release: v0.1.2 — recategorize three v0.1.1 CHANGELOG entries (#28)

[cmeans-claude-dev]

Summary

Documentation-only patch release. Closes #28.

v0.1.1 shipped with three CHANGELOG entries miscategorized under ### Fixed when they belong under ### Added by Keep-a-Changelog category:

• deploy/README.md Tailscale Funnel section (originally Added in PR docs(deploy): document Tailscale Funnel as an alternative HTTPS exposure #22)
• README ## Acknowledgments and ## License sections (PR docs(readme): Acknowledgments + License sections, BigQuery dataset link, GCP credential setup pointer #15)
• README ## Install pointer to pypinfo's GCP setup (PR docs(readme): Acknowledgments + License sections, BigQuery dataset link, GCP credential setup pointer #15)

The orphaning was introduced at PR #25, which added new ### Added and ### Changed blocks at the top of the then-active ## [Unreleased] without repositioning entries from earlier PRs (#15, #22). PR #27 then inserted a ### Fixed block between ### Changed and the orphans, leaving them visually attributed to the wrong subsection. v0.1.1 shipped with that layout.

Three mechanical edits

• pyproject.toml: version = "0.1.1" → version = "0.1.2"
• CHANGELOG.md:
  • Insert ## [0.1.2] - 2026-04-27 between ## [Unreleased] and ## [0.1.1] - 2026-04-26 with one ### Fixed bullet describing the recategorization (refs CHANGELOG: three orphaned entries in [Unreleased] now appear under the wrong subsection #28).
  • Move three orphan bullets from ## [0.1.1] ### Fixed (lines 32-62 on main) to the bottom of ## [0.1.1] ### Added (after line 17 on main). The four legitimate ### Fixed bullets in v0.1.1 (Coverage 100% closes Push coverage from 99% → 100% on src/ #37, stale_threshold_days closes config: stale_threshold_days is loaded but never read — staleness warning never fires #33, _write_health closes collector: _write_health OSError escapes per-package isolation and bypasses structured exit #32, shields.io URL closes docs(readme): canonicalize shields.io endpoint-badge URL form (line 11) #16) are unchanged.
  • Update reference links: [Unreleased] now compares from v0.1.2, new [0.1.2] compares v0.1.1...v0.1.2.
• uv.lock: refreshed by uv lock so the locked pypi-winnow-downloads version (0.1.2) matches pyproject.toml.

What this does NOT do

The v0.1.1 git tag, the v0.1.1 GitHub release page, and the v0.1.1 CHANGELOG.md as it exists at the v0.1.1 tag are immutable historical artifacts and remain miscategorized. This PR corrects the in-tree CHANGELOG.md going forward; the corrected categorization ships with v0.1.2.

If you want the v0.1.1 GitHub release page notes also fixed, that's a separate manual edit on the release page after this lands — happy to do it as a follow-up.

What this is NOT

The auto-CHANGELOG workflow's underlying ordering bug for ### Changed insertion is a different issue (#26) handled in PR #41, currently Ready for QA. The orphaning fixed by this PR was a one-time manual mistake at PR #25 time, not an automation bug, so no workflow change is required.

Verification

• uv run pytest --cov — 71/71 passed, 100% coverage gate green
• uv run ruff check src/ tests/ — clean
• uv run ruff format --check src/ tests/ — clean
• uv run mypy src/pypi_winnow_downloads/ — clean
• uv.lock matches pyproject.toml — pypi-winnow-downloads v0.1.1 → v0.1.2
• CHANGELOG link refs verified by hand
• No code, dependency, or runtime behavior changes (confirmed by git diff --stat: 3 files, +23/-17, all in CHANGELOG.md + pyproject.toml + uv.lock)

After this merges

git tag v0.1.2 <merge-sha>
git push origin v0.1.2

Then verify the release at https://pypi.org/project/pypi-winnow-downloads/0.1.2/ — the trusted-publisher OIDC path is the same as v0.1.0 / v0.1.1.

🤖 Generated with Claude Code

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[cmeans]

LGTM

[cmeans]

QA round 1 — PASS

Documentation-only patch release. Verified end-to-end at head e43c365.

Release-mechanic verification

Item	Result
pyproject.toml version bump	0.1.1 → 0.1.2
uv.lock minimal pair	only pypi-winnow-downloads version line changed; no incidental dep churn
## [Unreleased] empty post-insertion of ## [0.1.2] - 2026-04-27	confirmed
## [0.1.2] ### Fixed recategorization bullet	references PR #25 / #27 history; closes #28
Three orphans relocated to ## [0.1.1] ### Added	Tailscale Funnel, Acknowledgments + License, Install pointer — confirmed at file body
Four legitimate ## [0.1.1] ### Fixed bullets retained	Coverage 100% (#37), stale_threshold_days (#33), _write_health (#32), shields.io (#16) — confirmed
Link refs	[Unreleased] compares from v0.1.2; new [0.1.2] compares v0.1.1...v0.1.2; existing entries unchanged
Date matches today	2026-04-27
Single commit since v0.1.1 tag	no missing CHANGELOG entries; ## [Unreleased] correctly empty
Issue #28 scope	fully delivered (issue suggested moving orphans under ### Added; PR does so under v0.1.1's ### Added since v0.1.1 already shipped, with the recat itself filed as a v0.1.2 ### Fixed entry)

Functional verification (worktree at head sha)

• uv run pytest --cov → 71 passed, 100.00% total coverage (fail_under = 100 gate satisfied)
• uv run ruff check src/ tests/ → clean
• uv run ruff format --check src/ tests/ → clean
• uv run mypy src/pypi_winnow_downloads/ → clean
• python -c "import pypi_winnow_downloads; print(pypi_winnow_downloads.__version__)" → 0.1.2 (resolves via importlib.metadata post-install)
• CI all green at e43c365: lint, typecheck, test 3.11 / 3.12 / 3.13, deploy-smoke

SemVer fit

Patch release appropriate — the change is documentation-only inside CHANGELOG.md plus the version bump in pyproject.toml and uv.lock. No public-API delta, no dependency change, no runtime behavior change. The PR's "What this does NOT do" section honestly captures that the v0.1.1 git tag, GitHub release page, and tagged-CHANGELOG remain immutably miscategorized; this release fixes the in-tree forward record only. The follow-up offer to manually edit the v0.1.1 GitHub release page notes after merge is the right approach.

Findings

None.

Informational (not blocking)

• .github/ISSUE_TEMPLATE/bug_report.yml:14 placeholder still reads "0.1.0 or abc1234" — same lag I noted on PR #39 round 1, operator's call to refresh whenever convenient. Not in this PR's scope.

Verdict

Ready for maintainer QA Approved.

[cmeans]

Switching label: QA Active → Ready for QA Signoff. Round 1 PASS, zero findings, CI fully green at e43c365. Awaiting maintainer QA Approved.