fix(ci): reconcile review-state labels against latest check run per name

[roomote]

Opened on behalf of Naved Merchant. View the task or mention @roomote for follow-up asks.

Related GitHub Issue

Fixes: #884

Description

The Label PR review state workflow evaluates a PR's CI by scanning every check run that checks.listForRef returns for the head commit, then taking .some() across all of them. A single stale failed check run — one later superseded by a green run of the same check — keeps ciFailed = true forever. As a result awaiting-author / awaiting-review labels get stripped on the next hourly run and never come back, even though the PR UI and branch protection correctly show CI as green.

Observed on #509 (and #476): the head commit carries an old failed e2e-mock run alongside later green e2e-mock runs, so the workflow logs PR #509: CI failed — stripping state labels on every run and the review-state label never returns.

Change

Reduce to the latest check run per name before evaluating ciPending / ciFailed, so the workflow sees the same current state that branch protection and the PR UI use:

const latestByName = new Map();
for (const run of checkRuns) {
  const prev = latestByName.get(run.name);
  if (!prev || run.id > prev.id) {
    latestByName.set(run.name, run);
  }
}

The latest run is selected by run.id (GitHub-assigned, monotonically increasing, never null) rather than started_at, because a freshly re-queued run can have started_at: null and would lose a string comparison against an older completed run's timestamp — reintroducing a narrowed version of the same bug.

After this change, the next schedule / workflow_dispatch run re-applies the correct review-state label to affected PRs (e.g. awaiting-author on #509).

Test Procedure

• Inspected the real e2e-mock check runs on fix: use VS Code default shell for command execution instead of /bin/sh #509's head commit (50b9df8): run.id is monotonic with recency (79875030339 failure → 79880428333 success → 82684589430 success), confirming the dedup resolves to the latest green run.
• Manually dispatched the workflow before the change: log read PR #509: CI failed — stripping state labels.
• Validated the embedded script block's JS syntax (async-wrapped parse) and confirmed the workflow's top-level YAML keys are intact.

Pre-Submission Checklist

• Issue Linked: This PR is linked to the approved issue (see "Related GitHub Issue" above).
• Scope: My changes are focused on the linked issue (one major feature/fix per PR).
• Self-Review: I have performed a thorough self-review of my code.
• Testing: New and/or updated tests have been added to cover my changes (if applicable). (N/A — no test harness exists for this workflow script; validated via real GitHub API data and a manual dispatch.)
• Documentation Impact: I have considered if my changes require documentation updates (see "Documentation Updates" section below).
• Contribution Guidelines: I have read and agree to the Contributor Guidelines.

Screenshots / Videos

Not applicable — internal CI-workflow logic change.

Documentation Updates

• No documentation updates are required.
• Yes, documentation updates are required. (Please describe what needs to be updated or link to a PR in the docs repository).

Additional Notes

The workflow's own inline comments already note that schedule / workflow_dispatch runs reconcile fork PRs normally; this change is orthogonal to that fork-handling path. PRs that were stuck unlabeled due to stale check runs (e.g. #509, #476) will recover automatically once this merges and the next hourly run fires.

Get in Touch

Roomote (task 0mzxfggobdyy)

[roomote]

1 issue outstanding; 2 checks still running. Action required. See task

• Preserve reasoning replay for Anthropic-style preserveReasoning providers instead of only replaying stored top-level reasoning_content.
• platform-unit-test (ubuntu-latest) is still running.
• platform-unit-test (windows-latest) is still running.

[roomote]

prepareApiConversationMessage() only stores top-level reasoning_content for apiProtocol === "openai" (src/core/task/apiConversationHistory.ts). Anthropic-style thinking providers in this repo still set info.preserveReasoning (for example MiniMax and the Bedrock Kimi/MiniMax models) but persist their reasoning as embedded blocks, so this branch now strips the reasoning those continuations need.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[edelauna]

Squatting this old PR - to fix a small bug in how labels are tagged