[Bug tree-optimization/95663] static_cast checks for null even when the pointer is dereferenced

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95663

--- Comment #5 from rguenther at suse dot de <rguenther at suse dot de> ---
On Mon, 15 Jun 2020, redi at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95663
> 
> --- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
> Or to be more clear:
> 
> struct Large {
>   char pad[1024*1024];
>   int x;
> };
> 
> Large* p = 0;
> int i = p->x;

Sure, but this isn't the same if C++ mandates the static_cast
of a null evaluates to null and not the offset of the base class.
So what clang does is not unsafe but wrong since the offset is
missing and it returns Base1::x instead of Base2::y?

Note for getter and clang I see

_Z6getterP5Base2:                       # @_Z6getterP5Base2
        .cfi_startproc
# %bb.0:
        leaq    -4(%rdi), %rax
        testq   %rdi, %rdi
        cmoveq  %rdi, %rax
        movl    4(%rax), %eax
        retq

So either static_cast<Derived*>(base) should evaluate to zero
or not.  If it does then clang dereferences the wrong address
in the null case (but only in 'field').

So, what does C++ say here?