Propagate the encoded data length through OnDownloadData delegate method

content/browser/loader/async_resource_handler.cc

Index: content/browser/loader/async_resource_handler.cc
diff --git a/content/browser/loader/async_resource_handler.cc b/content/browser/loader/async_resource_handler.cc
index 02e5552b901e97955e22bdac5d26d1d8ca1384e8..992eb652d1c9c3931cf4e2b0bc132c0fb66e0e26 100644
--- a/content/browser/loader/async_resource_handler.cc
+++ b/content/browser/loader/async_resource_handler.cc
@@ -284,8 +284,11 @@ bool AsyncResourceHandler::OnReadCompleted(int request_id, int bytes_read,
 
 void AsyncResourceHandler::OnDataDownloaded(
     int request_id, int bytes_downloaded) {
+  int encoded_data_length =
+      DevToolsNetLogObserver::GetAndResetEncodedDataLength(request_);

[michaeln, 2013/08/29 20:19:05]

I'm not familiar with the inspector instrumentation but i see the exact same
handling for the non-download-to-file case in OnReadCompleted so i trust it
returns valid results in this case too since this step replace that step in the
d-t-f control flow.

[Nico, 2013/08/31 20:49:47]

I'm also curious about
this.DevToolsNetLogObserver::GetAndResetEncodedDataLength() seems to return -1
if dev_tools_net_log_observer == NULL, and instance_ in
content/browser/devtools/devtools_netlog_observer.cc seems to be only non-NULL
after Attach() was called (and before Detach() was called), which after skimming
seems to be only the case while devtools is active (?)

[yusukesuzuki, 2013/09/02 02:05:10]

Right. The encoded_data_length is used for the Inspector, so it becomes valid
while devtools is active.
The encoded_data_length may become -1 and it represents that the encoded data
length is unknown.

+
   filter_->Send(new ResourceMsg_DataDownloaded(
-      routing_id_, request_id, bytes_downloaded));
+      routing_id_, request_id, bytes_downloaded, encoded_data_length));
 }
 
 bool AsyncResourceHandler::OnResponseCompleted(