Massive refactor of the Android invalidation code.

components/invalidation.gypi

Index: components/invalidation.gypi
diff --git a/components/invalidation.gypi b/components/invalidation.gypi
index 3da96511bd131a33c7f9ba928c7b99af6961edf8..4f2496d1b1dc1b34cf981ba37c56d8f84dc22d4f 100644
--- a/components/invalidation.gypi
+++ b/components/invalidation.gypi
@@ -42,6 +42,8 @@
         'invalidation/invalidation_prefs.cc',
         'invalidation/invalidation_prefs.h',
         'invalidation/invalidation_service.h',
+        'invalidation/invalidation_service_android.cc',
+        'invalidation/invalidation_service_android.h',
         'invalidation/invalidation_service_util.cc',
         'invalidation/invalidation_service_util.h',
         'invalidation/invalidation_state_tracker.cc',
@@ -101,6 +103,15 @@
               'invalidation/ticl_settings_provider.h',
             ],
           }],
+        ['OS == "android"', {
+          'dependencies': [
+            'invalidation_jni_headers',
+          ],
+          'sources': [
+            'invalidation/android/component_jni_registrar.cc',
+            'invalidation/android/component_jni_registrar.h',
+          ],
+        }],
       ],
     },
 
@@ -131,6 +142,7 @@
         'invalidation/fake_invalidation_state_tracker.h',
         'invalidation/fake_invalidator.cc',
         'invalidation/fake_invalidator.h',
+        'invalidation/invalidation_service_android_unittest.cc',
         'invalidation/invalidation_service_test_template.cc',
         'invalidation/invalidation_service_test_template.h',
         'invalidation/invalidation_test_util.cc',
@@ -153,4 +165,34 @@
       ],
     },
   ],
+  'conditions': [
+    ['OS == "android"', {
+      'targets': [
+        {
+          'target_name': 'invalidation_java',
+          'type': 'none',
+          'dependencies': [

[nyquist, 2014/09/04 09:01:56]

Doesn't this also depend on the cacheinvalidation_javalib target?

[maxbogue, 2014/09/05 16:42:45]

Done.

+            '../base/base.gyp:base',
+            '../sync/sync.gyp:sync_java',

[nyquist, 2014/09/04 09:01:56]

In the future, would this go away? I think we should aim towards that going
forward.
There will always be the class ModelType, so somewhere there needs to be some
kind of conversion, but preferably, this component should not depend on //sync.

[maxbogue, 2014/09/05 16:42:45]

I agree that it's preferable to not depend on //sync, but for now it's fine.

+          ],
+          'variables': {
+            'java_in_dir': 'invalidation/android/java',
+          },
+          'includes': [ '../build/java.gypi' ],
+        },
+        {
+          'target_name': 'invalidation_jni_headers',
+          'type': 'none',
+          'sources': [
+            'invalidation/android/java/src/org/chromium/components/invalidation/InvalidationService.java',
+          ],
+          'variables': {
+            'jni_gen_package': 'components/invalidation',
+          },
+          'includes': [ '../build/jni_generator.gypi' ],
+        },
+      ],
+     },
+    ],
+  ],
 }