Skip to content

chore: post-merge cleanups (release-prep deletion + npm Windows ARM64 strip) - #287

Merged
aram-devdocs merged 2 commits into
mainfrom
chore/post-merge-cleanups
May 8, 2026
Merged

chore: post-merge cleanups (release-prep deletion + npm Windows ARM64 strip)#287
aram-devdocs merged 2 commits into
mainfrom
chore/post-merge-cleanups

Conversation

@aram-devdocs

Copy link
Copy Markdown
Owner

Summary

Two small cleanups that emerged from the share-readiness audit's merge train:

  1. Re-delete docs/src/ci/release-prep.md and drop the dead install-smoke-validate.sh guard that greps it. PR docs: remove stale blog, complete config reference, sweep dead PRD refs #276 had to revert the deletion when this guard surprised it.

  2. Strip wrong-arch aarch64-pc-windows-msvc mapping from the npm tarball before npm publish. cargo-dist 0.28.0 has no knob to filter supportedPlatforms, and PR docs(cli): npm-first README + flag wrong-arch Windows ARM64 mapping #277 documented the gap but couldn't fix the publish job. Closes audit finding H7.

Why

Both items were planned but deferred during the parallel Wave A dispatch to avoid scope creep on their parent PRs.

Test plan

  • tests/install-smoke-validate.sh passes after the guard removal.
  • The node one-liner is a no-op when the bad entry isn't present (idempotent / safe to keep around).
  • After v0.0.12 ships: npm view plumb-cli supportedPlatforms does not include aarch64-pc-windows-msvc.

🤖 Generated with Claude Code

PR #276 had to revert the deletion of docs/src/ci/release-prep.md because
tests/install-smoke-validate.sh greps the file for "Issue #51 verified" and
"Issue #52 wired" sentinel strings. Those issues are now resolved (Homebrew
tap and unscoped npm publish both ship as of 0.0.10/0.0.11), so the doc
page is internal noise that doesn't belong in the public book.

This commit removes both the page (already deleted via git rm in the same
working tree) and the guard that referenced it.
cargo-dist 0.28.0 emits an aarch64-pc-windows-msvc entry in the npm
package's supportedPlatforms that points to the x86_64 Windows zip. On
Windows ARM that silently installs the wrong arch.

The npm-publish job now extracts the tarball, removes the bad mapping
from package.json via a node one-liner, repacks, and publishes. When
cargo-dist ships a filter knob or we add a real Windows ARM64 target,
the strip step can come out.

Closes audit finding H7.
@aram-devdocs
aram-devdocs merged commit fdf48d2 into main May 8, 2026
16 of 24 checks passed
@github-actions

github-actions Bot commented May 8, 2026

Copy link
Copy Markdown
Contributor

I have all the information needed. Here is the full review.


PR #286 — chore(main): release 0.0.12

Type: Release-please automation. No Rust source changes.
Files touched: .release-please-manifest.json · CHANGELOG.md · Cargo.toml
Bucket: deps / ci / docs


1. Determinism

Not applicable — no Rust code touched.

2. Workspace layering

Not applicable — no Rust code touched.

3. Error handling

Not applicable — no Rust code touched.

4. Test coverage

Not applicable — no new API surface.

5. Documentation (CHANGELOG correctness)

All feat / fix / docs commits merged into main since v0.0.11 are accounted for (#276#282, #284, #285). The two chore commits (#283, #287) are correctly excluded per Conventional Commits convention. HTML entity &lt;id&gt; in the rules.<id> entry will render as <id> in GitHub markdown — correct.


Cargo.toml integrity

  • [workspace.package].version: 0.0.110.0.12
  • All six internal path-dep version pins (plumb-core, plumb-format, plumb-cdp, plumb-config, plumb-codegen, plumb-mcp): uniformly bumped to 0.0.12
  • Every crate uses version.workspace = true — no per-crate Cargo.toml edits needed ✓

Punch list

No findings.

Verdict: APPROVE

aram-devdocs added a commit that referenced this pull request May 12, 2026
…292)

* fix(ci): unstick Preflight + tolerate Dogfood warnings/nextjs flake

PR #287 deleted docs/src/ci/release-prep.md and removed the matching
guard in tests/install-smoke-validate.sh, but missed the parallel guard
in tests/release-security-validate.sh. The validator has been failing
on every CI run since 2026-05-07, skipping every downstream job (MSRV,
deny, Test, Size, Coverage, Determinism, Docs).

Drop the stale RELEASE_PREP_DOC declaration and the corresponding
sentinel-grep block. The Homebrew + npm contracts are still validated
authoritatively against dist-workspace.toml and install-smoke.yml in
sections 5 and 6 of the same script.

Two follow-on dogfood fixes since the daily run has been red since
2026-05-02 (after the Phase 7 rule additions in #202/#204/#207):

- lint-canonical-docs now tolerates exit 3 (warnings/info only)
  alongside exit 0, mirroring the test-sites leg in the same workflow.
  Errors still fail. The 318-violation steady state on the live docs
  site is a separate content/theme cleanup, not a CI bug.
- nextjs leg gets continue-on-error on both plumb invocations to mirror
  the e2e-sites stance on the upstream chromiumoxide WebSocket flake
  tracked in #233.

Also drops the dead docs/src/ci/release-prep.md reference from the
install-smoke.yml header comment.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* test(cdp): gate macos_candidate ordering test to unix

The `macos_candidate_order_lists_system_apps_then_user_apps` test in
crates/plumb-cdp/src/chrome_path.rs asserts that paths returned by
`macos_candidates` start with `/Users/example/Applications/...` —
forward slashes against `Path::display()`. On Windows, `Path::join`
uses `\`, so display[3] becomes
`/Users/example\Applications\Google Chrome.app\...` and the
starts_with check fails.

`detect_with` short-circuits to `None` on non-macOS targets, so the
function is never reached in production on Windows. Gate the test to
`#[cfg(unix)]` to keep the existing Linux coverage and drop the
unreachable Windows leg.

This test was previously masked because Preflight has been failing on
every CI run since 2026-05-07, which skipped Test (windows-latest)
entirely. Surfacing now as a side effect of the validator fix in this
same PR.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(cdp): gate macos_candidates import to unix in tests

Follow-up to 890b14c. Gating the test to `#[cfg(unix)]` left
`macos_candidates` as an unused import on Windows, breaking the
`Test (windows-latest)` build with `error: unused import:
\`macos_candidates\``. Mirror the gate on the import.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant