Add a new function for getting a deterministic machine ID

chrome/browser/prefs/tracked/device_id_unittest.cc

+// Copyright 2015 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "chrome/browser/prefs/tracked/device_id.h"
+#include "build/build_config.h"
+#include "testing/gtest/include/gtest/gtest.h"
+
+TEST(GetDeterministicMachineSpecificIdTest, IsDeterministic) {
+  std::string first_machine_id;
+  std::string second_machine_id;
+
+  const MachineIdStatus kExpectedStatus =
+#if defined(OS_WIN)
+      MachineIdStatus::SUCCESS;
+#else
+      MachineIdStatus::NOT_IMPLEMENTED;
+#endif
+
+  ASSERT_EQ(kExpectedStatus,
+            GetDeterministicMachineSpecificId(&first_machine_id));
+  ASSERT_EQ(kExpectedStatus,
+            GetDeterministicMachineSpecificId(&second_machine_id));
+
+  // The reason for using |EXPECT_TRUE| with one argument instead of |EXPECT_EQ|
+  // with two arguments is a compiler bug in gcc that results in a "converting
+  // 'false' to pointer type" error when the first argument to |EXPECT_EQ| is a
+  // compile-time const false value. See also this bug report:
+  // https://code.google.com/p/googletest/issues/detail?id=322
+  EXPECT_TRUE((kExpectedStatus == MachineIdStatus::SUCCESS) ==

[grt (UTC plus 2), 2015/01/30 14:39:08]

how about getting rid of the hack?
  if (kExpectedStatus == MachineIdStatus::SUCCESS)
    EXPECT_TRUE(!first_machine_id.empty());
  else
    EXPECT_TRUE(first_machine_id.empty();

[gab, 2015/01/30 17:37:16]

I considered that syntax, but found it just as weird (i.e. still requires the
comment to justify it). In that light I prefer the current syntax because it's 2
lines rather than 4.

In practice if anyone ever reads this code, they'll find it weird, then read the
comment and understand. The syntax used currently doesn't make it hacky or
unreadable, it just makes it feel a little overkill until one reads the comment
above IMO.

[grt (UTC plus 2), 2015/01/30 17:48:17]

It seems pretty clear to me, and I don't think it requires a comment about a GCC
bug. That said, do as you wish.

+              !first_machine_id.empty());
+  EXPECT_EQ(first_machine_id, second_machine_id);
+}