Windows specific changes. All tests passing on Windows.

src/GlobalJsObject.cpp

Index: src/GlobalJsObject.cpp
===================================================================
--- a/src/GlobalJsObject.cpp
+++ b/src/GlobalJsObject.cpp
@@ -50,7 +50,8 @@
       Sleep(delay);
       const v8::Locker locker(isolate);
       const v8::HandleScope handleScope;
-      function->Call(function, functionArguments.size(), &functionArguments[0]);
+      v8::Handle<v8::Value> argv = functionArguments.size() > 0 ? functionArguments[0] : v8::Handle<v8::Value>();
+      function->Call(function, functionArguments.size(), &argv);

[Wladimir Palant, 2013/04/09 05:47:07]

No, this isn't right. You *copy* the first argument and then use the pointer to
this copy. What if there is more than one argument?

Ideally we would use functionArguments.data() here but that's C++11 again. So if
you really want to handle empty vectors properly I would go with:

  v8::Handle<v8::Value> *argv = functionArguments.empty() ? 0 :
&(functionArguments.front());

[Oleksandr, 2013/04/09 05:54:21]

Ah. You're right. My bad. 

It's not that I *wanted* to explicitly handle the empty vectors case, its just
that the test fails otherwise.

 
On 2013/04/09 05:47:07, Wladimir Palant wrote:

[Felix Dahlke, 2013/04/09 08:18:06]

Ah, just saw the original change here, didn't show up in the other diff so I
thought you had just replaced size() with empty() and the copy with a pointer.

We can actually just pass 0 here if it's empty, so I vote for:

function->Call(function, functionArguments.size(), functionArguments.size() ?
functionArguments[0] : 0);

       delete this;
     }