Re: drivers/infiniband/hw/qib/qib_sysfs.c:413:1: error: static_assert expression is not an integral constant expression

[Jason Gunthorpe <jgg@nvidia.com>]

On Mon, Sep 13, 2021 at 10:00:02AM -0700, Nick Desaulniers wrote:

> > This macro would like to know that the passed in member name has a u64
> > type, all the things I've come up with fail on clang - but many work
> > fine on gcc. Frankly I think this case is a clang bug myself..
> 
> Perhaps, though this assertion looks a bit like offsetof() to me. I
> wonder if that can help here?

The assertion would logically like to be this:

         static_assert(typecheck(((struct qib_port *)0)->N, u64))

Which doesn't compile because typecheck is not a constexpr :\

typecheck also can't be used as a build bug on zero in the initializer
for the same reason.

My original attempt was

                .counter = &((struct qib_ibport *)0)->rvp.n_##N - (u64 *)0,    \

Which is fairly simple opencoding of offsetof_end but clang whines
overly pedantically that NULL subtraction is undefined behavior.

The current version is this:

       static_assert(&((struct qib_ibport *)0)->rvp.n_##N != (u64 *)NULL);

Which *should* be perfectly fine, but clang explodes for some reason
complaining about -> on NULL. I think it is broken and doesn't
understand that this -> is not an actual deref but pointer/type logic,
much like this:

#define sizeof_field(TYPE, MEMBER) sizeof((((TYPE *)0)->MEMBER))

Which does work.

So to my mind clang is being buggy, and I'm probably going to just
delete the line and a give up on type checking here unless someone has
a better idea.

Jason