Ignore OnKeyEvent for VKEY_POWER because it is handled by PowerButtonEvent

ash/system/chromeos/power/tablet_power_button_controller_unittest.cc

Index: ash/system/chromeos/power/tablet_power_button_controller_unittest.cc
diff --git a/ash/system/chromeos/power/tablet_power_button_controller_unittest.cc b/ash/system/chromeos/power/tablet_power_button_controller_unittest.cc
index a6f2b5c0878942d595eef22dc2bf100a79895ea9..62d29218235ef2520200abb3d45104ce8c070879 100644
--- a/ash/system/chromeos/power/tablet_power_button_controller_unittest.cc
+++ b/ash/system/chromeos/power/tablet_power_button_controller_unittest.cc
@@ -21,6 +21,7 @@
 #include "base/test/simple_test_tick_clock.h"
 #include "chromeos/dbus/dbus_thread_manager.h"
 #include "chromeos/dbus/fake_power_manager_client.h"
+#include "ui/events/event.h"
 #include "ui/events/test/event_generator.h"
 
 namespace ash {
@@ -299,5 +300,28 @@ TEST_F(TabletPowerButtonControllerTest, ConvertibleOnMaximizeMode) {
   generator_->ExitPenPointerMode();
 }
 
+// Tests that a single set of power button pressed-and-released operation should
+// cause only one SetBacklightsForcedOff call.
+TEST_F(TabletPowerButtonControllerTest, IgnorePowerOnKeyEvent) {
+  ui::KeyEvent power_key_pressed(ui::ET_KEY_PRESSED, ui::VKEY_POWER,
+                                 ui::EF_NONE);
+  ui::KeyEvent power_key_released(ui::ET_KEY_RELEASED, ui::VKEY_POWER,
+                                  ui::EF_NONE);
+
+  int expectation =
+      power_manager_client_->num_set_backlights_forced_off_calls() + 1;

[Daniel Erat, 2016/11/22 00:48:19]

does this count really not start out at zero? it looks like you're creating a
new FakePowerManagerClient in SetUp (which is run for each test), so this is
surprising.

[Qiang(Joe) Xu, 2016/11/22 01:02:50]

It should start with zero. Thanks for pointing out.

+  // There can be multiple |power_key_pressed| and |power_key_released| events
+  // depends on the duration of holding.
+  tablet_controller_->OnKeyEvent(&power_key_pressed);
+  PressPowerButton();
+  tablet_controller_->OnKeyEvent(&power_key_pressed);
+  tablet_controller_->OnKeyEvent(&power_key_pressed);
+  tablet_controller_->OnKeyEvent(&power_key_released);

[Daniel Erat, 2016/11/22 00:48:19]

are there really multiple release events? i'd expect there to be multiple press
events (for autorepeat) but only one release event.

[Qiang(Joe) Xu, 2016/11/22 01:02:51]

Yes, there are multiple release events. I made the logs before I wrote the
tests. Please see https://paste.googleplex.com/6168565862367232.

+  ReleasePowerButton();
+  tablet_controller_->OnKeyEvent(&power_key_released);
+  EXPECT_EQ(power_manager_client_->num_set_backlights_forced_off_calls(),

[Daniel Erat, 2016/11/22 00:48:19]

nit: the order should be EXPECT_EQ(expected, actual) in order for the failure
messages to make sense

[Qiang(Joe) Xu, 2016/11/22 01:02:50]

Done.

+            expectation);
+}
+
 }  // namespace test
 }  // namespace ash