Implementation of PostTaskAndReply() in MessageLoopProxy and BrowserThread.

content/browser/browser_thread.cc

Index: content/browser/browser_thread.cc
diff --git a/content/browser/browser_thread.cc b/content/browser/browser_thread.cc
index b4d95ce32f93f4c71be291cb1bd697abc90eb2f7..09d71860cdd939a26d32612bc4dd50ea12b0df8d 100644
--- a/content/browser/browser_thread.cc
+++ b/content/browser/browser_thread.cc
@@ -4,6 +4,7 @@
 
 #include "content/browser/browser_thread.h"
 
+#include "base/bind.h"
 #include "base/message_loop.h"
 #include "base/message_loop_proxy.h"
 #include "base/threading/thread_restrictions.h"
@@ -221,6 +222,20 @@ bool BrowserThread::PostNonNestableDelayedTask(
 }
 
 // static
+bool BrowserThread::PostTaskAndReply(
+    ID identifier,
+    const tracked_objects::Location& from_here,
+    const base::Closure& task,
+    const base::Closure& reply) {
+  using base::internal::PostTaskAndReplyRelay;
+  PostTaskAndReplyRelay* relay =
+      new PostTaskAndReplyRelay(from_here, task, reply);
+  return PostTask(identifier, from_here,
+                  base::Bind(&PostTaskAndReplyRelay::Run,

[darin (slow to review), 2011/08/17 17:20:39]

It almost seems like there could be room for a static function defined within
the base:: namespace that takes (from_here, task, reply) and returns a
"PostTaskAndReplyRelay" base::Closure.  This way you don't need to replicate
this somewhat subtle setup and invocation code for using PostTaskAndReplyRelay.

Plus, using the base::internal:: namespace outside of base/ is just wrong.  We
should have a presubmit check that no one does this.

  return PostTask(identifier, from_here,
                  base::CreatePostTaskAndReplyRelay(from_here, task, reply));

We could probably come up with an even better name than
CreatePostTaskAndReplyRelay.

[awong, 2011/08/17 18:46:46]

Good point...especially considering I *just* made a mistake with this little
fork.  Grr.

I changed the code to call the version on MessageLoopProxy.  Trying to extract a
function out of this is ridiculously painful cause you have to ensure that

 1) relay is new-ed on the right thread
 2) you can delete relay on a failure.

+                             base::Unretained(relay)));
+}
+
+// static
 bool BrowserThread::GetCurrentThreadIdentifier(ID* identifier) {
   // We shouldn't use MessageLoop::current() since it uses LazyInstance which
   // may be deleted by ~AtExitManager when a WorkerPool thread calls this