ci: faster, stricter CI (instrumented gate, parallel jobs, test forks) (#1348)

.github/workflows/android.yml

@@ -97,11 +97,8 @@ jobs:
       - name: Add local bin to PATH
         run: echo "$HOME/.local/bin" >> $GITHUB_PATH
 
-      - name: Run ktlintCheck
-        run: ./gradlew ktlintCheck
-
-      - name: Run checkColorLiterals (hardcoded Color guard, issue #622)
-        run: ./gradlew checkColorLiterals
+      - name: Run ktlintCheck + checkColorLiterals (hardcoded Color guard, issue #622)
+        run: ./gradlew ktlintCheck checkColorLiterals --continue
 
       - name: Upload ktlint reports
         if: always()
@@ -185,6 +182,25 @@ jobs:
       - name: Run test
         run: ./gradlew test
 
+      # Telemetry only, never a gate: surfaces the slowest classes so perf work
+      # targets real offenders.
+      - name: Slowest unit-test classes
+        if: always()
+        run: |
+          python3 - <<'EOF' >> "$GITHUB_STEP_SUMMARY"
+          import glob, xml.etree.ElementTree as ET
+          rows = []
+          for path in glob.glob("app/build/test-results/testDebugUnitTest/*.xml"):
+              root = ET.parse(path).getroot()
+              rows.append((float(root.get("time", 0)), int(root.get("tests", 0)), root.get("name", path)))
+          rows.sort(reverse=True)
+          print("### Slowest unit-test classes\n")
+          print(f"{len(rows)} classes, {sum(r[0] for r in rows):.0f}s summed across forks\n")
+          print("| Time | Tests | Class |\n|---:|---:|---|")
+          for secs, count, name in rows[:20]:
+              print(f"| {secs:.1f}s | {count} | `{name.rsplit('.', 1)[-1]}` |")
+          EOF
+
       - name: Upload test reports
         if: always()
         uses: actions/upload-artifact@v7
@@ -194,11 +210,11 @@ jobs:
           retention-days: 7
 
   # ─────────────────────────────────────────────────────────────────────
-  # Job 4: Build Debug APK — depends on all three above passing.
+  # Job 4: Build Debug APK — runs in parallel; CI Summary gates the result.
   # ─────────────────────────────────────────────────────────────────────
   build:
     name: Build Debug APK
-    needs: [changes, ktlint, android-lint, unit-tests]
+    needs: [changes]
     if: needs.changes.outputs.code == 'true'
     runs-on: ubuntu-latest
     timeout-minutes: 10
@@ -241,7 +257,7 @@ jobs:
   # ─────────────────────────────────────────────────────────────────────
   release-compile:
     name: Release Compile Check
-    needs: [changes, ktlint, android-lint, unit-tests]
+    needs: [changes]
     if: needs.changes.outputs.code == 'true'
     runs-on: ubuntu-latest
     timeout-minutes: 10
@@ -339,7 +355,7 @@ jobs:
   # ─────────────────────────────────────────────────────────────────────
   ci-summary:
     name: CI Summary
-    needs: [changes, ktlint, android-lint, unit-tests, build, release-compile]
+    needs: [changes, ktlint, android-lint, unit-tests, build, release-compile, instrumented-tests]
     runs-on: ubuntu-latest
     timeout-minutes: 2
     if: always()
@@ -355,11 +371,12 @@ jobs:
           TEST_RESULT: ${{ needs.unit-tests.result }}
           BUILD_RESULT: ${{ needs.build.result }}
           RELEASE_COMPILE_RESULT: ${{ needs.release-compile.result }}
+          INSTRUMENTED_RESULT: ${{ needs.instrumented-tests.result }}
         run: |
           LABEL="ci-red"
           FAILED=0
           FAILED_JOBS=""
-          for entry in "ktlint:$KTLINT_RESULT" "android-lint:$LINT_RESULT" "unit-tests:$TEST_RESULT" "build:$BUILD_RESULT" "release-compile:$RELEASE_COMPILE_RESULT"; do
+          for entry in "ktlint:$KTLINT_RESULT" "android-lint:$LINT_RESULT" "unit-tests:$TEST_RESULT" "build:$BUILD_RESULT" "release-compile:$RELEASE_COMPILE_RESULT" "instrumented-tests:$INSTRUMENTED_RESULT"; do
             name="${entry%%:*}"
             result="${entry#*:}"
             echo "::notice::$name = $result"

app/build.gradle.kts

@@ -309,6 +309,13 @@ tasks.named("check") {
     dependsOn("checkColorLiterals")
 }
 
-tasks.withType<Test> {
+val includeBenchmarks = providers.gradleProperty("includeBenchmarks").isPresent
+
+tasks.withType<Test>().configureEach {
     useJUnitPlatform()
+    maxParallelForks = 2
+    // Assertion-free timing probes are opt-in: ./gradlew testDebugUnitTest -PincludeBenchmarks
+    if (!includeBenchmarks) {
+        filter { excludeTestsMatching("*Benchmark") }
+    }
 }

app/src/main/java/com/m57/hermescontrol/data/ws/HermesWsClient.kt

@@ -415,6 +415,7 @@ object HermesWsClient {
 
     @VisibleForTesting
     internal fun probeLivenessOnTransportChange(
+        timeoutMs: Long = LIVENESS_PROBE_TIMEOUT_MS,
         onFailureAction: () -> Unit = {
             synchronized(outboundLock) {
                 if (connected.get()) webSocket?.cancel()
@@ -426,7 +427,7 @@ object HermesWsClient {
         transportProbeJob =
             wsScope.launch {
                 Log.d(TAG, "Network transport changed — probing WebSocket liveness")
-                val alive = runCatching { ping(LIVENESS_PROBE_TIMEOUT_MS) }.isSuccess
+                val alive = runCatching { ping(timeoutMs) }.isSuccess
                 if (alive) {
                     Log.d(TAG, "WebSocket liveness probe succeeded on new transport")
                     return@launch

app/src/main/java/com/m57/hermescontrol/notification/NotificationReplyReceiver.kt

@@ -25,6 +25,9 @@ open class NotificationReplyReceiver : BroadcastReceiver() {
         const val EXTRA_SESSION_ID = "extra_session_id"
         private const val REPLY_TIMEOUT_MS = 5_000L
 
+        /** Overall deadline for persisting and submitting one notification reply. */
+        private const val PERSISTENCE_TIMEOUT_MS = 5_000L
+
         /**
          * Turn-boundary read budget inside the reply deadline. Deliberately tight:
          * the probe sits on the reply path, so a slow or unreachable gateway must
@@ -44,6 +47,9 @@ open class NotificationReplyReceiver : BroadcastReceiver() {
      */
     internal open fun goAsyncCompat(): BroadcastReceiver.PendingResult = goAsync()
 
+    /** Overridable so tests can prove the timeout path without waiting 5s. */
+    internal open val persistenceTimeoutMs: Long get() = PERSISTENCE_TIMEOUT_MS
+
     /**
      * Test-friendly wrapper for notification creation. Override in tests to
      * avoid [NotificationCompat.Builder.build()] calling Android framework
@@ -75,7 +81,7 @@ open class NotificationReplyReceiver : BroadcastReceiver() {
             val pendingResult = goAsyncCompat()
             replyScope.launch {
                 try {
-                    withTimeout(5000L) {
+                    withTimeout(persistenceTimeoutMs) {
                         withContext(Dispatchers.IO) {
                             val db =
                                 com.m57.hermescontrol.data.local.HermesDatabase

app/src/test/java/com/m57/hermescontrol/data/ws/HermesWsClientTest.kt

@@ -2188,12 +2188,13 @@ class HermesWsClientTest {
 
         val failureLatch = CountDownLatch(1)
         HermesWsClient.probeLivenessOnTransportChange(
+            timeoutMs = 50L,
             onFailureAction = {
                 failureLatch.countDown()
             },
         )
 
-        assertTrue("Expected failure action to be invoked on failed ping", failureLatch.await(6, TimeUnit.SECONDS))
+        assertTrue("Expected failure action to be invoked on failed ping", failureLatch.await(3, TimeUnit.SECONDS))
     }
 
     @Test

app/src/test/java/com/m57/hermescontrol/notification/NotificationReplyReceiverTest.kt

@@ -412,16 +412,23 @@ class NotificationReplyReceiverTest {
             kotlinx.coroutines.delay(10000)
         }
 
-        receiver.onReceive(mockContext, mockIntent)
+        // Same timeout path as production, with a short deadline instead of 5s.
+        val shortTimeoutReceiver =
+            object : NotificationReplyReceiver() {
+                override fun goAsyncCompat(): BroadcastReceiver.PendingResult = mockPendingResult
 
-        // Wait longer than the 5-second timeout, but less than the 10-second delay
-        Thread.sleep(6000)
+                override fun buildReplyNotification(context: Context): Notification = mockNotification
+
+                override val persistenceTimeoutMs: Long = 50L
+            }
+
+        shortTimeoutReceiver.onReceive(mockContext, mockIntent)
 
         // Timeout should be caught in catch block and logged
-        verify { android.util.Log.e("NotificationReply", any<String>(), any()) }
+        verify(timeout = 3_000) { android.util.Log.e("NotificationReply", any<String>(), any()) }
 
         // Pending result must still finish after timeout
-        verify { mockPendingResult.finish() }
+        verify(timeout = 3_000) { mockPendingResult.finish() }
     }
 
     // ── Helpers ────────────────────────────────────────────────────────────