tests/mpq/t-get_d.c failure on m68k-linux

[Niels Möller]

nisse at lysator.liu.se (Niels Möller) writes:

> On the other hand, maybe the test ought to be rewritten. Using mpz_get_d
> seems unnecessary.

Below is a rewritten check_random (in tests/mpq/t-get_d.c), with fewer
floating point operations. Works for me (but fails if I increase the
size of A to 54 bits). Can you check if it solves the problem?

Regards,
/Niels

void
check_random (int argc, char **argv)
{
  gmp_randstate_ptr rands = RANDS;
  
  double d;
  mpq_t q;
  mpz_t a, t;
  int exp;

  int test, reps = 100000;
  int i;

  if (argc == 2)
     reps = 100 * atoi (argv[1]);

  mpq_init (q);
  mpz_init (a);
  mpz_init (t);
  
  for (test = 0; test < reps; test++)
    {
      mpz_rrandomb (a, rands, 53);
      mpz_urandomb (t, rands, 32);
      exp = mpz_get_ui (t) % (2*MAXEXP) - MAXEXP;

      d = my_ldexp (mpz_get_d (a), exp);
      mpq_set_d (q, d);
      /* Check that n/d = a * 2^exp, or
	 d*a 2^{exp} = n */
      mpz_mul (t, a, mpq_denref (q));
      if (exp > 0)
	mpz_mul_2exp (t, t, exp);
      else
	{
	  if (!mpz_divisible_2exp_p (t, -exp))
	    goto fail;
	  mpz_div_2exp (t, t, -exp);
	}
      if (mpz_cmp (t, mpq_numref (q)) != 0)
	{
	fail:
	  printf ("ERROR (check_random test %d): bad mpq_set_d results\n", test);
	  printf ("%.16g\n", d);
	  gmp_printf ("%Qd\n", q);
	  abort ();
	}
    }
  mpq_clear (q);
  mpz_clear (t);
  mpz_clear (a);
}

-- 
Niels Möller. PGP-encrypted email is preferred. Keyid C0B98E26.
Internet email is subject to wholesale government surveillance.