Replace uses of raw uint32's with a type-checked RtpTimeTicks data type.

media/cast/common/expanded_value_base_unittest.cc

Index: media/cast/common/expanded_value_base_unittest.cc
diff --git a/media/cast/common/expanded_value_base_unittest.cc b/media/cast/common/expanded_value_base_unittest.cc
new file mode 100644
index 0000000000000000000000000000000000000000..587f682b5d8bbc39d166a74956ccfabb78dc832b
--- /dev/null
+++ b/media/cast/common/expanded_value_base_unittest.cc
@@ -0,0 +1,107 @@
+// 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 "media/cast/common/expanded_value_base.h"
+#include "testing/gtest/include/gtest/gtest.h"
+
+namespace media {
+namespace cast {
+
+namespace {
+
+// A basic subclass of ExpandedValueBase to use for testing.
+class TestValue : public ExpandedValueBase<int64_t, TestValue> {
+ public:
+  explicit TestValue(int64_t value) : ExpandedValueBase(value) {}
+};
+
+}  // namespace
+
+// Tests the various scenarios of truncating and then re-expanding values, and
+// confirming the correct behavior.  Note that, while the code below just tests
+// truncation/expansion to/from 8 bits, the 16- and 32-bit cases are implicitly
+// confirmed because the Expand() method uses the compiler to derive all of its
+// constants (based on the type of its argument).
+TEST(ExpandedValueBaseTest, TruncationAndExpansion) {
+  // Test that expansion works when the reference is always equal to the value
+  // that was truncated.
+  for (int64_t i = -512; i <= 512; ++i) {
+    const TestValue original_value(i);
+    const uint8_t truncated = original_value.lower_8_bits();
+    const TestValue reference(i);
+    ASSERT_EQ(original_value, reference.Expand(truncated)) << "i=" << i;
+  }
+
+  // Test that expansion works when the reference is always one less than the
+  // value that was truncated.
+  for (int64_t i = -512; i <= 512; ++i) {
+    const TestValue original_value(i);
+    const uint8_t truncated = original_value.lower_8_bits();
+    const TestValue reference(i - 1);
+    ASSERT_EQ(original_value, reference.Expand(truncated)) << "i=" << i;
+  }
+
+  // Test that expansion works when the reference is always one greater than the
+  // value that was truncated.
+  for (int64_t i = -512; i <= 512; ++i) {
+    const TestValue original_value(i);
+    const uint8_t truncated = original_value.lower_8_bits();
+    const TestValue reference(i + 1);
+    ASSERT_EQ(original_value, reference.Expand(truncated)) << "i=" << i;
+  }
+
+  // Test cases where the difference between the original value and the fixed
+  // reference is within the range [-128,+127].  The truncated value should
+  // always be re-expanded to the original value.
+  for (int64_t bias = -5; bias <= 5; ++bias) {
+    for (int64_t i = -128; i <= 127; ++i) {
+      const TestValue original_value(bias + i);
+      const uint8_t truncated = original_value.lower_8_bits();
+      const TestValue reference(bias);
+      ASSERT_EQ(original_value, reference.Expand(truncated)) << "bias=" << bias
+                                                             << ", i=" << i;
+    }
+  }
+
+  // Test cases where the difference between the original value and the fixed
+  // reference is within the range [+128,+255].  When the truncated value is
+  // re-expanded, it should be 256 less than the original value.
+  for (int64_t bias = -5; bias <= 5; ++bias) {
+    for (int64_t i = 128; i <= 255; ++i) {
+      // Example: Let |original_value| be 192.  Then, the truncated 8-bit value
+      // will be 0xc0.  When a |reference| of zero is asked to expand 0xc0 back
+      // to the original value, it should produce -64 since -64 is closer to
+      // |reference| than 192.
+      const TestValue original_value(bias + i);
+      const uint8_t truncated = original_value.lower_8_bits();
+      const TestValue reexpanded_value(bias + i - 256);
+      ASSERT_EQ(reexpanded_value.lower_8_bits(), truncated);
+      const TestValue reference(bias);
+      ASSERT_EQ(reexpanded_value, reference.Expand(truncated))
+          << "bias=" << bias << ", i=" << i;
+    }
+  }
+
+  // Test cases where the difference between the original value and the fixed
+  // reference is within the range [-256,-129].  When the truncated value is
+  // re-expanded, it should be 256 more than the original value.
+  for (int64_t bias = -5; bias <= 5; ++bias) {
+    for (int64_t i = -256; i <= -129; ++i) {
+      // Example: Let |original_value| be -192.  Then, the truncated 8-bit value
+      // will be 0x40.  When a |reference| of zero is asked to expand 0x40 back
+      // to the original value, it should produce 64 since 64 is closer to the
+      // |reference| than -192.

[Irfan, 2015/12/10 01:40:38]

Really good explanation, thanks!

+      const TestValue original_value(bias + i);
+      const uint8_t truncated = original_value.lower_8_bits();
+      const TestValue reexpanded_value(bias + i + 256);
+      ASSERT_EQ(reexpanded_value.lower_8_bits(), truncated);
+      const TestValue reference(bias);
+      ASSERT_EQ(reexpanded_value, reference.Expand(truncated))
+          << "bias=" << bias << ", i=" << i;
+    }
+  }
+}
+
+}  // namespace cast
+}  // namespace media