Fixes for re-enabling more MSVC level 4 warnings: sandbox/ edition

sandbox/win/src/handle_dispatcher.cc

Index: sandbox/win/src/handle_dispatcher.cc
diff --git a/sandbox/win/src/handle_dispatcher.cc b/sandbox/win/src/handle_dispatcher.cc
index 6acb6f9ceb3029ad1c95456d4ee3c04a0c46e2aa..fda7aacaaef351f480f821c58753b07088a35be3 100644
--- a/sandbox/win/src/handle_dispatcher.cc
+++ b/sandbox/win/src/handle_dispatcher.cc
@@ -44,7 +44,6 @@ bool HandleDispatcher::DuplicateHandleProxy(IPCInfo* ipc,
                                             DWORD target_process_id,
                                             DWORD desired_access,
                                             DWORD options) {
-  NTSTATUS error;
   static NtQueryObject QueryObject = NULL;
   if (!QueryObject)
     ResolveNTFunctionPtr("NtQueryObject", &QueryObject);
@@ -65,9 +64,10 @@ bool HandleDispatcher::DuplicateHandleProxy(IPCInfo* ipc,
   OBJECT_TYPE_INFORMATION* type_info =
       reinterpret_cast<OBJECT_TYPE_INFORMATION*>(buffer);
   ULONG size = sizeof(buffer) - sizeof(wchar_t);
-  error = QueryObject(handle, ObjectTypeInformation, type_info, size, &size);
+  NTSTATUS error =
+      QueryObject(handle, ObjectTypeInformation, type_info, size, &size);
   if (!NT_SUCCESS(error)) {
-    ipc->return_info.win32_result = error;
+    ipc->return_info.nt_status = error;

[Peter Kasting, 2014/07/09 18:55:18]

Note that because win32_result and nt_status are in a union, this doesn't
actually result in the sort of behavior change it might look like.  (Similar
comments apply elsewhere in this CL; in general, changes like this are attempts
to make sure we read and write the fields that match the types of the local
variables we're reading and writing to.)

[cpu_(ooo_6.6-7.5), 2014/07/10 00:49:34]

yeah that reads like it was a bug, but it can become one easily if one would
remove the union which its an optimization if my memory is correct.

[Peter Kasting, 2014/07/10 01:19:31]

I don't know if I understand this comment.

As far as I can tell, it was already the case that we sometimes wrote one format
and then read the other.  In other words, the union is not just for space
savings, it's actually functionally necessary (it effectively casts types back
and forth).

Given that, I don't know how to interpret what you're saying I should be doing
here.

     return false;
   }
   type_info->Name.Buffer[type_info->Name.Length / sizeof(wchar_t)] = L'\0';