Hook up TABLET_MODE switch to maximize_mode_controller

ash/wm/maximize_mode/maximize_mode_controller.cc

Index: ash/wm/maximize_mode/maximize_mode_controller.cc
diff --git a/ash/wm/maximize_mode/maximize_mode_controller.cc b/ash/wm/maximize_mode/maximize_mode_controller.cc
index 9d8b18dae29de82390b61d4e439eca666dfd59db..939907a511927c053e812506feee7e95e32237fa 100644
--- a/ash/wm/maximize_mode/maximize_mode_controller.cc
+++ b/ash/wm/maximize_mode/maximize_mode_controller.cc
@@ -103,6 +103,9 @@ MaximizeModeController::MaximizeModeController()
       lid_open_past_180_(false),
       touchview_usage_interval_start_time_(base::Time::Now()),
       tick_clock_(new base::DefaultTickClock()),
+#if defined(OS_CHROMEOS)
+      tablet_mode_(false),
+#endif
       lid_is_closed_(false) {
   Shell* shell = Shell::GetInstance();
   shell->AddShellObserver(this);
@@ -224,6 +227,20 @@ void MaximizeModeController::LidEventReceived(bool open,
   LeaveMaximizeMode();
 }
 
+void MaximizeModeController::TabletModeEventReceived(
+    bool on,
+    const base::TimeTicks& time) {
+  if (on) {
+    tablet_mode_ = true;
+    EnterMaximizeMode();
+  } else {
+    tablet_mode_ = false;
+    // Set lid_open_past_180_ so that we get a chance to decide whether to
+    // leave maximize mode.
+    lid_open_past_180_ = true;

[Daniel Erat, 2016/06/03 15:00:18]

this doesn't seem right. this member feels like it should only be updated by
HandleHingeRotation().

[jonross, 2016/06/03 16:27:24]

Though the lid_open_past_180_ may actually be what the sensor is telling us.

[jcliang, 2016/06/06 04:39:22]

This is removed. I use another tablet_mode_swithc_was_on_ to achieve the same
result in HandleHingeRotation.

+  }
+}
+
 void MaximizeModeController::SuspendImminent() {
   // The system is about to suspend, so record TouchView usage interval metrics
   // based on whether TouchView mode is currently active.
@@ -323,6 +340,11 @@ void MaximizeModeController::EnterMaximizeMode() {
 }
 
 void MaximizeModeController::LeaveMaximizeMode() {
+#if defined(OS_CHROMEOS)
+  if (tablet_mode_)

[Daniel Erat, 2016/06/03 15:00:18]

i don't think i've looked at this code before, but adding this check here,
rather than in the places that call LeaveMaximizeMode(), seems a bit weird. i'd
prefer that the logic to decide whether we should be in maximize mode or not
didn't live inside of this method. (a more common approach might be something
like an UpdateState() method that looks at various factors to decide which mode
it should be in and then enters or leaves the mode.)

[jonross, 2016/06/03 16:27:24]

The 'update state' method is HandleHingeRotation. Which could be a more
appropriate place for this exit. These methods are separated out for use with a
testing flag, and for unittesting.

[Daniel Erat, 2016/06/03 16:59:51]

what if the "decide what to do" code from HandleHingeRotation were separated out
into a new UpdateState or SelectMode or whatever method that
HandleHingeRotation, TabletModeEventReceived, etc. call after updating the
lid_open_past_180_, tablet_mode_, etc. members?

[jcliang, 2016/06/06 04:39:22]

This is removed. Everything is handled in HandleHingeRotation now.

+    return;
+#endif
+
   event_blocker_.reset();
 
   if (base::CommandLine::ForCurrentProcess()->HasSwitch(