[Offline Pages] Clear remaining requests after test harness finishes.

chrome/android/javatests/src/org/chromium/chrome/browser/offlinepages/OfflinePageSavePageLaterEvaluationTest.java

Index: chrome/android/javatests/src/org/chromium/chrome/browser/offlinepages/OfflinePageSavePageLaterEvaluationTest.java
diff --git a/chrome/android/javatests/src/org/chromium/chrome/browser/offlinepages/OfflinePageSavePageLaterEvaluationTest.java b/chrome/android/javatests/src/org/chromium/chrome/browser/offlinepages/OfflinePageSavePageLaterEvaluationTest.java
index 6abc5c0d48fe5f9948d21b7b2d297411df1baf65..48b4d9d74d08390902d67d43adfb2eb9329d0ebc 100644
--- a/chrome/android/javatests/src/org/chromium/chrome/browser/offlinepages/OfflinePageSavePageLaterEvaluationTest.java
+++ b/chrome/android/javatests/src/org/chromium/chrome/browser/offlinepages/OfflinePageSavePageLaterEvaluationTest.java
@@ -82,6 +82,7 @@ public class OfflinePageSavePageLaterEvaluationTest
     private static final String RESULT_OUTPUT_FILE_PATH = "paquete/offline_eval_results.txt";
     private static final int GET_PAGES_TIMEOUT_MS = 5000;
     private static final int PAGE_MODEL_LOAD_TIMEOUT_MS = 5000;
+    private static final int REMOVE_REQUESTS_TIMEOUT_MS = 5000;

[Pete Williamson, 2016/11/07 23:45:26]

Maybe allow a longer timeout.  A svelte device might be busy, and we don't
really want to fail the test if the timeout happens (or stop trying to clean
up).  30s might be a better timeout than 5s.

[romax, 2016/11/08 00:18:35]

Done.
I've also updated the timeouts above.

 
     private OfflinePageEvaluationBridge mBridge;
     private OfflinePageEvaluationObserver mObserver;
@@ -110,11 +111,34 @@ public class OfflinePageSavePageLaterEvaluationTest
 
     @Override
     protected void tearDown() throws Exception {
-        super.tearDown();
         NotificationManager notificationManager =
                 (NotificationManager) ContextUtils.getApplicationContext().getSystemService(
                         Context.NOTIFICATION_SERVICE);
         notificationManager.cancelAll();
+        final Semaphore mClearingSemaphore = new Semaphore(0);

[Pete Williamson, 2016/11/07 23:45:26]

I see where we decrement the count and release the semaphore, where do we
increment the count?

[romax, 2016/11/08 00:18:35]

This is used as a way to block the thread until timeout or the callback has been
called. So seems there's no need to increment the count. I might be wrong here
but this is how I understand this usage...

[Pete Williamson, 2016/11/08 00:51:35]

Ah, so it is only called once, not once per request.

+        ThreadUtils.runOnUiThread(new Runnable() {
+            @Override
+            public void run() {
+                mBridge.getRequestsInQueue(new Callback<SavePageRequest[]>() {
+                    @Override
+                    public void onResult(SavePageRequest[] results) {
+                        ArrayList<Long> ids = new ArrayList<Long>(results.length);
+                        for (int i = 0; i < results.length; i++) {
+                            ids.add(results[i].getRequestId());
+                        }
+                        mBridge.removeRequestsFromQueue(ids, new Callback<Integer>() {
+                            @Override
+                            public void onResult(Integer removedCount) {
+                                mClearingSemaphore.release();
+                            }
+                        });
+                    }
+                });
+            }
+        });
+        checkTrue(mClearingSemaphore.tryAcquire(REMOVE_REQUESTS_TIMEOUT_MS, TimeUnit.MILLISECONDS),
+                "Timed out when clearing remaining requests!");
+        super.tearDown();
     }
 
     @Override