Repository navigation
fix(release): name the package npm refuses to stage, and check trusted publishers first - #450
Conversation
…d publishers first A trusted publisher set up while releases went out directly may not allow npm stage publish, which the next release is the first to use. The publish job now names the refused package and the fix.
Codex Review Summary: ❌ FailedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
There was a problem hiding this comment.
✅ No new issues found. No P0 or P1 findings remain.
Reviewed changes Reviewed the release workflow’s staging-failure handling and the accompanying trusted-publisher guidance.
- Package-specific diagnostics: Failed staging now identifies the package and version, suggests checking trusted-publisher permissions for E403/E404 responses, and preserves failure propagation so dependent waves do not run.
- Publisher checks: Release instructions now call for manually inspecting existing trusted publishers before the first staged release.
Validation: All three tests in test/npm-trust.test.ts passed. Shell syntax and stubbed executions of the actual publish step also passed for successful staging, E403 failure, and already-published versions. Live npm publisher settings and OIDC staging were not exercised.
gpt-6.1-sol(xhigh) | 𝕏
comark-annotate
comark-badge
comark-bundle
comark-cards
comark-carousel
comark-cast
comark-charts
comark-chat
comark-chips
comark-cjk
comark-code
comark-code-notes
comark-colors
comark-comments
comark-critic
comark-deflist
comark-details
comark-diagrams
comark-diff
comark-embed
comark-fields
comark-figure
comark-file-tree
comark-footnotes
comark-git-graph
comark-github
comark-icon
comark-include
comark-keys
comark-links
comark-linter
comark-magic-move
comark-mark
markdown-it-to-comark
marked-to-comark
comark-mathml
comark-media
comark-metrics
migrate-to-comark
comark-outline
comark-properties
comark-qr
comark-quiz
comark-reading-time
remark-to-comark
comark-sources
comark-steps
comark-table
comark-tabs
comark-tests
comark-time
comark-ui
comark-xref
comark-zoom
commit: |

With pnpm's
trustPolicy: no-downgrade, comark-ui@0.12.2 and comark-linter@0.12.2 don't install. Drop's report skill hits it onvp dlx --package comark-linter --package comark-ui comark-lint:With a fresh store, pnpm 11.18 refuses three packages, one after another: comark-zoom, comark-annotate and comark-carousel, all at 0.12.2. Their first releases (0.1.0 to 0.2.0) were staged and approved with 2FA. #42 switched to direct
npm publishon 2026-10-20, so every version since then has one rank less evidence: annotate and carousel from 0.4.1, zoom from 0.7.1. The other 51 packages were never staged, so pnpm accepts them.#435 already moved the release back to
npm stage publish, with the same command and flags uppt's publish action used for the staged versions. It merged about four hours after 0.12.2 went out, so no release has used it yet. The next one, v0.13.0 (#448), stages every package, and pnpm accepts a staged version after a staged one. Installs ofcomark-ui@latestwork again once its packages are approved. The 0.4.1 to 0.12.2 versions stay blocked, and since there won't be a 0.12.3, projects on^0.12need to move to^0.13.One thing could still stop that release. A trusted publisher that only allows
npm publishrefusesnpm stage publish. The publishers changed after the staged releases: comark-zoom@0.2.0 and @0.12.2 carry differentoidcConfigIds. I couldn't read the current settings: the npmjs.com page needs the security key, andnpm trust listneeds a one-time password. This PR names the refused package and the fix in the publish job's error, instead of only failing the wave, and AGENTS.md now says to check the publishers withnpm trust listbefore the first staged release.test/npm-trust.test.tspasses. I also ran the publish function against a stubbednpmthat answers E403: it prints the new error and xargs exits 123, so the later waves aren't staged. Only a real release can prove that staging goes through.