Unless int is 16-bit. Code like this is potentially UB; you should use int32_t as the target.

You should use long, and don't ever assume it's exactly 32-bits. The fixed size types are often an overused crutch that hampers future portability.

There are no mainstream 16-bit int platforms. It's fine to know what you target.

The promotions that are really surprising are from uint64_t bitfields to int (because it's based on value representability).

  struct {
    uint64_t a : 33,
             b : 15;
  } s;
  // s.b gets "promoted" to int, s.a does not