Add a PrintTo function for rtc::Optional to aid with testing.

webrtc/base/optional.h

 /*
  *  Copyright 2015 The WebRTC Project Authors. All rights reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
 #ifndef WEBRTC_BASE_OPTIONAL_H_
 #define WEBRTC_BASE_OPTIONAL_H_
 
 #include <algorithm>
 #include <memory>
+#include <ostream>
 #include <utility>
 
 #include "webrtc/base/array_view.h"
 #include "webrtc/base/checks.h"
 #include "webrtc/base/sanitizer.h"
 
 namespace rtc {
 
 namespace optional_internal {
 
@@ skipping 263 matching lines @@
     char empty_;
     // By placing value_ in a union, we get to manage its construction and
     // destruction manually: the Optional constructors won't automatically
     // construct it, and the Optional destructor won't automatically destroy
     // it. Basically, this just allocates a properly sized and aligned block of
     // memory in which we can manually put a T with placement new.
     T value_;
   };
 };
 
+// Add a PrintTo function for Optionals for use with gtest, only if
+// there is already a valid operator<< available for the enclosed type.
+// The cast in decltype will ensure PrintTo has a void return type.
+template <typename T>
+auto PrintTo(const Optional<T>& opt, std::ostream* os)

[ossu, 2017/02/16 14:23:48]

Turns out the compiler does clever things that linux_memcheck doesn't like. I'll
have to replicate what value_or() does to get it to stop complaining.

[tommi, 2017/02/17 13:46:23]

Can we call this PrintToForTest?

There's a risk that more and more ostream related things start to creep into the
non-test code base and there is already a precedence for "ForTest" methods,
including a presubmit check that prevents non-test code from calling those
methods.

[ossu, 2017/02/17 13:50:06]

Alas, no, it needs to be called PrintTo for gtest to pick it up, c.f.
https://github.com/google/googletest/blob/master/googletest/docs/AdvancedGuid...

+    -> decltype(static_cast<void>(*os << *opt)) {

[ossu, 2017/02/16 13:20:05]

The SFINAE here is maybe not super-obvious if one's not worked with such
constructs before. Should I explain it better?
Essentially, the template will be skipped if *os << *opt isn't a valid
expression, since the function will not have a valid return type. Regardless of
the return type of that expression, I'm forcing the decltype() to be void, if
the expression is at all valid, so the return type will be void.

[kwiberg-webrtc, 2017/02/16 14:07:37]

Yes, more explanation is probably called for. How about documenting what it is
and how it behaves in one paragraph, and explaining the details on how it works
in a second paragraph? (The explanation in your message here was good, I think.)

[ossu, 2017/02/16 14:23:48]

Will do!

+  if (opt) {
+    *os << *opt;

[kwiberg-webrtc, 2017/02/16 14:07:37]

This is test code, so maybe indicate that the value is contained in an optional?

[ossu, 2017/02/16 14:23:48]

I considered that but was unsure if it would be more confusing than not. I think
I'd rather not, if it's not very important. I'll think about it some more,
though.

+  } else {
+    *os << "<empty optional>";
+  }
+}
+
 }  // namespace rtc
 
 #endif  // WEBRTC_BASE_OPTIONAL_H_