appdistribution: fix flaky test due to race condtion in testing logic: checkForNewRelease_whenCalledMultipleTimes_onlyFetchesReleasesOnce (#7241)

Author: dconeybe

firebase-appdistribution/src/main/java/com/google/firebase/appdistribution/impl/TaskCache.java

@@ -14,6 +14,7 @@
 
 package com.google.firebase.appdistribution.impl;
 
+import androidx.annotation.VisibleForTesting;
 import com.google.android.gms.tasks.Task;
 import com.google.android.gms.tasks.TaskCompletionSource;
 import com.google.firebase.annotations.concurrent.Lightweight;
@@ -66,6 +67,20 @@ Task<T> getOrCreateTask(TaskProducer<T> producer) {
     return taskCompletionSource.getTask();
   }
 
+  /**
+   * Returns a task that completes when this object's internal sequential executor has finished
+   * processing all enqueued operations.
+   * <p>
+   * This method should never be called in production code; however, it is useful in unit tests to
+   * ensure deterministic behavior.
+   */
+  @VisibleForTesting
+  Task<Void> newTaskForSequentialExecutorFlush() {
+    TaskCompletionSource<Void> taskCompletionSource = new TaskCompletionSource<>();
+    sequentialExecutor.execute(() -> taskCompletionSource.setResult(null));
+    return taskCompletionSource.getTask();
+  }
+
   private static <T> boolean isOngoing(Task<T> task) {
     return task != null && !task.isComplete();
   }

firebase-appdistribution/src/test/java/com/google/firebase/appdistribution/impl/NewReleaseFetcherTest.java

@@ -112,6 +112,7 @@ public void checkForNewRelease_whenCalledMultipleTimes_onlyFetchesReleasesOnce()
 
     // Don't set the result until after calling twice, to make sure that the task from the first
     // call is still ongoing.
+    awaitTask(newReleaseFetcher.cachedCheckForNewRelease.newTaskForSequentialExecutorFlush());
     task.setResult(null);
     awaitTask(checkForNewReleaseTask1);
     awaitTask(checkForNewReleaseTask2);