Makes waiting SSLConnectJobs use the message loops to resume their connection.

net/socket/ssl_client_socket_openssl.cc

Index: net/socket/ssl_client_socket_openssl.cc
diff --git a/net/socket/ssl_client_socket_openssl.cc b/net/socket/ssl_client_socket_openssl.cc
index 06b57ba23dcfa29ba62d0936183b465d2a2720e0..dd1083efb2567c35f4bc8238319ab79715168646 100644
--- a/net/socket/ssl_client_socket_openssl.cc
+++ b/net/socket/ssl_client_socket_openssl.cc
@@ -16,6 +16,7 @@
 #include "base/memory/singleton.h"
 #include "base/metrics/histogram.h"
 #include "base/synchronization/lock.h"
+#include "base/thread_task_runner_handle.h"
 #include "crypto/ec_private_key.h"
 #include "crypto/openssl_util.h"
 #include "net/base/net_errors.h"
@@ -395,7 +396,7 @@ void SSLClientSocketOpenSSL::SetIsLeader() {
 
 void SSLClientSocketOpenSSL::OnSocketFailure() {
   if (is_leader_) {
-    error_callback_.Run();
+    base::ThreadTaskRunnerHandle::Get()->PostTask(FROM_HERE, error_callback_);

[wtc, 2014/07/11 18:55:51]

IMPORTANT: Based on my understanding of the code, I suggest running this error
callback directly. This will allow us to remove the comment for
SetSocketFailureCallback, which would introduce a behavior uncommon in the
various Socket interfaces of our network stack:

 // NOTE: This callback may still be executed even if this socket has
 // been disconnected or deleted.

The error callback (SSLConnectJobMessenger::OnJobFailed) already posts a task.
So posting a task here would be redundant.

[mshelley, 2014/07/14 20:30:06]

Done.

     error_callback_ = base::Closure();
     is_leader_ = false;
   }