Component updater on-demand logic with ResourceThrottle

chrome/browser/renderer_host/chrome_resource_dispatcher_host_delegate.cc

Index: chrome/browser/renderer_host/chrome_resource_dispatcher_host_delegate.cc
===================================================================
--- chrome/browser/renderer_host/chrome_resource_dispatcher_host_delegate.cc	(revision 226022)
+++ chrome/browser/renderer_host/chrome_resource_dispatcher_host_delegate.cc	(working copy)
@@ -9,8 +9,10 @@
 #include "base/base64.h"
 #include "base/logging.h"
 #include "base/metrics/histogram.h"
+#include "base/strings/string_util.h"
 #include "chrome/browser/browser_process.h"
 #include "chrome/browser/chrome_notification_types.h"
+#include "chrome/browser/component_updater/component_updater_service.h"
 #include "chrome/browser/content_settings/host_content_settings_map.h"
 #include "chrome/browser/download/download_request_limiter.h"
 #include "chrome/browser/download/download_resource_throttle.h"
@@ -191,6 +193,27 @@
   }
 }
 
+void AppendComponentUpaterThrottles(
+    net::URLRequest* request,
+    content::ResourceContext* resource_context,
+    ResourceType::Type resource_type,
+    ScopedVector<content::ResourceThrottle>* throttles) {
+
+  char* crx_id = NULL;
+
+  // Each component can go bananas here with custom checks. Here is an
+  // example of pNaCL doing a custom check.
+  if (resource_type == ResourceType::OBJECT) {
+    if (EndsWith(request->url().ExtractFileName(), ".pexe", false))
+      crx_id = "hnimpnehoodheedghdeeijklkeaacbdc";

[darin (slow to review), 2013/10/03 21:31:43]

we probably want to base this on mime type instead of file extension. it is
possible for the file extension to be anything. the mime type is what matters.

That said, the approach you are taking nicely avoids the problem of the TCP
socket stream being left hanging for a long time. Hmm...

Perhaps we can get a more clear signal that it is NaCl requesting a PEXE.

[jvoung (off chromium), 2013/10/09 17:05:22]

Separately in a mail thread, I suggested adding something to the request
headers. E.g., could we make the NaCl pexe URLloader set "Accept:
application/x-pnacl" in the request header?

Darin, is that the appropriate use of "Accept:" ?

What is the issue about having the TCP socket stream being left hanging for a
long time?  Is it that if we try to get something based on a *response* we will
have started the stream, and that will be left hanging, while if we check
something that doesn't require a response, we won't have started the stream yet?

[cpu_(ooo_6.6-7.5), 2013/10/20 23:27:25]

Yeah, a special header sounds nice to me.
The server might close the socket or otherwise time out the request.

+  }
+
+  if (crx_id) {
+    ComponentUpdateService* cus = g_browser_process->component_updater();
+    throttles->push_back(cus->GetResourceThrottle(request, crx_id));
+  }
+}
+
 }  // end namespace
 
 ChromeResourceDispatcherHostDelegate::ChromeResourceDispatcherHostDelegate(
@@ -317,6 +340,12 @@
                                   resource_context,
                                   resource_type,
                                   throttles);
+  if (!is_prerendering) {
+    AppendComponentUpaterThrottles(request,
+                                   resource_context,
+                                   resource_type,
+                                   throttles);
+  }
 
   if (io_data->resource_prefetch_predictor_observer()) {
     io_data->resource_prefetch_predictor_observer()->OnRequestStarted(