Issues #276, #1163 - introduce class IncomingParam

src/plugin/COM_Value.h

Index: src/plugin/COM_Value.h
===================================================================
--- a/src/plugin/COM_Value.h
+++ b/src/plugin/COM_Value.h
@@ -115,6 +115,27 @@
        */
       BSTR_Argument& operator=( BSTR_Argument&& ); // = delete
     };
+
+    /**
+     * Constructor class for BSTR [in] parameters arriving from our COM entry points.
+     *
+     * The life cycle of a BSTR value that comes to us as an [in] parameter is entirely managed by the caller.
+     * Thus this class simply derives from std::wstring and provides an appropriate constructor.
+     *
+     * This class is a narrow technique toward a larger goal of eliminating CComBSTR,
+     *   which had been used for these parameters.
+     * All the uses, however, had been for IDispatch entry points, which are all VARIANT.
+     * In all cases, the caller much first check the type of the VARIANT structure before constructing an instance.
+     * Perhaps better would be a class, say, wstring_Incoming_Param that took a VARIANT as a constructor argument
+     *   and converted it to a string.
+     * Such a class, however, is beyond the scope of the work that this class was a part of.
+     */
+    class Incoming_Param
+      : public std::wstring 
+    {
+    public:
+      Incoming_Param( BSTR b );

[sergei, 2014/07/28 09:24:31]

Instead of commenting all bad things here, I will say only the desired result.
Replace this class by the method
`std::wstring our-namespace::to_wstring(const BSTR value)`.

[Eric, 2014/07/29 19:53:08]

Oh, that's just awful. If there's one thing I've learned about dealing with COM
interface types, it's that conversion functions that ignore their lifecycle
issues are a Very Bad Idea. Using a generic name for a conversion function is
asking for trouble.

Whatever. This was supposed to be a temporary class anyway, as the comment very
clearly indicated. I've just rewritten the whole thing as I was going to get to
later.

+    };
   }
 }