[mojo-edk] Fix uninitialized reads in EDK dispatchers

mojo/edk/system/message_pipe_dispatcher.cc

 // Copyright 2015 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "mojo/edk/system/message_pipe_dispatcher.h"
 
 #include <limits>
 #include <memory>
 
 #include "base/macros.h"
@@ skipping 481 matching lines @@
   base::AutoLock lock(signal_lock_);
 
   // We stop observing our port as soon as it's transferred, but this can race
   // with events which are raised right before that happens. This is fine to
   // ignore.
   if (port_transferred_)
     return;
 
 #if !defined(NDEBUG)
   ports::PortStatus port_status;
-  node_controller_->node()->GetStatus(port_, &port_status);
+  if (node_controller_->node()->GetStatus(port_, &port_status) != ports::OK)

[Anand Mistry (off Chromium), 2016/05/03 05:51:54]

Now debug and non-debug builds have different semantics (because of the awake
below the #if), which is bad for testing. Either zero-initalise port_status, or
encapsulate the entire block below in a condition.

[Ken Rockot(use gerrit already), 2016/05/03 06:16:15]

Good catch, fixed

+    return;
+
   if (port_status.has_messages) {
     ports::ScopedMessage unused;
     size_t message_size = 0;
     node_controller_->node()->GetMessageIf(
         port_, [&message_size](const ports::Message& message) {
           message_size = message.num_payload_bytes();
           return false;
         }, &unused);
     DVLOG(1) << "New message detected on message pipe " << pipe_id_
              << " endpoint " << endpoint_ << " [port=" << port_.name()
              << "; size=" << message_size << "]";
   }
   if (port_status.peer_closed) {
     DVLOG(1) << "Peer closure detected on message pipe " << pipe_id_
              << " endpoint " << endpoint_ << " [port=" << port_.name() << "]";
   }
 #endif
 
   awakables_.AwakeForStateChange(GetHandleSignalsStateNoLock());
 }
 
 }  // namespace edk
 }  // namespace mojo