bug#12350: Composites identified as primes in factor.c (when HAVE_GMP)

[Jim Meyering]

Pádraig Brady wrote:
> On 10/07/2012 10:00 AM, Jim Meyering wrote:
>> Torbjorn Granlund wrote:
>>> Jim Meyering <address@hidden> writes:
>>>
>>>    How about this in place of the final sentence?
>>>
>>>                                                        The new program also
>>>        runs a deterministic primality test for each prime factor, not just
>>>        a probabilistic test.
>>>
>>> That's better, thanks.
>>
>> I pushed the actual bug fix (for the issue mentioned in the Subject)
>> long ago, so I'm closing this "issue".
>>
>> Regarding your upcoming improvements, please start a new thread
>> when you're ready to discuss them, so that your comments are not
>> lost in the volume of with this now-"done" bug report.
>
> A small amendment I'm going to push is not to rely on GMP5.
> GMP4 on my fedora 15 system doesn't have mpz_inits().
> i.e. support for initializing multiple variables at once.
>
> Patch is simple enough...

Hi Pádraig,

Thanks, but wouldn't that be a slight "pessimization"?
What do you think about providing the missing function instead?
Maybe not worth the hassle, but still, it would avoid adding those 12
in-function lines.  factor.c is already large and complex enough that
every little bit helps.

BTW, Fedora 15 passed "end of life" back in June.

> diff --git a/src/factor.c b/src/factor.c
> index 5bfbfdc..1857297 100644
> --- a/src/factor.c
> +++ b/src/factor.c
> @@ -1335,7 +1335,10 @@ mp_prime_p (mpz_t n)
>    if (mpz_cmp_ui (n, (long) FIRST_OMITTED_PRIME * FIRST_OMITTED_PRIME) < 0)
>      return true;
>
> -  mpz_inits (q, a, nm1, tmp, NULL);
> +  mpz_init (q);
> +  mpz_init (a);
> +  mpz_init (nm1);
> +  mpz_init (tmp);
...