Cancel prerender when we discover a download starting from a page we are prer...

chrome/browser/renderer_host/download_resource_handler.cc

Index: chrome/browser/renderer_host/download_resource_handler.cc
===================================================================
--- chrome/browser/renderer_host/download_resource_handler.cc	(revision 74377)
+++ chrome/browser/renderer_host/download_resource_handler.cc	(working copy)
@@ -15,13 +15,29 @@
 #include "chrome/browser/download/download_file_manager.h"
 #include "chrome/browser/history/download_create_info.h"
 #include "chrome/browser/renderer_host/global_request_id.h"
+#include "chrome/browser/renderer_host/render_view_host.h"
 #include "chrome/browser/renderer_host/resource_dispatcher_host.h"
 #include "chrome/browser/renderer_host/resource_dispatcher_host_request_info.h"
+#include "chrome/common/notification_service.h"
 #include "chrome/common/resource_response.h"
 #include "net/base/io_buffer.h"
 #include "net/http/http_response_headers.h"
 #include "net/url_request/url_request_context.h"
 
+namespace {
+void NotifyOnUI(NotificationType::Type notification_type,
+                int process_id, int view_id) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  RenderViewHost* rvh = RenderViewHost::FromID(process_id, view_id);
+  if (!rvh)
+    return;
+
+  NotificationService::current()->Notify(notification_type,
+                                         Source<RenderViewHost>(rvh),
+                                         NotificationService::NoDetails());
+}
+}  // namespace
+
 DownloadResourceHandler::DownloadResourceHandler(
     ResourceDispatcherHost* rdh,
     int render_process_host_id,
@@ -46,6 +62,10 @@
       rdh_(rdh),
       is_paused_(false),
       url_check_pending_(false) {
+   BrowserThread::PostTask(
+      BrowserThread::UI, FROM_HERE,
+      NewRunnableFunction(&NotifyOnUI, NotificationType::DOWNLOAD_INITIATED,
+                          render_process_host_id, render_view_id_));

[Randy Smith (Not in Mondays), 2011/02/13 03:08:36]

I'd rather not notify from the DownloadResourceHandler constructor, as that'll
mean in the case where a download comes off of the usual path (i.e. through the
DownloadThrottlingResourceHandler) you'll get two notifications that the same
download is starting.  I don't like the fragility of putting the notification in
ResourceDispatcherHost::BeginDownload, but it has the advantage of covering all
the (current :-{) pathways through which a download can be initiated while still
giving you the notification you're looking for when the initiation doesn't
actually go through to creating a DownloadResourceHandler.

[dominich, 2011/02/14 16:07:12]

I was considering download_util::DownloadUrl as an alternative to
RDH::BeginDownload as it may be a little more future-proof. What do you think?

[Randy Smith (Not in Mondays), 2011/02/14 19:20:51]

Neither one's particularly future-proof, I'm afraid :-{.  I have (vague) plans
to get rid of download_util based on a recent rant of Brett's on chromium-dev,
and no current plans to get rid of RDH::BeginDownload, but I don't consider
either of them part of a particularly clean architecture.  It's sort of like the
breakfast cereal adds: "Standing next to ^W ^W Part of this complete
architecture" :-} :-J.

Having said that, I'd vote in favor of RDH just because there's a path that
initiales a download that doesn't get notified if we use download_util:
ResourceMessageFilter::OnDownloadUrl -> ResourceDispatcherHost::BeginDownload->
DownloadResourceHandler().  That happens when the renderer processes sends us a
message requesting a download; I'm still not sure under what circumstances that
happens, but if the path's in the code, I'd like to catch it.

 }
 
 bool DownloadResourceHandler::OnUploadProgress(int request_id,