Skip race, upgrade and integration jobs in the giga-1 merge queue

[masih]

The giga-1 merge queue re-ran the whole PR suite for every queued entry: four race-detection shards plus the upgrade tests (Go Test), the state_db race run (Sei-DB), RocksDB tests, and the full Docker integration matrix. Those had all already passed on the PR itself, so in the queue they only added wall-clock time and runner minutes, and with several PRs landing back-to-back the queue kept re-running them on each rebase.

Merge-queue entries whose base is giga-1 now skip all of those. The required check names are preserved so the queue still gets a green result: the Go Test Race Detection roll-up accepts skipped from the shards and upgrade tests only for the giga-1 queue, and the Sei-DB Race Detection job is skipped at the job level with the same predicate #4178 introduced, since a skipped job satisfies a required check. RocksDB tests are already skipped for giga-1 by #4178. Docker Integration Test drops giga-1 from its merge_group filter while still running on giga-1 pushes and pull requests. The giga-1 queue is left with lint, UCI Go lint and Buf; main and release/** are unchanged.

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[github-actions]

The latest Buf updates on your PR. Results from workflow Buf / buf (pull_request).

Build	Format	Lint	Breaking	Updated (UTC)
✅ passed	✅ passed	✅ passed	✅ passed	Sep 16, 2026, 9:03 AM

[cursor]

PR Summary

Low Risk
CI workflow predicates only; no application code or test logic changes, with PR and non-giga-1 merge queues unchanged.

Overview
Speeds up the giga-1 merge queue by not re-running heavy suites that already passed on the PR.

In Go Test (go-test.yml), the four race-detection shards and Upgrade tests are skipped when merge_group.base_ref is refs/heads/giga-1. The Race Detection rollup still runs and treats both dependencies as OK when they are skipped only for that queue, so required check names stay green.

In Sei-DB (sei-db-tests.yml), the Race Detection job uses the same skip predicate (a skipped job still satisfies branch protection).

Docker Integration Test (integration-test.yml) no longer triggers on merge_group for giga-1 (push/PR to giga-1 unchanged). main and release/ merge queues are unchanged.

Coverage jobs only get comment tweaks explaining why merge-queue runs stay as no-ops via per-step ifs.

^{Reviewed by Cursor Bugbot for commit 8a2e0c5. Bugbot is set up for automated code reviews on this repo. Configure here.}

[seidroid]

CI-only change that trims the giga-1 merge queue. The go-test and sei-db skips correctly preserve the required check names, but dropping giga-1 from integration-test.yml's merge_group branch filter suppresses the workflow entirely, so its check runs never report for giga-1 queue entries — a different (and riskier) mechanism than the skip pattern the rest of the PR and #4178 use.

Findings: 1 blocking | 2 non-blocking | 1 posted inline

Blockers

• None at the file/PR level.
• 1 blocking issue(s) flagged inline on specific lines.

Non-blocking

• [suggestion] This PR now uses three different mechanisms for the same "skip in the giga-1 queue" goal: job-level if (go-test race-shard/upgrade-test), per-step if via SKIP_SLOW (sei-db test), and removing the trigger (integration-test). The sei-db per-step form was copied from coverage, whose comment claims a job-level skip would leave the "result not being present", while #4178's rationale states the opposite ("a skipped job satisfies a required status check") and this PR's own race-shard gate relies on that. Only one can be right; worth reconciling and stating it once (e.g. in a comment next to the giga-1 predicate) so the next workflow author does not have to guess between seven step-level if:s and one job-level if:.
• [suggestion] After this change the giga-1 merge queue executes no Go tests and no integration tests at all (lint, UCI Go lint and Buf only). The queue's distinct value over per-PR CI is catching semantic conflicts between PRs that each pass in isolation but not when stacked, and that signal is now gone for giga-1. If the wall-clock cost is the concern, keeping one cheap gate in the queue — e.g. a single race shard, or a non-race go build ./... && go vet ./... — would retain compile-level conflict detection for a fraction of the current time.

[masih]

@seidroid review

[seidroid]

CI-only change that trims the giga-1 merge queue; the job-level skip predicate matches the #4178 precedent, the Race Detection roll-up correctly stays on always() and only short-circuits when both gated jobs report skipped in that queue, and the previous round's blocker about the integration-test trigger was resolved by verifying giga-1's ruleset. One documentation inconsistency remains inside sei-db-tests.yml.

Findings: 0 blocking | 2 non-blocking | 1 posted inline

Blockers

• None at the file/PR level.

Non-blocking

• None at the file/PR level.
• 1 suggestion(s)/nit(s) flagged inline on specific lines.
• 1 non-blocking pre-existing issue(s) listed below under pre-existing issues.

Pre-existing issues

• [suggestion] go-test.yml and sei-db-tests.yml both declare a workflow_call: trigger, but no workflow in .github/workflows/ invokes either via uses: ./.github/workflows/.... Under workflow_call the github.event context belongs to the caller, so the new giga-1 predicates would silently inherit the caller's merge-queue base — harmless today, but the unused trigger is a trap worth dropping.

Superseded: latest AI review found no blocking issues.