sync: Add pre-commit update avoidance mode + flag

sync/sessions/sync_session_context.h

Index: sync/sessions/sync_session_context.h
diff --git a/sync/sessions/sync_session_context.h b/sync/sessions/sync_session_context.h
index b694162c15637e046d1997f51d0d8bcde6d42cb2..bdac0a6f6e6035ee9525d60587e6871d0418f569 100644
--- a/sync/sessions/sync_session_context.h
+++ b/sync/sessions/sync_session_context.h
@@ -55,6 +55,7 @@ class SYNC_EXPORT_PRIVATE SyncSessionContext {
                      DebugInfoGetter* debug_info_getter,
                      TrafficRecorder* traffic_recorder,
                      bool keystore_encryption_enabled,
+                     bool always_fetch_updates_before_commit,
                      const std::string& invalidator_client_id);
 
   ~SyncSessionContext();
@@ -129,6 +130,15 @@ class SYNC_EXPORT_PRIVATE SyncSessionContext {
     return invalidator_client_id_;
   }
 
+  const bool should_fetch_updates_before_commit() const {
+    return !(pre_commit_update_avoidance_experiment_enabled_ ||
+             force_enable_pre_commit_update_avoidance_);
+  }
+
+  void set_pre_commit_update_avoidance_experiment_enabled(bool value) {
+    pre_commit_update_avoidance_experiment_enabled_ = value;
+  }
+
  private:
   // Rather than force clients to set and null-out various context members, we
   // extend our encapsulation boundary to scoped helpers that take care of this
@@ -181,6 +191,15 @@ class SYNC_EXPORT_PRIVATE SyncSessionContext {
   // prevent us from receiving notifications of changes we make ourselves.
   const std::string invalidator_client_id_;
 
+  // Flag to enable or disable the no pre-commit GetUpdates experiment.  When
+  // this flag is set to false, the syncer has the option of not performing at
+  // GetUpdates request when there is nothing to fetch.
+  bool pre_commit_update_avoidance_experiment_enabled_;

[Nicolas Zea, 2013/07/16 00:24:32]

Do we plan for this to stick around (i.e. is the eventual state for whether we
do pre-commit GU going to remain server controlled)? I wonder if we shouldn't
call it an experiment then.

[rlarocque, 2013/07/16 01:08:27]

Good point.  I was hoping to remove this flag and make this behaviour the
default eventually, but it could be a long time before we get to that point.  I
could go either way on the name.  

Raz, any thoughts?

+
+  // If true, indicates that we've been passed a command-line flag to force
+  // enable the pre-commit update avoidance experiment described above.
+  bool force_enable_pre_commit_update_avoidance_;
+
   DISALLOW_COPY_AND_ASSIGN(SyncSessionContext);
 };