[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 — workflows or build tooling

---

Description of your pull request and other information

Phase 2 of #2060: publish the win-64 conda package. Phase 1 passed on main (run 40052211743): every windows-latest leg, 3.11–3.15, green with no warnings.

• Conda Update conda-dist: the Windows legs upload win-64 with anaconda-client, like the Intel legs (ci(conda): upload the macos-26-intel legs with the anaconda CLI (#2044) #2049). The new step is "Upload conda packages to Anaconda (Windows)", run in a bash -el login shell. The build-only dispatch guard still skips it, and test_build_only_dispatch treats the upload as a publishing marker.

• Create Release conda: gains a windows-latest leg with the same upload, gated on check_leg like the others.

• util/release_published.py: SUBDIRS (PCAPKIT_CONDA_SUBDIRS) adds win-64, so a release counts as complete only once win-64 is on Anaconda.

• util/release_conda.py: upload resolves Scripts\anaconda.exe on Windows.

• The Windows leg now blocks like the other platforms. Its continue-on-error, Windows summary step and conda-windows summary tier are removed.

• Unchanged: no new secrets or environments; it uses the existing ANACONDA_TOKEN in anaconda. The GitHub Release upload is unchanged; only ubuntu-latest does that.

• Tests: test_conda_windows_leg_2060.py is reworked, and the pins in the other conda/release tests are updated. Without the change, 28 of 56 tests fail. A 68-file selection of the workflow-reading tests gives 1224 passed. actionlint 1.7.12 is clean.

• Unverified until a real run:

  • The release job's Windows leg has never run.
  • The first real win-64 upload is the first scheduled Conda Update after this merges.
  • With win-64 in SUBDIRS, the current published version reads as incomplete. The conda job still skips it while PCAPKIT_RELEASED_ELSEWHERE is true, so win-64 first ships with the next release.

• Completeness (round 2): a version counts as complete against the platforms its tagged commit published (release_published.released_subdirs(), reading SUBDIRS at the tag). So 1.5.0b14 is not newly flagged for win-64. It is still flagged for its missing osx-64 files, which predates this PR: release: 1.5.0b14 has no osx-64 conda files, so Create Release warns it is unfinished on every released-elsewhere run #2140.

• A first real release exercises: on create-release.yml's windows-latest conda leg, the pre-setup-python steps Check the release tags name this run's commit and Check whether this leg is already on Anaconda (python3 under Git Bash, /tmp mapping, curl), then Upload conda packages to Anaconda (Windows) in both jobs. The first scheduled Conda Update after merge exercises the upload; a build-only dispatch does not.

[JarryShaw]

GOOD TO GO at 214d266fa (Sonnet review), with follow-ups in progress.

• No accidental publishing. Conda Update triggers only on schedule, workflow_run and workflow_dispatch. The Windows upload is gated event_name != 'workflow_dispatch' && windows-latest, and a manual run gets environment: ''. Only the upload steps read ANACONDA_TOKEN. In Create Release, the new upload is gated by check_leg behind the unchanged job gate. Tests pin all of this per leg.
• Upload parity with Intel: --skip-existing, token via secrets.ANACONDA_TOKEN in anaconda, a bash -el shell, and Scripts\anaconda.exe resolved on Windows. The edited conda and release tests were updated, not weakened (891 passed across 34 files).
• Being fixed:
  • With win-64 in SUBDIRS, the current published version reads incomplete (9 of 12 files). A released-elsewhere Create Release run then warns "unfinished … re-run the Create Release run", which is the wrong advice. The completeness check will exempt versions released before win-64 publishing.
  • A stale comment in experimental.yml still names the removed Windows summary step.
• Unverified until the first real release: Create Release's Windows check_leg / release_tags.py check steps, which call python3 in Git Bash before setup-python.

[JarryShaw]

GOOD TO GO at b6024227b (Sonnet delta review).

• New rule: released_subdirs() reads SUBDIRS at the version's tagged commit. A tag that predates the file reads the file's first commit, 79934ddd9. Only version_check calls it, and it checks out with fetch-depth: 0, the same pattern as release_tiers.py. ^{commit} handles annotated tags.
• 1.5.0b14: confirmed through the anaconda.org API at 5 linux-64, 5 osx-arm64 and 0 osx-64 files. The rule expects the same three platforms as before, so the "unfinished" warning (release: 1.5.0b14 has no osx-64 conda files, so Create Release warns it is unfinished on every released-elsewhere run #2140) is neither added nor removed.
• Failure mode: an unreadable tag line falls back to the current SUBDIRS, a superset, so the error leans towards "incomplete". It warns as Platforms of v… not read (#2060).
• Tests: the new test pins the SUBDIRS = '…' line format that the regex reads. The new class fails 5 of 8 cases on 214d266fa, and the release and conda tests give 432 passed.

[github-actions]

Coverage: 91.55% (unit tier, Python 3.15, b6024227b, 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	2589	75	816	37	96.48%
pcapkit/dumpkit	282	3	102	3	98.44%
pcapkit/foundation	2971	23	1036	25	98.80%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17418	219	4574	185	98.14%
pcapkit/toolkit	869	36	298	10	95.20%
pcapkit/utilities	498	7	144	5	98.13%
pcapkit/vendor	4432	1988	1018	167	51.87%

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

[JarryShaw]

Ready to merge at b6024227b.

• Review: GOOD TO GO, from the Sonnet review and the round-2 delta review.
• CI: 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 main d7aa41535: 2246 tests, 0 failures.
• After merge:
  • First real Windows upload: the first scheduled Conda Update uploads win-64; a build-only dispatch does not.
  • Create Release: the first release exercises its Windows leg.
  • 1.5.0b14: it keeps its existing "unfinished" warning for the missing osx-64 files (release: 1.5.0b14 has no osx-64 conda files, so Create Release warns it is unfinished on every released-elsewhere run #2140).

This closes #2060 (phase 2).