store: Retry polling event queue on error

The new TODO comments are all things that we don't do today in
zulip-mobile, and haven't been aware of causing problems for people.
So there's no rush to do them here.

Fixes: #184

Author: gnprice

lib/model/store.dart

@@ -14,6 +14,7 @@ import '../api/model/initial_snapshot.dart';
 import '../api/model/model.dart';
 import '../api/route/events.dart';
 import '../api/route/messages.dart';
+import '../api/backoff.dart';
 import '../log.dart';
 import '../notifications.dart';
 import 'autocomplete.dart';
@@ -573,6 +574,8 @@ class UpdateMachine {
   }
 
   void poll() async {
+    final backoffMachine = BackoffMachine();
+
     while (true) {
       if (_debugLoopSignal != null) {
         await _debugLoopSignal!.future;
@@ -595,10 +598,22 @@ class UpdateMachine {
             debugLog('… Event queue replaced.');
             return;
 
+          case Server5xxException() || NetworkException():
+            assert(debugLog('Transient error polling event queue for $store: $e\n'
+                'Backing off, then will retry…'));
+            // TODO tell user if transient polling errors persist
+            // TODO reset to short backoff eventually
+            await backoffMachine.wait();
+            assert(debugLog('… Backoff wait complete, retrying poll.'));
+            continue;
+
           default:
-            assert(debugLog('Error polling event queue for $store: $e'));
-            // TODO(#184) handle errors on get-events; retry with backoff
-            rethrow;
+            assert(debugLog('Error polling event queue for $store: $e\n'
+                'Backing off and retrying even though may be hopeless…'));
+            // TODO tell user on non-transient error in polling
+            await backoffMachine.wait();
+            assert(debugLog('… Backoff wait complete, retrying poll.'));
+            continue;
         }
       }
 

test/model/store_test.dart

@@ -249,6 +249,53 @@ void main() {
       await Future.delayed(Duration.zero);
       check(store.userSettings!.twentyFourHourTime).isTrue();
     }));
+
+    void checkRetry(void Function() prepareError) {
+      awaitFakeAsync((async) async {
+        await prepareStore(lastEventId: 1);
+        updateMachine.debugPauseLoop();
+        updateMachine.poll();
+        check(async.pendingTimers).length.equals(0);
+
+        // Make the request, inducing an error in it.
+        prepareError();
+        updateMachine.debugAdvanceLoop();
+        async.flushMicrotasks();
+        checkLastRequest(lastEventId: 1);
+
+        // Polling doesn't resume immediately; there's a timer.
+        check(async.pendingTimers).length.equals(1);
+        updateMachine.debugAdvanceLoop();
+        async.flushMicrotasks();
+        check(connection.lastRequest).isNull();
+        check(async.pendingTimers).length.equals(1);
+
+        // Polling continues after a timer.
+        connection.prepare(json: GetEventsResult(events: [
+          HeartbeatEvent(id: 2),
+        ], queueId: null).toJson());
+        async.flushTimers();
+        checkLastRequest(lastEventId: 1);
+        check(updateMachine.lastEventId).equals(2);
+      });
+    }
+
+    test('retries on Server5xxException', () {
+      checkRetry(() => connection.prepare(httpStatus: 500, body: 'splat'));
+    });
+
+    test('retries on NetworkException', () {
+      checkRetry(() => connection.prepare(exception: Exception("failed")));
+    });
+
+    test('retries on ZulipApiException', () {
+      checkRetry(() => connection.prepare(httpStatus: 400, json: {
+        'result': 'error', 'code': 'BAD_REQUEST', 'msg': 'Bad request'}));
+    });
+
+    test('retries on MalformedServerResponseException', () {
+      checkRetry(() => connection.prepare(httpStatus: 200, body: 'nonsense'));
+    });
   });
 
   group('UpdateMachine.registerNotificationToken', () {