gmp

[Stephen Weeks]

> I want to implement their gcd since that is required for all the uses I
> want.

The Polyspace guys have already done this, and I will port their
changes in soon.

> re compiling gmp for the machine you are running on, this is the big argument
> in favor of using shared libraries.  The big disadvantage is that  then  your
> program won't work at all on machines that don't have any copy of gmp.

There are two separate issues, right?

1. Whether or not we ship gmp with mlton (source or binary rpm)
2. Whether or not MLton uses libgmp.a or libgmp.so

As to 1, I propose to completely eliminate gmp from our rpms.  Most
people have it, and if not they can get it.  As to 2, I propose to
stick with our current practice of linking with libgmp.a, but I don't
really care one way or the other.