Break rtp_rtcp_format out of rtp_rtcp, to resolve circular dependencies

webrtc/modules/rtp_rtcp/source/rtp_header_extension_map.cc

 /*
  *  Copyright (c) 2012 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.
  */
 
 #include "webrtc/modules/rtp_rtcp/include/rtp_header_extension_map.h"
 
 #include "webrtc/modules/rtp_rtcp/source/rtp_header_extensions.h"
-#include "webrtc/modules/rtp_rtcp/source/rtp_utility.h"
 #include "webrtc/rtc_base/arraysize.h"
 #include "webrtc/rtc_base/checks.h"
 #include "webrtc/rtc_base/logging.h"
 
 namespace webrtc {
 namespace {
 
-using RtpUtility::Word32Align;
-
 struct ExtensionInfo {
   RTPExtensionType type;
   const char* uri;
 };
 
 template <typename Extension>
 constexpr ExtensionInfo CreateExtensionInfo() {
   return {Extension::kId, Extension::kUri};
 }
 
@@ skipping 61 matching lines @@
   static constexpr size_t kRtpOneByteHeaderLength = 4;
   // Header size of each individual extension, see RFC5285 Section 4.2
   static constexpr size_t kExtensionHeaderLength = 1;
   size_t values_size = 0;
   for (const RtpExtensionSize& extension : extensions) {
     if (IsRegistered(extension.type))
       values_size += extension.value_size + kExtensionHeaderLength;
   }
   if (values_size == 0)
     return 0;
-  return Word32Align(kRtpOneByteHeaderLength + values_size);
+  size_t size = kRtpOneByteHeaderLength + values_size;
+  // Round up to the nearest size that is a multiple of 4.
+  // Which is same as round down (size + 3).
+  return size + 3 - (size + 3) % 4;

[eladalon, 2017/09/12 15:25:24]

I still think this would be safer, clearer and more efficient:
  return size & (std::numeric_limits<size_t>::max() - 3);
1. If I am not mistaken, bitwise-and will always be more efficient than modulo.
2. The left side of the binary-and operator is compile-time, not runtime,
computable.

[danilchap, 2017/09/12 15:40:31]

For any reasonable use (there are up to 16 types of extension each of up to 16
bytes long, so size is always very small) I find both solutions equally safe.

personally I find rounding with reminder clearer than rounding with bitwise
operator (last one rely on knowledge how integers are layed out in memory, first
one rely on a bit of math)

bitwise is more efficient, but extra efficiency I fine negligible.

 }
 
 int32_t RtpHeaderExtensionMap::Deregister(RTPExtensionType type) {
   if (IsRegistered(type)) {
     uint8_t id = GetId(type);
     types_[id] = kInvalidType;
     ids_[type] = kInvalidId;
   }
   return 0;
 }
@@ skipping 24 matching lines @@
     return false;
   }
   RTC_DCHECK(!IsRegistered(type));
 
   types_[id] = type;
   ids_[type] = id;
   return true;
 }
 
 }  // namespace webrtc