Possible resolution:

1. Change in 7.6.6 [expr.add] paragraph 5 as follows:

   The result of subtracting two pointer expressions P and Q is a prvalue of type std::ptrdiff_t (17.2.4 [support.types.layout]).

   • If P and Q both evaluate to null pointer values, the value is 0.
   • Otherwise, if P and Q point to, respectively, array elements i and j of the same array object x, the expression P - Q has the value i - j. [Note 2: If ; if the that value i - j is not in the range of representable values of type std::ptrdiff_t, the behavior is undefined (7.1 [expr.pre] [expr.sub.pointers.rep]). —end note]
   • Otherwise, the behavior is undefined (F.3.31 [ub:expr.add.sub.diff.pointers]).

2. Add a subclause before F.3.31 [ub:expr.add.sub.diff.pointers] as follows:

   [ub:expr.sub.pointers.rep]
   Specified in: 7.6.6 [expr.add]
   Subtracting two expressions of pointer type where the difference is not representable in std::ptrdiff_t is undefined behavior. [Example:

     char big[static_cast<std::uintmax_t>(PTRDIFF_MAX) + 1];
     std::ptrdiff_t d = (big + sizeof(big) - big);   // undefined behavior
   

   —end example]

(From submission #922.)

Paper P3899R3 (Clarify the behavior of floating-point overflow), adopted in June, 2026, accidentally removed undefined behavior for the case where the result of subtracting two pointer values is not representable as a std::ptrdiff_t. Undefined behavior was limited to arithmetic expressions, which pointer subtraction is not.