[mldangelo-oai]

Summary

Three pull request retargets started CI on a test merge whose first parent was still the previous base. In two of those cases, the checkout omitted changes from the requested target, including the shard runner in one case and the CI classifier in the other. Shard runner checkout · Classifier checkout

Changes

On a pull request retarget, compare the checked-out merge commit's ordered parents with the base and head SHAs in GitHub's event payload before selecting jobs. Report the requested base ref and both SHA pairs. A mismatch fails classification and propagates through the existing required gate.

Testing

• All 10 Node CI selection and result-gate tests passed.
• All 18 workflow-quality tests passed with 196 assertions. The Git fixture checks matching parents, a stale base, and a stale head. Each mismatch exited before emitting selections.
• A hosted draft changed target in both directions without changing its source head. In each direction, the event carried the new base and Actions fetched the previous merge. The comparison rejected both before source jobs started. The final patch generalizes the diagnostic wording. Forward retarget · Return to main
• Replaying the hosted event's actual SHAs against both Git merge objects at depth two rejected the stale commit with zero outputs and selected the full matrices on the correct commit.
• The focused release dependency test passed with 69 assertions. Actionlint with ShellCheck, Prettier, and the whitespace check passed.

Risk and rollout

A retarget with a stale checkout stops in the Linux classifier before the expensive jobs start. A new pull request event after GitHub refreshes the merge ref can test the updated tree. GitHub uses the original commit when rerunning an existing attempt. GitHub rerun documentation

Public disclosure review

• No customer, partner, prospect, or user identities, data, or identifying details are included.
• No credentials, personal data, private source, scan findings, or nonpublic links or tickets are included.
• I reviewed the branch name, title, description, commits, changes, comments, logs, screenshots, attachments, and links for public disclosure.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-24T17:06:13.043394Z	c4e2c8a	PR opened
🔒 Security Review	✅ Completed 2026-10-24T17:06:43.755049Z	c4e2c8a	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[alandelong-oai]

Thanks — no blocking findings in this owner-level merge-safety pass. Retargeted runs now stop before emitting selections when the checked-out merge parents differ from the event, and the existing aggregate gate propagates that failure. The change stays confined to the classifier and its fixture.

Validation: all 10 Node selection/gate tests and 19 focused Bun workflow/release tests passed (265 Bun assertions), along with SDK/MCP typechecking, changed-file formatting and whitespace checks. Reviewed both hosted retarget failures and their actual parent pairs. Full platform CI was not rerun locally.