Re: Behavior of out parameters and dependencies

• Previous message: Alexander Nickolov: "Re: Behavior of out parameters and dependencies"
• In reply to: Alexander Nickolov: "Re: Behavior of out parameters and dependencies"
• Next in thread: Alexander Nickolov: "Re: Behavior of out parameters and dependencies"
• Reply: Alexander Nickolov: "Re: Behavior of out parameters and dependencies"
• Messages sorted by: [ date ] [ thread ]

Date: Thu, 20 Jan 2005 12:49:03 -0800

Thanks a lot. This will save some bugs later in the project.
In most cases it will be ok to just declare the pointers as ref, but in
scenarios it can be usefull to pass null parameters which depends on each
other, this seems to be the perfect solution. In my case I will do my stuff
in the ..._Stub method.

Thanks.
- Laurents C. R. Meyer

"Alexander Nickolov" wrote:

> This is one of the finer points of marshaling. You need the [local]
> and [call_as] atributes here:
>
> [pointer_default(unique), ...]
> interface IFoo : IUnknown
> {
> [local]
> HRESULT GetData( DWORD* pdwSize, BYTE** ppBuffer );
> [call_as(GetData)]
> HRESULT RemGetData( [out]DWORD* pdwSize, [out, size_is(
> ,*pdwSize)]BYTE** ppBuffer );
> };
>
> Then you need to write a small C file to link with the rest of the
> proxy/stub code that does the conversion from the local method
> into the remote method at the proxy, and back at the stub. Your
> clients always call the local method, and that's what you implement
> at the server. However, it's always the remote method that travels
> on the wire. Of course, you need to modify the proxy/stub DLL
> project to compile and link your file as well.
>
> HRESULT WINAPI IFoo_GetData_Proxy(IFoo* pThis, DWORD* pdwSize, BYTE**
> ppBuffer)
> {
> if (IsBadWritePtr(pdwSize, sizeof(DWORD*)) || IsBadWritePtr(ppBuffer,
> sizeof(BYTE**))) {
> return S_FALSE;
> }
> return IFoo_RemGetData_Proxy(pThis, pdwSize, ppBuffer);
> }
>
> HRESULT WINAPI IFoo_RemGetData_Stub(IFoo* pThis, DWORD* pdwSize, BYTE**
> ppBuffer)
> {
> return pThis->lpVtbl->GetData(pThis, pdwSize, ppBuffer);
> }
>
> Note the docs for [call_as] is wrong on numerous points. The
> needed prototypes are declared in the header file produced
> by MIDL.
>
> --
> =====================================
> Alexander Nickolov
> Microsoft MVP [VC], MCSD
> email: agnickolov@mvps.org
> MVP VC FAQ: http://www.mvps.org/vcfaq
> =====================================
>
> "Laurents C. R. Meyer" <Laurents C. R. Meyer@discussions.microsoft.com>
> wrote in message news:CFE5270F-EAC9-433A-A95F-13C4C7ABBC70@microsoft.com...
> > Hoi!
> > I am currently thinking about an interesting exception in the COM usage
> > and
> > would like to hear if somebody knows about the behavior.
> >
> > Let's say there is Some Interface called IFoo. This interface offers one
> > method called GetData. Because the caller does not know how large the
> > buffer
> > has to be, it is allocated by GetData using CoTaskMemAlloc. The user has
> > the
> > possibility to omit a parameter, which results in an return code of
> > S_FALSE.
> > The declaration of the interface would look something like this:
> >
> > [pointer_default(unique), ...]
> > interface IFoo : IUnknown
> > {
> > HRESULT GetData( [out]DWORD* pdwSize, [out, size_is( ,*pdwSize)]BYTE**
> > ppBuffer );
> > };
> >
> > Now, on a usual call like:
> >
> > DWORD dwSize;
> > BYTE* pData;
> > GetData( &dwSize, &pData );
> >
> > everything will be alright.
> > But what is the behavior, if I call like this:
> >
> > BYTE* pData;
> > GetData( NULL, &pData );
> >
> > Even if GetData returns immediately after a paremeter check (allocates no
> > memory for ppBuffer), COM would try to dereference pdwSize (i.e.
> > *pdwSize),
> > which would result in an error.
> > Am I right, or is COM checking the pointer (pdwSize) first if it is zero
> > of
> > not and in case it is zero assumes the dereferenced pointer (*pdwSize) is
> > zero too?
> >
> > In case COM returns an error, is it one which can be handled or is COM
> > terminating the application?
> > Of course this is only a point when using a proxy.
> >
> > Can anybody help me? Thanks in advance.
> >
> > Laurents C. R. Meyer
>
>
>

• Previous message: Alexander Nickolov: "Re: Behavior of out parameters and dependencies"
• In reply to: Alexander Nickolov: "Re: Behavior of out parameters and dependencies"
• Next in thread: Alexander Nickolov: "Re: Behavior of out parameters and dependencies"
• Reply: Alexander Nickolov: "Re: Behavior of out parameters and dependencies"
• Messages sorted by: [ date ] [ thread ]