> I have VB and C++ client apps using a C++ ATL/COM object that has some
> initializing code in its in its constructor.  I want to be able to throw an
> exception from the COM object's constructor if something goes wrong during
> the initialization process.  If I call _com_raise_error(S_FALSE,
> pErrorInfo), my VB client returns the string "Out of memory" when I call
> Err.GetDescription().  This string "Out of memory" is different from the
> string that I used to initialize the pErrorInfo object passed to
> _com_raise_error.  How can I get my custom error messages across to the VB
> (or C++) client when throwing an exception from the constructor of a COM
> object?

> I realize that I could create an Init method on the COM object and have all
> the initializing done in there, and call return Error(...) if an exception
> needs to be thrown to the higher scope.  But I am wondering if there is a
> slick way to accomplish the same from the constructor without creating a new
> method.

> Shouldn't this sort of thing go in FinalConstruct, which returns an
> HRESULT to indicate success or failure?

> > Shouldn't this sort of thing go in FinalConstruct, which returns an
> > HRESULT to indicate success or failure?

> I looked up the FinalConstruct() method and found that it is only available
> on the base classes CComAggObject, CComCachedTearOffObject, and
> CComPolyObject.  The COM object I am working with is a regular COM
> object....

> No, that's not the case, although the way the cross-referencing in the
> documentation is set up does lead you to think that.  The root
> implementation is located in CComObjectRootBase, which is the base
> class of CComObjectRootEx.  The documentation for CComObjectRootEx does
> include FinalConstruct, although it only discusses using it for
> creating aggregates.

}

> > No, that's not the case, although the way the cross-referencing in the
> > documentation is set up does lead you to think that.  The root
> > implementation is located in CComObjectRootBase, which is the base
> > class of CComObjectRootEx.  The documentation for CComObjectRootEx does
> > include FinalConstruct, although it only discusses using it for
> > creating aggregates.

> Thanks - you're right.  I tried it, and the overridden FinalConstruct
method
> does indeed get called.  In that method, I only have one line of code:

> HRESULT MyObj::FinalConstruct()
> {
>   return Error("my error msg");
> }

> On the VB client, when I try to create MyObj, I get an error, and
> Err.Description returns "Automation Error" and not "my error msg" as I was
> expecting.

> Also, I put a breakpoint in MyObj::MyObj() which has no source code in it,
> and a breakpoint in the FinalConstruct method shown above.  I was
expecting
> each of these breakpoints to only be hit once each, but both these
> breakpoints were hit 4 times each even though the VB client only has one
> line of code that reads like:

> set pObj = new MyObj

> Any thoughts on what I may be doing wrong and how I can pass an
> exception/error string to the VB client during the construction of a COM
> object?