Revert of Workaround for clang bug http://llvm.org/PR28348.

webrtc/media/base/videoframe_unittest.h

Index: webrtc/media/base/videoframe_unittest.h
diff --git a/webrtc/media/base/videoframe_unittest.h b/webrtc/media/base/videoframe_unittest.h
index 06afd0eb4c707b672332282dac3f6f1d23434186..f19d0325d6a103b5e2f87d68fbcc21ac1ab0b10f 100644
--- a/webrtc/media/base/videoframe_unittest.h
+++ b/webrtc/media/base/videoframe_unittest.h
@@ -1216,10 +1216,7 @@
       EXPECT_FALSE(expected_result);  // NULL is okay if failure was expected.
       return;
     }
-    // TODO(pbos): Remove kPad once http://llvm.org/PR28348 is fixed and the fix
-    // is rolled in.
-    volatile const int kPad = 4095;
-    data_ptr += kPadToHeapSized + (-(static_cast<int>(data_size)) & kPad);
+    data_ptr += kPadToHeapSized + (-(static_cast<int>(data_size)) & 4095);

[tommi, 2016/06/30 10:39:18]

instead of 4095, can we use 0xfff?  I find it a lot easier to understand what's
going on then.
Also, we're using this constant in several places inside of this function. Maybe
it's worth having a mask constant at the top.

[pbos-webrtc, 2016/06/30 12:16:54]

I agree, but I find this hard to read overall, I'm not even sure what to call
the constant. :\

     memcpy(data_ptr, sample, std::min(data_size, sample_size));
     for (int i = 0; i < repeat_; ++i) {
       EXPECT_EQ(expected_result, frame.Validate(fourcc, kWidth, kHeight,