Likely overflow and some logic question in `add_random_linear_form`

[vneiger]

At lines 707-724 of src/msolve/msolve.c, there is

      int j = 0;
      int32_t sum = 0;
      for (i = 2 * len_old; i < 2 * len_new; i += 2) {
          gens->random_linear_form[j] = ((int32_t)(rand()));

          while (gens->random_linear_form[j] == 0) {
              gens->random_linear_form[j] = ((int32_t)(rand()));
          }
          if (i < 2 * len_new - 1) {
              sum += nvars_old * abs(gens->random_linear_form[j]);
          } else {
              gens->random_linear_form[j] = sum;
          }
          mpz_set_si(*(gens->mpz_cfs[i]), gens->random_linear_form[j]);
          mpz_set_ui(*(gens->mpz_cfs[i + 1]), 1);

          j++;
      }

It seems that the line 716 (sum += ...) may overflow since we use 32 bit signed integers and we multiply a random one (likely close to 29-31 bits) by nvars_old.

Also, if I get the logic well, the else is never entered: i is an even integer less than 2 * len_new, so at all iterations i will be less than 2 * len_new - 1... isn't it? And in that case, the value of sum is actually never used(?).