ci: fold serialized suites into existing server shards (BLO-22428)

.github/workflows/pr.yml

@@ -130,6 +130,9 @@ jobs:
       - name: Test verify job cancelled/skipped-vs-failed lane outcome
         run: node --test ./scripts/__tests__/pr-verify-lane-outcome.test.mjs
 
+      - name: Test PR server-shard folding
+        run: node --test ./scripts/__tests__/pr-ci-shard-folding.test.mjs
+
       - name: Validate release package manifest
         run: node ./scripts/release-package-map.mjs check
 
@@ -456,13 +459,17 @@ jobs:
           fi
           pnpm test:run:general -- "${args[@]}"
 
+      # Reuse the checkout, dependency install, cache restore, and embedded
+      # Postgres setup already paid for by each server shard. This removes four
+      # ARC pod cold starts without reducing serialized-suite coverage.
+      - name: Run serialized server test shard
+        if: matrix.group == 'general-server'
+        run: pnpm test:run:serialized -- --shard-index ${{ matrix.shard_index }} --shard-count ${{ matrix.shard_count }}
+
   verify:
     # Preserve the legacy required-check name while the underlying work runs in parallel.
-    # BLO-20869: verify_serialized_server is folded in here (rather than made
-    # required on its own) so a cancelled/skipped/never-scheduled serialized
-    # shard fails this required check instead of being invisible to the merge
-    # gate. See the issue for the measured latency tradeoff of putting a
-    # ~90-minute job on the required path this way.
+    # Serialized server suites run inside the four general-server matrix jobs,
+    # so general_tests remains the required aggregate for both test families.
     name: verify
     if: ${{ always() }}
     needs:
@@ -473,7 +480,6 @@ jobs:
         worktree_install,
         opencode_responses_replay,
         build,
-        verify_serialized_server,
       ]
     runs-on: arc-light
     timeout-minutes: 5
@@ -507,7 +513,6 @@ jobs:
           WORKTREE_INSTALL_RESULT: ${{ needs.worktree_install.result }}
           OPENCODE_RESPONSES_REPLAY_RESULT: ${{ needs.opencode_responses_replay.result }}
           BUILD_RESULT: ${{ needs.build.result }}
-          VERIFY_SERIALIZED_SERVER_RESULT: ${{ needs.verify_serialized_server.result }}
         run: |
           lane_names=(
             helm_chart
@@ -516,7 +521,6 @@ jobs:
             worktree_install
             opencode_responses_replay
             build
-            verify_serialized_server
           )
           lane_results=(
             "$HELM_CHART_RESULT"
@@ -525,7 +529,6 @@ jobs:
             "$WORKTREE_INSTALL_RESULT"
             "$OPENCODE_RESPONSES_REPLAY_RESULT"
             "$BUILD_RESULT"
-            "$VERIFY_SERIALIZED_SERVER_RESULT"
           )
 
           cancelled_lanes=()
@@ -602,63 +605,6 @@ jobs:
       - name: Build
         run: pnpm build
 
-  verify_serialized_server:
-    name: Verify serialized server suites (${{ matrix.shard_label }})
-    needs: [policy, general_tests]
-    # BLO-22428: route merge_group traffic to the dedicated arc-merge-queue
-    # pool so pull_request and merge_group traffic cannot starve each other.
-    # See typecheck_release_registry above.
-    runs-on: ${{ github.event_name == 'merge_group' && 'arc-merge-queue' || 'arc-paperclip-general' }}
-    timeout-minutes: 90
-    strategy:
-      fail-fast: false
-      max-parallel: 4
-      matrix:
-        include:
-          - shard_index: 0
-            shard_count: 4
-            shard_label: 1/4
-          - shard_index: 1
-            shard_count: 4
-            shard_label: 2/4
-          - shard_index: 2
-            shard_count: 4
-            shard_label: 3/4
-          - shard_index: 3
-            shard_count: 4
-            shard_label: 4/4
-
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v6
-
-      - name: Setup pnpm
-        uses: pnpm/action-setup@v6
-        with:
-          version: 9.15.4
-
-      - name: Restore regenerated PR lockfile (if policy uploaded one)
-        uses: actions/download-artifact@v4
-        continue-on-error: true
-        with:
-          name: pr-lockfile
-          path: .
-
-      - name: Setup Node.js
-        uses: actions/setup-node@v6
-        with:
-          node-version: 24
-          cache: pnpm
-
-      - name: Set up GHA-backed sccache
-        uses: mozilla-actions/sccache-action@v0.0.9
-
-      - name: Install dependencies
-        run: pnpm install --frozen-lockfile
-
-      - name: Run serialized server test shard
-        run: pnpm test:run:serialized -- --shard-index ${{ matrix.shard_index }} --shard-count ${{ matrix.shard_count }}
-
   canary_dry_run:
     name: Canary Dry Run
     needs: [policy]

scripts/__tests__/pr-ci-shard-folding.test.mjs

@@ -0,0 +1,54 @@
+import assert from "node:assert/strict";
+import { readFileSync } from "node:fs";
+import { test } from "node:test";
+
+const workflow = readFileSync(new URL("../../.github/workflows/pr.yml", import.meta.url), "utf8");
+
+function jobBlock(name, nextName) {
+  const start = workflow.indexOf(`\n  ${name}:\n`);
+  assert.notEqual(start, -1, `pr.yml must define ${name}`);
+  const end = workflow.indexOf(`\n  ${nextName}:\n`, start + 1);
+  assert.notEqual(end, -1, `pr.yml must define ${nextName} after ${name}`);
+  return workflow.slice(start, end);
+}
+
+test("server shards run general and serialized suites in the same four jobs", () => {
+  const general = jobBlock("general_tests", "verify");
+  const serverEntries = general.match(
+    /          - group: general-server\n(?:            [^\n]*\n)*/g,
+  ) ?? [];
+  assert.equal(serverEntries.length, 4, "general_tests must retain four isolated server shards");
+  for (const [index, entry] of serverEntries.entries()) {
+    assert.match(
+      entry,
+      new RegExp(`group_label: server ${index + 1}/4`),
+      `server shard ${index} must retain its matching label`,
+    );
+    assert.match(entry, new RegExp(`shard_index: ${index}`));
+    assert.match(entry, /shard_count: 4/);
+  }
+  assert.match(general, /pnpm test:run:general -- "\$\{args\[@\]\}"/);
+  assert.match(
+    general,
+    /- name: Run serialized server test shard\n        if: matrix\.group == 'general-server'\n        run: pnpm test:run:serialized -- --shard-index \$\{\{ matrix\.shard_index \}\} --shard-count \$\{\{ matrix\.shard_count \}\}/,
+    "serialized suites must run only in the corresponding server shard",
+  );
+  assert.ok(
+    general.indexOf("pnpm test:run:general") < general.indexOf("pnpm test:run:serialized"),
+    "each server shard must run general suites before serialized suites",
+  );
+});
+
+test("serialized coverage is aggregated through general_tests without a second job matrix", () => {
+  assert.doesNotMatch(workflow, /\n  verify_serialized_server:\n/);
+  assert.equal(
+    workflow.match(/pnpm test:run:serialized -- --shard-index/g)?.length,
+    1,
+    "the PR workflow should declare one matrix-driven serialized command",
+  );
+
+  const verify = jobBlock("verify", "build");
+  assert.match(verify, /\n        general_tests,/);
+  assert.doesNotMatch(verify, /verify_serialized_server/);
+  assert.match(verify, /GENERAL_TESTS_RESULT: \$\{\{ needs\.general_tests\.result \}\}/);
+});

scripts/__tests__/pr-verify-lane-outcome.test.mjs

@@ -55,7 +55,6 @@ function runVerifyStep(results) {
     WORKTREE_INSTALL_RESULT: results.worktree_install ?? "success",
     OPENCODE_RESPONSES_REPLAY_RESULT: results.opencode_responses_replay ?? "success",
     BUILD_RESULT: results.build ?? "success",
-    VERIFY_SERIALIZED_SERVER_RESULT: results.verify_serialized_server ?? "success",
   };
   return spawnSync("bash", ["-c", script], { env, encoding: "utf8" });
 }
@@ -122,31 +121,6 @@ test("verify step annotates both a real failure and a cancellation when a run ha
   assert.match(result.stdout, /general_tests/);
 });
 
-// BLO-20869: verify_serialized_server must be treated exactly like the other
-// required lanes -- cancelled, skipped, or never-scheduled (which also reads
-// as "skipped" via `needs`) must all fail this required check instead of
-// being invisible to the merge gate.
-test("verify step fails when the serialized server suite is cancelled", () => {
-  const result = runVerifyStep({ verify_serialized_server: "cancelled" });
-  assert.notEqual(result.status, 0);
-  assert.match(result.stdout, /::error title=verify: lane cancelled::/);
-  assert.match(result.stdout, /verify_serialized_server/);
-});
-
-test("verify step fails when the serialized server suite is skipped", () => {
-  const result = runVerifyStep({ verify_serialized_server: "skipped" });
-  assert.notEqual(result.status, 0);
-  assert.match(result.stdout, /::error title=verify: lane skipped::/);
-  assert.match(result.stdout, /verify_serialized_server/);
-});
-
-test("verify step fails when the serialized server suite genuinely fails", () => {
-  const result = runVerifyStep({ verify_serialized_server: "failure" });
-  assert.notEqual(result.status, 0);
-  assert.match(result.stdout, /::error title=verify: lane failure::/);
-  assert.match(result.stdout, /verify_serialized_server/);
-});
-
 test("verify step annotates a skipped lane as an unmet dependency, not a failure", () => {
   const result = runVerifyStep({ worktree_install: "skipped" });
   assert.notEqual(result.status, 0);