[BackgroundSync] Trigger Background Sync events when Chrome is backgrounded on Android

content/browser/android/background_sync_network_observer_android.cc

Index: content/browser/android/background_sync_network_observer_android.cc
diff --git a/content/browser/android/background_sync_network_observer_android.cc b/content/browser/android/background_sync_network_observer_android.cc
new file mode 100644
index 0000000000000000000000000000000000000000..48e92573d34c14a9ed7291ae2083bf7fd82b85c5
--- /dev/null
+++ b/content/browser/android/background_sync_network_observer_android.cc
@@ -0,0 +1,48 @@
+// Copyright 2015 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "content/browser/android/background_sync_network_observer_android.h"
+
+#include "content/public/browser/browser_thread.h"
+#include "jni/BackgroundSyncNetworkObserver_jni.h"
+
+namespace content {
+
+// static
+bool BackgroundSyncNetworkObserverAndroid::RegisterNetworkObserver(
+    JNIEnv* env) {
+  return RegisterNativesImpl(env);
+}
+
+BackgroundSyncNetworkObserverAndroid::BackgroundSyncNetworkObserverAndroid(
+    const base::Closure& network_changed_callback)
+    : BackgroundSyncNetworkObserver(network_changed_callback) {
+  DCHECK_CURRENTLY_ON(BrowserThread::IO);

[jkarlin, 2015/09/11 00:05:23]

I'm pretty tired so I may be wrong but I believe this class needs to live on the
UI thread to talk to Java code.

[iclelland, 2015/09/11 14:42:15]

Well... JNI is usable from any thread, and this class is created on the IO
thread (it's parent class constructor has a similar DCHECK) so we need to stay
on IO at least long enough to call RemoveNetworkChangeObserver.

But perhaps you're right in that the Java-side objects should be created on the
UI thread. The NetworkChangeNotifierAutodetect.Observer notes that notifications
always happen on the main thread.

[jkarlin, 2015/09/11 15:16:06]

Good point. This class needs to be on the IO thread. 
I'd like the java side of this to all be on one thread. Ideally every class
lives on one thread. So perhaps there should be a new class that does the
communication with Java and it lives on the IO thread.

[jkarlin, 2015/09/11 20:58:55]

Sorry, I meant the class that talks to Java should be on the UI thread of
course.

+
+  // Remove the observer attached by the NetworkObserver constructor
+  net::NetworkChangeNotifier::RemoveNetworkChangeObserver(this);
+
+  // Attach a Java BackgroundSyncNetworkObserver object. Its lifetime will be
+  // scoped to the lifetime of this object.
+  JNIEnv* env = base::android::AttachCurrentThread();
+  base::android::ScopedJavaLocalRef<jobject> obj =
+      Java_BackgroundSyncNetworkObserver_createObserver(
+          env, base::android::GetApplicationContext(),
+          reinterpret_cast<jlong>(this));
+  observer_ = obj;
+}
+
+BackgroundSyncNetworkObserverAndroid::~BackgroundSyncNetworkObserverAndroid() {
+}
+
+void BackgroundSyncNetworkObserverAndroid::NotifyConnectionTypeChanged(
+    JNIEnv* env,
+    jobject jcaller,
+    jint new_connection_type) {
+  DCHECK_CURRENTLY_ON(BrowserThread::IO);

[jkarlin, 2015/09/11 00:05:23]

I believe this will be called on UI?

Probably explains the purple android bot.

[iclelland, 2015/09/11 14:42:15]

Apparently you weren't the only tired one. I fixed this on the *other* CL. I'll
backport that here.

+  OnNetworkChanged(static_cast<net::NetworkChangeNotifier::ConnectionType>(
+      new_connection_type));
+}
+
+}  // namespace content