Fixing a problem, where a hung renderer process is not killed when navigating away

content/browser/renderer_host/mock_render_process_host.cc

Index: content/browser/renderer_host/mock_render_process_host.cc
diff --git a/content/browser/renderer_host/mock_render_process_host.cc b/content/browser/renderer_host/mock_render_process_host.cc
index 626960ac672eb7ee9a9f483290f442e5b1044a9b..8110471dedbe65944a69a6db2160c07ebc544474 100644
--- a/content/browser/renderer_host/mock_render_process_host.cc
+++ b/content/browser/renderer_host/mock_render_process_host.cc
@@ -213,6 +213,13 @@ bool MockRenderProcessHost::OnMessageReceived(const IPC::Message& msg) {
 void MockRenderProcessHost::OnChannelConnected(int32 peer_pid) {
 }
 
+bool MockRenderProcessHost::AllowTerminateOnUnresponsive() const {
+  if (listeners_.size() > 1)
+    return false;

[Charlie Reis, 2012/03/01 21:08:17]

Is this necessary for tests?  Usually the Mock version just hardcodes a
reasonable answer, like true.

[nasko, 2012/03/01 22:50:15]

Is there a reason to have a hardcoded value, especially with such a simple
function body?

[Charlie Reis, 2012/03/02 00:16:43]

In general, it might get out of date if we change the other one.  It might also
not do what you think, depending on how listeners_ is updated in the Mock.

[nasko, 2012/03/02 18:40:36]

If the function is now renamed to IsSharedByMultipleViews, does it make sense to
hard code the value, especially when the mock host does proper accounting for
listeners?

[Charlie Reis, 2012/03/02 19:17:02]

No, we can leave it as is.

+
+  return true;
+}
+
 MockRenderProcessHostFactory::MockRenderProcessHostFactory() {}
 
 MockRenderProcessHostFactory::~MockRenderProcessHostFactory() {