[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 of your pull request and other information

Part of #1660, the owner's option 1: GraalPy and free-threaded CPython run as experimental, non-blocking test legs, to be promoted only once green.

• Tier: .github/python-versions.toml gains an experimental tier with 3.14t and graalpy-25.4, spelt as setup-python v7 reads them. util/python_versions.py gives each of them one test leg and nothing more: no integration, Compat, wheel or conda leg, and no required context.
• Non-blocking:
  • The legs are named Python <version> (experimental) and run on the KTLO mechanics: the step guard, warning annotations and an Experimental summary step.
  • On these legs only, setup-python, install and tests are capped at 5, 10 and 25 minutes. A hung step therefore fails under continue-on-error, instead of the 45-minute job cancel that Required checks passed refuses (actions/runner StepsRunner.cs:322-353).
  • Every other leg's caps read 360, so its job cap binds as before.
• Updater: experimental lines are hand-maintained, and update_python_versions.py keeps them as written.
• util/ci_timing.py: it now reads a step cap written as an expression as no cap, where it used to crash on int().
• Docs: the tiers section and step caps in workflows.rst. The line citations in workflows.rst and releasing.rst follow the moved lines.
• Tests:
  • The tier, KTLO-leg, updater, unit-tests-queue, CI-timing and release-gate tests pass. The new ones fail on main.
  • Mutations of the caps are each caught.
  • actionlint reports no new finding.

[JarryShaw]

Verdict: GOOD TO GO at 296ad6b71. Sonnet cross-reviews (author: Opus) checked two rounds.

• Round 1 blocker, now fixed: a hung experimental leg could hit the 45-minute job cap and block every PR. The steps' caps now act first.
• Timeouts don't block: read from actions/runner itself (StepsRunner.cs:310-353, ExecutionContext.cs:1422-1440). A step that times out under continue-on-error ends with outcome failure and result success, so the job concludes success.
• Other legs unchanged: every job but test is identical to main. In test, the non-experimental caps evaluate to 360, above the job cap.
• Spellings: 3.14t and graalpy-25.4 are what setup-python v7 accepts, checked in its source and the python-versions manifest.
• Matrices: wheel, conda, Compat and integration are identical to main. Only test gains the two legs.
• Mutations: six cap mutations are each caught. The seventh, raising the job cap, rightly still passes.
• Non-blocking: the 25-minute test cap is a guess, since GraalPy's runtime is unmeasured. Running over shows as a warning.

[github-actions]

Coverage: 89.35% (unit tier, Python 3.15, 296ad6b71, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	3	20	3	95.16%
pcapkit/const	18860	954	2368	868	91.06%
pcapkit/corekit	2372	67	762	35	96.49%
pcapkit/dumpkit	275	3	100	3	98.40%
pcapkit/foundation	2885	144	1010	34	94.20%
pcapkit/interface	138	7	52	5	93.68%
pcapkit/protocols	17181	226	4590	190	98.07%
pcapkit/toolkit	1148	290	450	10	71.21%
pcapkit/utilities	500	4	150	4	98.77%
pcapkit/vendor	4417	2342	1010	157	43.38%

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