Proposed resolution (approved by CWG 2026-02-06):

Change in 6.8.2 [intro.object] paragraph 10 as follows:

... The address of a subobject of zero size is the address of an unspecified byte of storage occupied by the complete object of that subobject.

Given two complete objects with overlapping lifetimes, a byte in the object representation of one may have the same address as a byte in the object representation of the other if

• one object is nested within the other or
• they are both potentially non-unique objects;

otherwise, such bytes have distinct addresses.

Consider:

struct A {
  char c;
  int i;
} a;

char d;

There is likely going to be padding between a.c and a.i. For lack of an explicit prohibition, the implementation could put d into that padding. This is undesirable.