[Bug c++/110619] Dangling pointer returned from constexpr function converts in nullptr

["redi at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110619

--- Comment #9 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Fedor Chelnokov from comment #8)
> This program must be rejected because of relational comparison of unrelated
> pointers

I disagree. If GCC were to define the behaviour of returning invalid pointers
during constant evaluation as converting them to null pointers, then those
wouldn't be unrelated pointers. They'd be null pointers. And the comparison
would be well defined.

But I don't think that conversion was ever intended as the defined behaviour,
just an unintended consequence. If it was intended, it should have been
documented at
https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Implementation.html

The fact that we _don't_ do any such conversion should also be documented
there, of course. We should say that other uses of invalid pointers do not
modify the pointer values, so copying them preserves the invalid value (without
trapping, except on Itanium?) and pointer comparisons and pointer arithmetic
are undefined because they don't point to an object.