ci: quarantine platform-stalled E2E suites from push lanes (#10316)

.github/workflows/e2e.yml

@@ -200,10 +200,16 @@ jobs:
           # The docker leg runs vitest directly instead of through
           # test:integration:sandbox:docker: that script would rebuild the image
           # the step above just built.
+          # external-context-mem0-write, external-context-auto-recall,
+          # context-compress-interactive and qwen-serve-channel-workers stall
+          # at CLI startup on macOS and the ecs-qwen pool runners (#10272);
+          # with cron-interactive and channel-plugin they are quarantined to
+          # the nightly isolated matrix below until the root fix lands
+          # (#10316).
           if [[ "${{ matrix.sandbox }}" == "sandbox:docker" ]]; then
-            npx cross-env QWEN_SANDBOX=docker vitest run --root ./integration-tests --exclude '**/interactive/cron-interactive.test.ts' --exclude '**/channel-plugin.test.ts' --shard='${{ matrix.shard }}'
+            npx cross-env QWEN_SANDBOX=docker vitest run --root ./integration-tests --exclude '**/interactive/cron-interactive.test.ts' --exclude '**/channel-plugin.test.ts' --exclude '**/interactive/external-context-mem0-write.test.ts' --exclude '**/interactive/external-context-auto-recall.test.ts' --exclude '**/interactive/context-compress-interactive.test.ts' --exclude '**/cli/qwen-serve-channel-workers.test.ts' --shard='${{ matrix.shard }}'
           else
-            npm run test:integration:sandbox:none -- --exclude '**/interactive/cron-interactive.test.ts' --exclude '**/channel-plugin.test.ts' --shard='${{ matrix.shard }}'
+            npm run test:integration:sandbox:none -- --exclude '**/interactive/cron-interactive.test.ts' --exclude '**/channel-plugin.test.ts' --exclude '**/interactive/external-context-mem0-write.test.ts' --exclude '**/interactive/external-context-auto-recall.test.ts' --exclude '**/interactive/context-compress-interactive.test.ts' --exclude '**/cli/qwen-serve-channel-workers.test.ts' --shard='${{ matrix.shard }}'
           fi
 
       # The sandbox build retags the same image name every run, so on a
@@ -272,7 +278,7 @@ jobs:
           OPENAI_API_KEY: '${{ secrets.OPENAI_API_KEY }}'
           OPENAI_BASE_URL: '${{ secrets.OPENAI_BASE_URL }}'
           OPENAI_MODEL: '${{ secrets.OPENAI_MODEL }}'
-        run: 'npx cross-env VERBOSE=true KEEP_OUTPUT=true QWEN_SANDBOX=false vitest run --root ./integration-tests --exclude "**/interactive/cron-interactive.test.ts" --exclude "**/channel-plugin.test.ts" --shard="${{ matrix.shard }}"'
+        run: 'npx cross-env VERBOSE=true KEEP_OUTPUT=true QWEN_SANDBOX=false vitest run --root ./integration-tests --exclude "**/interactive/cron-interactive.test.ts" --exclude "**/channel-plugin.test.ts" --exclude "**/interactive/external-context-mem0-write.test.ts" --exclude "**/interactive/external-context-auto-recall.test.ts" --exclude "**/interactive/context-compress-interactive.test.ts" --exclude "**/cli/qwen-serve-channel-workers.test.ts" --shard="${{ matrix.shard }}"'
 
   isolated-nightly:
     name: '${{ matrix.label }} (nightly)'
@@ -288,6 +294,16 @@ jobs:
             test_file: 'interactive/cron-interactive.test.ts'
           - label: 'channel-plugin E2E'
             test_file: 'channel-plugin.test.ts'
+          # Quarantined from the push lanes (#10272, #10316); see the
+          # excludes in the push jobs above.
+          - label: 'external-context-mem0-write E2E'
+            test_file: 'interactive/external-context-mem0-write.test.ts'
+          - label: 'external-context-auto-recall E2E'
+            test_file: 'interactive/external-context-auto-recall.test.ts'
+          - label: 'context-compress-interactive E2E'
+            test_file: 'interactive/context-compress-interactive.test.ts'
+          - label: 'qwen-serve-channel-workers E2E'
+            test_file: 'cli/qwen-serve-channel-workers.test.ts'
     steps:
       - name: 'Checkout'
         uses: 'actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10' # v6.0.3

scripts/tests/e2e-workflow.test.js

@@ -32,4 +32,38 @@ describe('e2e workflow', () => {
     expect(group).toContain('github.event_name');
     expect(group).toContain('github.head_ref || github.ref_name');
   });
+
+  it('quarantines the nightly-isolated suites from every push lane', () => {
+    // cron-interactive is timing-flaky and external-context-mem0-write,
+    // external-context-auto-recall, context-compress-interactive and
+    // qwen-serve-channel-workers stall at CLI startup on macOS and the
+    // ecs-qwen pool runners (#10272, #10316), so they are quarantined to the
+    // nightly isolated matrix while the push lanes keep real signal. Every
+    // nightly canary must be excluded from every push lane, and every
+    // push-lane exclusion must keep its nightly canary — dropping either
+    // side silently loses coverage.
+    const nightlyFiles = yml.jobs['isolated-nightly'].strategy.matrix.include
+      .map((entry) => entry.test_file)
+      .sort();
+    const runStep = (job) =>
+      yml.jobs[job].steps.find((step) => step.name === 'Run E2E tests').run;
+    const countExcludes = (run) => {
+      const counts = new Map();
+      for (const match of run.matchAll(/--exclude ['"]([^'"]+)['"]/g)) {
+        const file = match[1].replace(/^\*\*\//, '');
+        counts.set(file, (counts.get(file) ?? 0) + 1);
+      }
+      return counts;
+    };
+    // The Linux job runs vitest twice (sandbox:docker and sandbox:none legs),
+    // so each quarantine exclude appears once per leg there.
+    const linux = countExcludes(runStep('e2e-test-linux'));
+    const macos = countExcludes(runStep('e2e-test-macos'));
+    expect([...linux.keys()].sort()).toEqual(nightlyFiles);
+    expect([...macos.keys()].sort()).toEqual(nightlyFiles);
+    for (const file of nightlyFiles) {
+      expect(linux.get(file)).toBe(2);
+      expect(macos.get(file)).toBe(1);
+    }
+  });
 });