Render process changes for ResourceTiming sizes.

content/public/child/request_peer.h

Index: content/public/child/request_peer.h
diff --git a/content/public/child/request_peer.h b/content/public/child/request_peer.h
index 2b054b38d77b62773b0d7dc7c9d3abab32d14a3a..64173cb6ee673d2074edbdfd021a0017c261722a 100644
--- a/content/public/child/request_peer.h
+++ b/content/public/child/request_peer.h
@@ -36,21 +36,30 @@ struct ResourceResponseInfo;
 class CONTENT_EXPORT RequestPeer {
  public:
   // This class represents data gotten from the Browser process. Each data
-  // consists of |payload|, |length| and |encoded_length|. The payload is
-  // valid only when the data instance is valid.
+  // consists of |payload|, |length|, |encoded_data_length| and
+  // |encoded_body_length|. The payload is valid only when the data instance is
+  // valid.
   // In order to work with Chrome resource loading IPC, it is desirable to
   // reclaim data in FIFO order in a RequestPeer in terms of performance.
-  // |payload|, |length| and |encoded_length| functions are thread-safe, but
-  // the data object itself must be destroyed on the original thread.
+  // |payload|, |length|, |encoded_data_length| and |encoded_body_length|
+  // functions
+  // are thread-safe, but the data object itself must be destroyed on the
+  // original thread.

[kinuko, 2016/07/04 04:08:12]

nit: weird line wrapping

[Adam Rice, 2016/07/04 06:33:42]

Fixed.

   class CONTENT_EXPORT ReceivedData {
    public:
     virtual ~ReceivedData() {}
     virtual const char* payload() const = 0;
     virtual int length() const = 0;
-    // The encoded_length is the length of the encoded data transferred
-    // over the network, which could be different from data length (e.g. for
-    // gzipped content).
-    virtual int encoded_length() const = 0;
+    // The encoded_data_length is the length of the encoded data transferred
+    // over the network, including headers. It is only set for responses
+    // originating from the network (ie. not the cache). It will usually be
+    // different from length(), and may be smaller if the content was
+    // compressed.
+    virtual int encoded_data_length() const = 0;

[kinuko, 2016/07/04 04:08:12]

We often seem to set this to -1 in tests and also in some code paths (while not
for other fields), should it be explicitly communicated here?  (I'm not fully
sure why we specifically use -1 for this one)

[Adam Rice, 2016/07/04 06:33:42]

I have added a comment that "-1 means this value is unavailable." It appears
that in practice 0 means "no data was received over the network for this chunk"
but that doesn't appear to be explicitly documented anywhere.

Since this field is actually an incremental value for just this chunk, -1 is
probably not a good idea, but I will have to audit all consumers of this value
to be sure we can change the semantics safely.

+    // The encoded_body_length is the size of the body as transferred over the
+    // network or stored in the disk cache. This will be different from the data

[kinuko, 2016/07/04 04:08:12]

'excluding headers' ...?

[Adam Rice, 2016/07/04 06:33:42]

Added.

+    // length if a content encoding was used.
+    virtual int encoded_body_length() const = 0;
   };
 
   // A ThreadSafeReceivedData can be deleted on ANY thread.