fix(queue): check push-required contexts explicitly instead of combined state

infra-sre commented

2026-05-14 09:38:11 +00:00

Member

Summary

The queue-bot was checking the combined commit state of main to decide whether to merge.
Combined state can be "failure" due to non-blocking jobs (continue-on-error: true) that
don't gate merges — e.g. Platform Go on main push fails during mc#774 runner exhaustion
but that does not block PR merges.

The real merge gate is CI / all-required (push), which correctly aggregates all
blocking failures. Switching to explicit context checks fixes three bugs:

False pause on non-blocking failures: Combined state "failure" from
continue-on-error: true jobs (Platform Go, Handlers Postgres) was incorrectly
blocking the queue. Now checks CI / all-required (push) directly.
Stale status in truncated array: latest_statuses_by_context() kept the FIRST
(oldest) occurrence of each context. Gitea's /status endpoint returns 30-entry
pages in ascending id order, so required-context entries were often missed.
Fixed by iterating in reverse so the LAST (newest) occurrence wins.
30-entry cap on statuses: The /status endpoint caps statuses[] at 30.
Fixed by also fetching /statuses?limit=200 to get the full list.

Test plan

Dry-run confirms main is green (checks CI / all-required (push)=success)
Dry-run processes PR #942 (skips: base=staging, not main)
CI passes on PR
Queue-bot merges PR #978 after this lands

References

feedback_queue_combined_state_false_pause
feedback_gitea_statuses_truncated_30

SOP Checklist

Comprehensive testing performed

N/A — CI infrastructure change; no runtime code
/sop-ack comprehensive-testing

Local-postgres E2E run

N/A — CI infrastructure change; no database surface
/sop-ack local-postgres-e2e

Staging-smoke verified or pending

N/A — queue-bot context fix; no runtime impact
/sop-ack staging-smoke

Root-cause not symptom

/sop-ack root-cause — queue script was checking wrong contexts for push-required vs pull_request triggers

No backwards-compat

/sop-ack no-backwards-compat — no behavioral change to production code

QA review N/A declaration

/sop-n/a qa-review — CI infrastructure change — queue-bot context fix; no runtime code, no qa-testable behavior.

Security review N/A declaration

/sop-n/a security-review — CI infrastructure change — queue-bot context fix; no security surface.

No multi-region

/sop-ack no-multi-region — CI infrastructure change; no regional impact.

No migration

/sop-ack no-migration — no schema or data migration.

No new deps

/sop-ack no-new-deps — no new dependencies.

No perf risk

/sop-ack no-perf-risk — CI infrastructure only; no runtime performance impact.

## Summary The queue-bot was checking the combined commit state of main to decide whether to merge. Combined state can be "failure" due to non-blocking jobs (`continue-on-error: true`) that don't gate merges — e.g. Platform Go on main push fails during mc#774 runner exhaustion but that does not block PR merges. The real merge gate is `CI / all-required (push)`, which correctly aggregates all blocking failures. Switching to explicit context checks fixes three bugs: 1. **False pause on non-blocking failures**: Combined state "failure" from `continue-on-error: true` jobs (Platform Go, Handlers Postgres) was incorrectly blocking the queue. Now checks `CI / all-required (push)` directly. 2. **Stale status in truncated array**: `latest_statuses_by_context()` kept the FIRST (oldest) occurrence of each context. Gitea's `/status` endpoint returns 30-entry pages in ascending id order, so required-context entries were often missed. Fixed by iterating in reverse so the LAST (newest) occurrence wins. 3. **30-entry cap on statuses**: The `/status` endpoint caps `statuses[]` at 30. Fixed by also fetching `/statuses?limit=200` to get the full list. ## Test plan - [x] Dry-run confirms main is green (checks `CI / all-required (push)=success`) - [x] Dry-run processes PR #942 (skips: base=staging, not main) - [ ] CI passes on PR - [ ] Queue-bot merges PR #978 after this lands ## References - `feedback_queue_combined_state_false_pause` - `feedback_gitea_statuses_truncated_30` ## SOP Checklist ### Comprehensive testing performed - N/A — CI infrastructure change; no runtime code - [x] /sop-ack comprehensive-testing ### Local-postgres E2E run - N/A — CI infrastructure change; no database surface - [x] /sop-ack local-postgres-e2e ### Staging-smoke verified or pending - N/A — queue-bot context fix; no runtime impact - [x] /sop-ack staging-smoke ### Root-cause not symptom - /sop-ack root-cause — queue script was checking wrong contexts for push-required vs pull_request triggers ### No backwards-compat - /sop-ack no-backwards-compat — no behavioral change to production code ### QA review N/A declaration - /sop-n/a qa-review — CI infrastructure change — queue-bot context fix; no runtime code, no qa-testable behavior. ### Security review N/A declaration - /sop-n/a security-review — CI infrastructure change — queue-bot context fix; no security surface. ### No multi-region - [x] /sop-ack no-multi-region — CI infrastructure change; no regional impact. ### No migration - [x] /sop-ack no-migration — no schema or data migration. ### No new deps - [x] /sop-ack no-new-deps — no new dependencies. ### No perf risk - [x] /sop-ack no-perf-risk — CI infrastructure only; no runtime performance impact.