Re: factor inconsistencies/limits?

[Jim Meyering]

"James Youngman" <address@hidden> wrote:
> On Fri, Jul 25, 2008 at 11:57 AM, Jim Meyering <address@hidden> wrote:
>> Actually, I've wanted coreutils' factor program to work
>> with arbitrary-precision numbers for a long time.
>> This is mentioned briefly in TODO.  For more detail, see these
>> search results:
>>
>>  
>> http://search.gmane.org/search.php?group=gmane.comp.gnu.coreutils.bugs&query=factor+gmp
>
> I have a working GMP-based version (essentially, it's the example
...

Great!
Thanks for working on that.

> Given these performance characteristics, I would not be adverse to
> using the GMP implementation for all sizes of input.  However, there
> is also the maintenance consideration.   We must maintain a non-GMP
> implementation for systems which lack GMP (or decline to use it).  I
> would in general be concerned about bit-rot in the non-GMP
> implementation, so I guess it's worth using it unconditionally for
> some subset of inputs in order to be sure it really works for those
> who must use it.

I don't think we need to worry much about bit rot...
if we add an option that will make factor use the
64-bit-limited code, and maybe another to use the GMP-based
version (or fail if it's not available).
Then it'd be easy to test both via "make check".