gmp 4.3.0: minor bug with gmp_version

[Paul Zimmermann]

>   Maybe GMP itself could provide a function doing that check?
>   
> Maybe.  But how?  Should we do this in mpz_init, mpq_init, mpf_init,
> mpf_init2?  Or should we add a gmp_init() and suggest that people use
> that?

I was thinking to a gmp_init() function.

> By "matches" you mean the major version is the same, the minor version is
> at least as large in gmp_version, and we ignore the patch level?

no, I mean all three numbers do match. Maybe gmp_init() can have arguments
that enable different matches.

Paul