Roll googletest to 1.8.0.

device/bluetooth/dbus/fake_bluetooth_media_endpoint_service_provider.h

Index: device/bluetooth/dbus/fake_bluetooth_media_endpoint_service_provider.h
diff --git a/device/bluetooth/dbus/fake_bluetooth_media_endpoint_service_provider.h b/device/bluetooth/dbus/fake_bluetooth_media_endpoint_service_provider.h
index 6a9e6ae6757f5fb50802ecf5e65be85f80479b80..f907e4da7324d4adb5c7338e151f9618ad8c6113 100644
--- a/device/bluetooth/dbus/fake_bluetooth_media_endpoint_service_provider.h
+++ b/device/bluetooth/dbus/fake_bluetooth_media_endpoint_service_provider.h
@@ -14,7 +14,7 @@
 #include "dbus/object_path.h"
 #include "device/bluetooth/bluetooth_export.h"
 #include "device/bluetooth/dbus/bluetooth_media_endpoint_service_provider.h"
-#include "testing/gtest/include/gtest/gtest_prod.h"
+#include "testing/gtest/include/gtest/gtest_prod.h"  // nogncheck

[Nico, 2017/04/28 00:46:44]

Can you explain why this is needed?

[pwnall, 2017/04/28 00:59:00]

This appeases "gn check".

Longer version: gn's deps check (completely separate from checkdeps.py, which
uses DEPS) uses a target's "deps" in BUILD.gn, and ignores files that are not in
the repository. For this reason, #including gtest_prod did not trigger "gn
check" while testing/gtest and testing/gmock were brought in via DEPS.

I think that the correct long-term fix is to add a gtest_prod target to GN,
which would only contain the gtest_prod header file. The GN targets with files
who #include gtest_prod would have to add gtest_prod to their deps.

One of the files #includes gtest_prod is base/gtest_prod_util.h (also patched in
this CL), so //base would need this dependency. I expect that this CL will be
tough to land as-is, so I used nogncheck instead of building up the gtest_prod
target to avoid growing the CL even further, and making it more difficult to
land.

Does this seem reasonable?

 
 namespace bluez {