fix(scheduler): retry OpenCode after coverage blockers clear by seonghobae · Pull Request #1266 · ContextualWisdomLab/.github

seonghobae · 2026-08-23T16:51:36Z

Protected-main replacement for #1152. It keeps the source-backed coverage-only retry and removes the stale audit-lock and unrelated docstring churn from the conflicted predecessor.

Root cause

A current-head OpenCode review can request changes solely because coverage-evidence was not yet available. Once exact-head coverage and Strix evidence become successful, the scheduler still returned at the existing CHANGES_REQUESTED gate, so the same head never received the review needed to clear that gate.

The original retry also used the environment-sensitive is_opencode_context() classifier to ignore its superseded failing OpenCode CheckRun. Under the central required-workflow deployment that classifier deliberately treats CheckRuns as non-authoritative placeholders, which made the placeholder failure self-block the retry.

Change

authenticate the latest exact-head OpenCode change request as coverage-only from the established gate markers;
require successful exact-head coverage and Strix evidence and no unrelated failed checks before dispatch;
ignore only the superseded OpenCode failure, including the central required-workflow placeholder;
keep unresolved review threads, ordinary source findings, unrelated failed checks, disabled dispatch, and active same-head runs fail-closed;
document the operational boundary in CHANGELOG.md.

Exact local evidence

Base: 885f2cd251999f21cf562cab3e2d9cc3cc3ec737
Head: 9f2ab9ae2f97be49fc4ab4dbcd13123057005b97
RED: the required-workflow placeholder regression returned block instead of review_dispatch before the fix.
Focused scheduler: 115 passed.
Full central suite in a fresh uv environment with the independently tracked pip bootstrap supplied through uv run --with pip: 1394 passed, 1 skipped, 16 subtests passed.
Coverage: 8385 statements and 3240 branches, 100%.
Production docstrings: 100% across scripts/ci.
compileall and git diff --check: pass.

No approval, protected check, or unresolved-thread requirement is bypassed. An independent exact-head verdict remains required before merge.

coderabbitai · 2026-08-23T16:52:10Z

Warning

Review limit reached

Next included review available in 48 minutes.

View limit details

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: bf66e18f-23e7-4081-9004-b89e0aaa0da1

📥 Commits

Reviewing files that changed from the base of the PR and between 0c6b9a6 and 0ffad34.

📒 Files selected for processing (3)

CHANGELOG.md
scripts/ci/pr_review_merge_scheduler.py
tests/test_pr_review_merge_scheduler.py

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

seonghobae · 2026-08-24T04:05:16Z

Implemented in 9cc3aca.

The same-head coverage retry-floor branch now disables an existing auto-merge request before returning the wait decision. Added a regression test covering an auto-merge-enabled current-head coverage-only review; the scheduler now remains fail-closed until the retry floor elapses without leaving auto-merge active.

Verification: 1408 passed, 1 skipped, 16 subtests; compileall; git diff --check.

seonghobae · 2026-08-24T05:45:42Z

Exact-head check RCA: Strix run 32688700442 failed closed after provider-infrastructure output, no report artifact, NVIDIA fallback failure, and direct fallback LLM Provider NOT provided. No source vulnerability evidence is present; revalidate after the provider/failover root is healthy.

seonghobae · 2026-08-24T06:07:10Z

Exact-head update: 2a30943

Fixed the valid coverage-retry finding. The same-head retry floor now uses the later of the exact-head coverage review submission and the latest completed exact-head OpenCode repository-dispatch start. If dispatch history cannot be read, the scheduler waits fail-closed. This prevents repeated dispatch when a review run completes without publishing a new review.

The central-deployment progress observation is retained as an intentional optimization boundary: CheckRun placeholders are non-authoritative, while dispatch_opencode_review remains the authoritative same-head active-run deduplicator.

Local verification on this exact head: 1415 passed, 1 skipped, 16 subtests; statement and branch coverage 100%; public scheduler docstrings 100%; ruff check, compileall, and git diff check passed.

Please re-review this exact head and publish substantive exact-head verdict evidence.

seonghobae · 2026-08-24T06:07:18Z

@opencode-agent review exact head 2a30943. Read the complete changed-file diff, verify the dispatch-history retry floor and central active-run boundary, then publish substantive review evidence tied to this exact head.

seonghobae · 2026-08-24T06:35:04Z

Exact-head finding disposition

Current exact head before restack: 2a30943

Central-deployment progress: the progress value is an optimization signal. Correctness remains guarded by the exact-head dispatch deduplication result from dispatch_opencode_review, so an OpenCode CheckRun that is not visible to the progress probe cannot create a duplicate dispatch.
Completed-run lookup: the additional request is limited to the non-dry-run coverage-only retry path and reads the newest bounded page; it supplies the retry anchor when no review publication exists. Removing it would reintroduce repeated hourly dispatch after a run completed without a review.

These are intentional bounded trade-offs rather than source defects. I will restack this branch onto the current main and request a fresh exact-head review after the base changes.

seonghobae · 2026-08-24T06:36:11Z

Restack complete: exact-head re-review required

The branch was updated through the normal PR path onto current main. Current exact head: 0ffad34. Current exact base: 0c6b9a6.

All predecessor checks and reviews are stale after the base update. The scheduler retry-anchor fix and its regression tests are present; the two informational findings were dispositioned with their correctness and bounded-cost rationale. Please inspect the complete current diff and publish a substantive review for this exact head. No merge is requested until required checks and independent approval are current.

seonghobae · 2026-08-24T07:20:30Z

Exact-head stack verification receipt

Exact root/base PR: fix(scheduler): require independent exact-head approval #1270 at 6cb714470c5a0c31b36cb64df8959cd020f22d52
Exact follow-up PR head: fix(scheduler): retry OpenCode after coverage blockers clear #1266 at 0ffad34c087a56f1ceb526148bc773b0667a2902
Retargeted base: codex/scheduler-independent-exact-head-approval-20260824
Synthetic stack commit: bac4bb9add4c27c199980d68c1568143ad1eff35
Verified stack tree: 4dde1e61e5583c9e9df4e5929dc5ed979d428376
Stack conflict: none
Full merge-result suite: 1428 passed, 1 skipped, 16 subtests
Production statement and branch coverage: 100%
Public docstrings: 100%
Ruff, Python compile, and diff check: passed
Decision: WAIT_AND_REMEDIATE while the new-base required Checks and exact-head independent review complete; normal squash auto-merge is enabled.

Predecessor main-base evidence is discarded after retarget. No force push, protected-branch write, approval, fake status, or bypass was used.

devin-ai-integration

Devin Review found 1 new potential issue.

devin-ai-integration · 2026-08-24T07:21:04Z

pr_review_merge_scheduler.py

+    review = latest_current_head_coverage_change_request(pr)
+    if review is None:
+        return None
+    submitted_at = parse_github_datetime(review.get("submittedAt"))
+    if submitted_at is None:
+        return "current-head OpenCode coverage review has no valid submission timestamp; defer same-head re-review"
+    retry_anchor = submitted_at
+    if repo and workflow:
+        try:
+            dispatch_started_at = latest_opencode_dispatch_started_at(repo, workflow, pr)
+        except RuntimeError:
+            return "same-head OpenCode dispatch history is unavailable; defer same-head re-review"
+        if dispatch_started_at and dispatch_started_at > retry_anchor:
+            retry_anchor = dispatch_started_at
+    current_time = now or datetime.now(timezone.utc)
+    if current_time < retry_anchor + timedelta(minutes=max(0, floor_minutes)):
+        return "same-head OpenCode coverage retry floor has not elapsed"
+    return None


📝 Info: Coverage-clear redispatch can wait up to the retry floor

coverage_retry_wait_reason (pr_review_merge_scheduler.py) anchors on the coverage-only review submission time. When coverage reruns and succeeds soon after the review, redispatch still waits up to floor_minutes (default 60) past the review timestamp. This is the intended same-head retry floor and is asserted by tests; noting the operational latency for reviewers.

Was this helpful? React with 👍 or 👎 to provide feedback.

seonghobae · 2026-08-24T07:23:05Z

Superseded PR audit

#1266's exact coverage-retry change is already included in #1270's current head ccc1220cd7710d023d4f10d79ccd170a636b4516.

fix(scheduler): require independent exact-head approval #1270 current tree: 4dde1e61e5583c9e9df4e5929dc5ed979d428376
Previously verified fix(scheduler): require independent exact-head approval #1270 + fix(scheduler): retry OpenCode after coverage blockers clear #1266 stack tree: 4dde1e61e5583c9e9df4e5929dc5ed979d428376
The merge-result suite for that tree passed: 1428 passed, 1 skipped, 16 subtests, statements/branches/docstrings 100%.
No source change is discarded; the root PR now carries the complete scheduler behavior.
fix(scheduler): retry OpenCode after coverage blockers clear #1266 is being closed as superseded; fix(scheduler): require independent exact-head approval #1270 remains the single current-head review and merge target.

No force push, direct protected-branch write, approval, fake status, or bypass was used.

seonghobae · 2026-08-24T07:38:44Z

Concurrent merge audit correction

Live state re-check confirms this PR was merged normally, not closed:

Pull request: fix(scheduler): retry OpenCode after coverage blockers clear #1266
Base: codex/scheduler-independent-exact-head-approval-20260824@6cb714470c5a0c31b36cb64df8959cd020f22d52
Head: codex/pr1152-current-main-replacement-20260824@0ffad34c087a56f1ceb526148bc773b0667a2902
Merge commit: ccc1220cd7710d023d4f10d79ccd170a636b4516
Result: the merged commit is fix(scheduler): require independent exact-head approval #1270's current head and retains the previously verified stack tree 4dde1e61e5583c9e9df4e5929dc5ed979d428376.

The earlier audit comment's “close as superseded” wording is superseded by this live GitHub merge record. No force-push, protected-branch direct push, approval fabrication, or bypass was used.

* fix: retry reviews after coverage blockers clear * fix(scheduler): match coverage blocker review body * test(scheduler): cover coverage evidence gate branches * test(scheduler): reproduce coverage retry self-block * fix(scheduler): unblock coverage-only review retry * fix(scheduler): preserve central coverage retry * test(scheduler): cover status-only coverage retry * fix(scheduler): prefer newest coverage rerun * fix(scheduler): keep coverage retries fail closed * fix(scheduler): ignore superseded coverage failures * fix(scheduler): bind coverage retries to one check snapshot * fix(scheduler): avoid duplicate coverage review dispatch * fix(scheduler): rate-limit same-head coverage retries * fix(scheduler): wait for active coverage re-review * fix(scheduler): disable auto-merge during coverage retry floor * fix(scheduler): bound coverage retry dispatches

seonghobae added 7 commits August 24, 2026 01:43

fix: retry reviews after coverage blockers clear

1420287

fix(scheduler): match coverage blocker review body

dba6317

test(scheduler): cover coverage evidence gate branches

787a292

test(scheduler): reproduce coverage retry self-block

f47e0e7

fix(scheduler): unblock coverage-only review retry

04d58e7

fix(scheduler): preserve central coverage retry

5116e30

test(scheduler): cover status-only coverage retry

9f2ab9a

seonghobae mentioned this pull request Aug 23, 2026

fix(opencode): retry OpenCode after coverage blockers clear #1152

Closed