CORS-RFC1918: Introduce the 'addressSpace' IDL attributes.

content/child/web_url_loader_impl.cc

Index: content/child/web_url_loader_impl.cc
diff --git a/content/child/web_url_loader_impl.cc b/content/child/web_url_loader_impl.cc
index 5abc0fbc26a4ab2e716825e319f5da77ddd61d8a..93e8ee4186dbb679efd8e9652758544a73023a97 100644
--- a/content/child/web_url_loader_impl.cc
+++ b/content/child/web_url_loader_impl.cc
@@ -955,7 +955,7 @@ void WebURLLoaderImpl::PopulateURLResponse(const GURL& url,
   response->setWasCached(!info.load_timing.request_start_time.is_null() &&
       info.response_time < info.load_timing.request_start_time);
   response->setRemoteIPAddress(
-      WebString::fromUTF8(info.socket_address.host()));
+      WebString::fromUTF8(info.socket_address.HostForURL()));

[philipj_slow, 2016/03/04 04:35:08]

Is this the "pretty bad bug"? I can't review here. Do any of the tests guard
against regressing this?

[Mike West, 2016/03/04 05:53:54]

Yes. `host` returns an IPv6 address as `::1`, while `HostForURL` returns it with
brackets, as `[::1]`. The latter format is what we need to pass into the
platform API that tells us whether or not the IP address is reserved.
I discovered the bug when writing the layout tests for this patch. I'll see if I
can add a unit test somewhere in //content that verifies the format.

[Mike West, 2016/03/04 08:58:21]

Added 'WebURLLoaderImplTest.ResponseIPAddress'.

   response->setRemotePort(info.socket_address.port());
   response->setConnectionID(info.load_timing.socket_log_id);
   response->setConnectionReused(info.load_timing.socket_reused);