Re: ole32 : wrong check in compobj.c

[Paul Vriens]

On Mon, 2005-01-10 at 18:02, Robert Shearman wrote:
> Paul Vriens wrote:
> > Hi,
> >
> > while looking through the code I saw:
> >
> > if (model & COINIT_MULTITHREADED)
> >
> > this will never work as COINIT_MULTITHREADED = 0.
>
> Good catch.
> > The attached patch fixes the 3 occurrences of a check against COINIT_*.
> >
> > This patch however makes the COM_CreateApartment in compobj.c:
> >
> > 545   if (!(apt = COM_CurrentInfo()->apt))
> >    546   {
> >    547     apt = COM_CreateApartment(dwCoInit);
> >    548     if (!apt) return E_OUTOFMEMORY;
> >    549   }
> >
> > return E_OUTOFMEMORY, always (?). And that makes sure that we have
> > a mismatch between CoInitialize and CoUninitialize.
> >
> > So there's definitely something else wrong as well.
> >
> > Any idea ?
>
> The parameter to CoInitializeEx is of type COINIT and so the caller can also specify another flag like COINIT_DISABLE_OLE1DDE along with the apartment flag.
> So we either need to convert the values to booleans before we call the lower level COM_CreateApartment or we need to be more careful and change the comparisons to this:
> if ((dwCoInit & (COINIT_APARTMENTTHREADED|COINIT_MULTITHREADED) == COINIT_MULTITHREADED) ...
> > -
> >    if ((apt->model & COINIT_APARTMENTTHREADED) && apt->win) DestroyWindow(apt->win);
> > -
> >    if ((apt->model == COINIT_APARTMENTTHREADED) && apt->win) DestroyWindow(apt->win);
>
> This change can be simplified to just check for the non-null apt->win.
>
> Rob

Any volunteers for this stuff? It's way over my head already :-). 
I will try as much as I can, but I'm sure the real men have this nailed
in a sec.

Cheers,

Paul.