Fix __UINTPTR_TYPE__ definition after r211657

[Joerg Sonnenberger]

On Sun, Jul 13, 2014 at 04:42:13PM -0700, Richard Smith wrote:
> On Sun, Jul 13, 2014 at 3:10 PM, Joerg Sonnenberger <joerg at britannica.bec.de
> > wrote:
> 
> > Hi all,
> > the attached patch ensures that __INTPTR_TYPE__ and __UINTPTR_TYPE__ are
> > consistent as the existing logic would pick e.g. long and unsigned int
> > on ARM. One issue I found during this is that the NVPTX backend is
> > using unsigned types here, which doesn't make sense to me.
> >
> 
> The change to SemaOverload.cpp looks unrelated? Other than that, this LGTM,

Yeah, test hack left-over in the tree.

> although I'm a little concerned that it might be intptr_t we were getting
> wrong on some of those targets rather than uintptr_t. Have you checked that
> this gives the same results as GCC?

That's why I CC'ed the owner of the NVPTX backend. I don't know if this
is really the desired behavior or not.

Joerg