[JarryShaw]

• 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 — 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?

• ci — changes to CI configuration files and scripts

---

Description of your pull request and other information

For #2060: the plan asks for a Windows unit-test run before the win-64 conda package is published.

• New windows job in experimental.yml: named Python 3.15 on Windows (experimental), on windows-latest. It uses 3.15, the latest tier, which is also the newest win-64 conda build. It has the experimental job's needs/if, shell: bash, and LF checkout.
• Stays green: continue-on-error on the job and every step, plus step caps (setup 5 min, install 10, tests 60) inside an 80-minute job. Failures become warnings through leg_summary.py experimental.
• Linux-only parts dropped: the memory scope, prlimit/GNU timeout, the resource monitor, the superseded watcher and the file-size cap, since Windows has no resource module.
• Updater change, for the owner: update_python_versions.py now ignores continue-on-error jobs when collecting the runners latest must have a build on. Without that, a missing Windows build would hold latest back for blocking jobs.
• Tests: test_windows_test_leg_2060 (24 tests) fails 21 against main and catches six planted mutations. The job-set pins in five workflow tests now list the new job. actionlint 1.7.12 is clean.

The first run will show what fails on Windows, so expect warnings, not red.

[JarryShaw]

GOOD TO GO at 416b5a788 (Sonnet review).

• Stays green: every step and the job carry continue-on-error, the step caps sit under the 80-minute job, the summary and upload steps tolerate missing files, and no /usr/bin/python3, sudo or /proc appears. The red-on-lost-runner limit is the same as on the existing experimental job.
• Gating: nothing gating references the job, and check_suite_report.py classifies it as non-blocking through its (experimental)$ pattern.
• Updater: the predicate is a literal job-level continue-on-error: true, so only the two experimental jobs are skipped. Other blocking latest jobs still pin the runner set. Recommendation: accept.
• Tests: they depend on the change; main's workflow fails 22 of 27, and three planted mutations each failed.
• Minor, no change:
  • A superseded leg runs to completion instead of being stopped, which costs minutes only.
  • Checkout and the LF step have no cap of their own.

[github-actions]

Coverage: 91.54% (unit tier, Python 3.15, 416b5a788, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19599	880	2618	866	92.12%
pcapkit/corekit	2565	71	812	37	96.56%
pcapkit/dumpkit	282	3	102	3	98.44%
pcapkit/foundation	2930	23	1022	25	98.79%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17414	219	4574	185	98.14%
pcapkit/toolkit	869	36	298	10	95.20%
pcapkit/utilities	473	7	136	5	98.03%
pcapkit/vendor	4432	1988	1018	167	51.87%

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

[JarryShaw]

Ready to merge at 416b5a788.

• Review and CI: the Sonnet review is GOOD TO GO, and every blocking check passed (Required checks passed SUCCESS). The 3.8–3.10 KTLO legs carry no warnings.
• Ordering: the tests/project plain-unittest ordering leg passed on a scratch merge into current main (d2b9b324b): 2138 tests, 0 failures.
• First Windows run: it behaved as designed. The check is green with a single tests failed … (experimental, non-blocking) warning. pytest reported 48 failed, 2074 passed, 1121 skipped and 4 errors in 4m12s.
  • The failures cluster in foundation/test_extraction* (no-EOF pipes, non-seekable streams), corekit/test_io, test_tier_guard and the vendor snapshot restore.
  • The 4 errors are SIGALRM-based helpers (time_limit, _deadline), which have no Windows counterpart.
  • Once this merges I'll file them, grouped by cause, as phase-2 prerequisites on Conda: bring back the Windows (win-64) package, in two phases #2060.