[NoStatePrefetch] Do not send responses to renderer in prefetch mode.

content/renderer/render_frame_impl.cc

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "content/renderer/render_frame_impl.h"
 
 #include <map>
 #include <string>
 #include <utility>
 #include <vector>
@@ skipping 4051 matching lines @@
   extra_data->set_allow_download(
       navigation_state->common_params().allow_download);
   extra_data->set_transition_type(transition_type);
   extra_data->set_should_replace_current_entry(should_replace_current_entry);
   extra_data->set_transferred_request_child_id(
       navigation_state->start_params().transferred_request_child_id);
   extra_data->set_transferred_request_request_id(
       navigation_state->start_params().transferred_request_request_id);
   extra_data->set_service_worker_provider_id(provider_id);
   extra_data->set_stream_override(std::move(stream_override));
+  bool is_prefetch = frame->document().isPrefetchOnly();
+  extra_data->set_is_prefetch(is_prefetch);
+  extra_data->set_download_to_network_cache_only(
+      is_prefetch &&
+      WebURLRequestToResourceType(request) != RESOURCE_TYPE_MAIN_FRAME);

[mattcary, 2016/08/29 09:27:54]

Why do we not set this for main frame resources?

[droger, 2016/08/29 09:32:42]

This flags prevents the responses from going back to the renderer.

When we do a NoStatePrefetch, we want the main resource to go to the renderer,
to be able to run the preload scanner on it.

[droger, 2016/08/29 09:35:55]

Basically I use this to distinguish the main HTML document from the
sub-resources like images and javascripts. I'm not using this to make a
distinction main-frame / sub-frame. Let me know if you think this is a bad use
of the flag.

[clamy, 2016/08/30 18:16:13]

Do we want iframes to also be downloaded to cache? If yes, this works, otherwise
you should use IsResourceTypeFrame to also exclude iframes (which have
RESOURCE_TYPE_SUB_FRAME).

Not that I wonder how this would work with more exotic types, such as
RESOURCE_TYPE_PING or RESOURCE_TYPE_SERVICE_WORKER. Do we want to try to
download them to cache, or just stop te request?

[droger, 2016/09/01 09:37:32]

For now we want to only let the main request go through, and everything else
goes to cache. We may add support for sub frames later.

About the more exotic requests, it's not very clear what kind of side effect
these can have. The current behavior with prerendering is not cancelling them,
so I think we can treat them like the other resources for now and see what
happens. Does that sounds good?

[clamy, 2016/09/01 18:30:25]

Could you add a comment explaining this? This way people reading the code will
know that it is intentional that subframes follow this as well. Maybe add a TODO
as well to check if all types of resources should go to cache.

   WebString error;
   extra_data->set_initiated_in_secure_context(
       frame->document().isSecureContext(error));
   request.setExtraData(extra_data);
 
   if (request.getLoFiState() == WebURLRequest::LoFiUnspecified) {
     if (is_main_frame_ && !navigation_state->request_committed()) {
       request.setLoFiState(static_cast<WebURLRequest::LoFiState>(
           navigation_state->common_params().lofi_state));
     } else {
@@ skipping 2266 matching lines @@
   // event target. Potentially a Pepper plugin will receive the event.
   // In order to tell whether a plugin gets the last mouse event and which it
   // is, we set |pepper_last_mouse_event_target_| to null here. If a plugin gets
   // the event, it will notify us via DidReceiveMouseEvent() and set itself as
   // |pepper_last_mouse_event_target_|.
   pepper_last_mouse_event_target_ = nullptr;
 #endif
 }
 
 }  // namespace content