median delay is 3.2 min.

[github-actions]

Coverage: 90.02% (unit tier, Python 3.14, e184be0ec, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18795	842	2356	756	92.09%
pcapkit/corekit	2204	67	712	34	96.26%
pcapkit/dumpkit	258	2	90	2	98.85%
pcapkit/foundation	2797	132	986	48	94.18%
pcapkit/interface	108	7	36	5	91.67%
pcapkit/protocols	16992	229	4546	190	98.03%
pcapkit/toolkit	635	98	198	5	83.31%
pcapkit/utilities	431	4	124	4	98.56%
pcapkit/vendor	4409	2342	1006	157	43.25%

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