Repository navigation
fix(ci): publish Rust releases before proposing the next PR - #7818
Conversation
Quest-Claim: 139d7862a3cb2b7a
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed head 18205b5. I found no actionable issue in the release-order change. Publishing before proposing addresses the observed stale proposal, and failure still stops the recipe before the next step. The quest correctly leaves the release-branch backport and live verification open.
Verification: GitHub-only static review of the recipe, workflow and pinned release-plz behavior. I did not run tests, publish a release, or exercise the deployed workflow.
(Written by review (OpenAI))
|
No actionable comments were generated in the recent review. WalkthroughThe Severity of issue fixed: Low Merge Risk:
|
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
Reviewed the two-file change that reorders the Rust release recipe and updates the retained quest.
- Recipe order —
rs/justfilenow runsrelease-plz releasebeforerelease-plz release-pr, so a just-merged version publishes to the registry before the proposal step runs.justaborts on the first failing line, preserving the intended failure boundary. - Quest record — the plan now carries the confirmed run timeline and keeps the
releasebackport plus live release verification under## Waiting.
The rationale checks out against upstream: release waits on registry publication (wait_until_published) before returning, and release_pr returns None when the update set is empty, so a merged release is no longer re-proposed. The no-dry-run and no-upstream-precedent notes are accurate, and the remaining deployment risk is documented rather than claimed.
deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) | 𝕏

Problem
On the release run for moq-dev/moq#7651,
release-prran at 19:56:23Z, compared local moq-tokio 0.19.33 with registry 0.19.32 at 19:56:50Z, and opened moq-dev/moq#7750 at 19:57:17Z after warning about duplicate 0.19.33 changelog entries. Only then didreleasestart at 19:57:18Z and publish 0.19.33 at 20:01:14Z. The proposal observed the merged release before it was published.Approach
Run
release-plz releasebeforerelease-plz release-prin the existing recipe. A publish failure stops the recipe before PR creation. Keep the quest accurate about the remaining deployment and verification.Primary-source checks:
releasesource waits for registry publication before returning success;release-prproposes only when the package update produces changes.c19c3f8. It also runsrelease-prfirst. This corrects the plan's assumption about copying its order; the recorded failure establishes why our recipe needs the reverse.release-plz release-pr --helphas no dry-run option. The action'sdry_runinput applies only torelease, so using its default command as a safe reproduction would still create a real PR.Impact
Alternatives
Keep the fix in the shared recipe. Copying the action's order retains the observed problem. Separate jobs or a new workflow add coordination without changing the publication boundary. No retries or delays are needed.
Validation
quest gate just ci origin/main: pass in the Nix dev shell, including 11,641 Rust tests (133 skipped by the existing profile), Rust checks, repository lints, and tooling tests.quest check --base origin/main: 603 documents pass;git diff --check: pass.release-plzcommand stub: original recipe reproduces the wrongrelease-pr,releaseorder. Fixed recipe runsrelease,release-pr; a failure inreleasestops beforerelease-pr, and either command's exit 42 propagates. Token forwarding is checked with a non-secret fixture value.git merge-tree --write-tree origin/main HEAD: no conflicts.No packages or release tags were published, and no real release or release PR was run for reproduction. The disposable command-stub check validates orchestration, not a deployed release outcome.
Review
Claude Code,
claude-opus-5-5at medium effort, round 1, reviewed18205b52e1b424822af875ab65fa019313b7c002againstorigin/mainatcd5a8476cb21681ad08f5ac96eb433253f57547c. No findings.Deployment
The implementation is ready to land independently. To finish the retained quest, land this fix on
main, then backport only the release recipe toreleasein a separate PR under the main-first backport rule. On the next approved release merge, verify the Release RS log publishes before proposing and opens no stale PR. No post-fix release has run; this PR claims the checked implementation, not deployment evidence. No maintainer design decision remains.Follow-ups
(Written by GPT-6.1 Sol)