Is it safe to cast pointer to integer, increment that integer, and cast back?

Question

Suppose I have a valid pointer p0:

T a[10];
T* p0 = &a[0];

I know that I can safely round-trip-cast it like this:

reinterpret_cast<T*>(reinterpret_cast<uintptr_t>(p0)) == p0;

But is it safe to do the following?

T* p1 = reinterpret_cast<T*>(reinterpret_cast<uintptr_t>(p0) + sizeof(T));

i.e. can I be sure that there is no UB and that p1 == &a[1]?

What do you mean by "can I be sure that there is no unsigned byte"? *Where*? — theV0ID, May 21 '15 at 09:45
@theV0ID UB is a popular acronym in C++ world for Undefined Behavior. — bashrc, May 21 '15 at 09:46
I'm pretty sure that'll fail on word-addressed systems, where incrementing an address by `n` makes it point `n*wordsize` bytes further. — user2357112, May 21 '15 at 09:46
duplicate: http://stackoverflow.com/questions/22624472/can-stduintptr-t-be-used-to-avoid-undefined-behavior-of-out-of-bounds-pointer — Thomas B., May 21 '15 at 09:58

score 3 · Accepted Answer · answered May 21 '15 at 09:56

This is implementation-defined behaviour. Your compiler should document whether or not pointer arithmetic is equivalent to integer arithmetic on the converted numeric values of pointers. That should be the case on modern computers with a "flat" byte-addressed memory space; but it's not guaranteed to work portably on all platforms.

Using char* rather than uintptr_t will work portably, as long as you stay within the array and ensure the pointer is correctly aligned for T before converting back.

Is it safe to cast pointer to integer, increment that integer, and cast back?

1 Answers1