Discussion related to cado-nfs

[Pierrick Gaudry <pierrick.gaudry@loria.fr>]

On Wed, Jul 30, 2025 at 11:25:00AM +0200, hermann@stamm-wilbrandt.de wrote:
> In another posting Paul Zimmermann mentioned:
> "... but as you know, users never read the documentation."
> 
> I tried to find cado-nfs doc, is there more than README.md?
> https://gitlab.inria.fr/cado-nfs/cado-nfs

No. Talking about "documentation of Cado-nfs" is kind of a joke...

> In that I found no requirement for the to be factored number to be
> semiprime. Is it correct that cado-nfs should work for non-semiprimes as
> well?

Yes. It should.

> If so, then it has a bug for 118 decimal digits square of a semiprime, see
> log further below. Factoring RSA-120 on my PC does take 28min:
> https://github.com/Hermann-SW/RSA_numbers_factored/tree/main/cado-nfs#factoring-rsa-x
> 
> I stopped the computation after 1.5h because it did hang.

Indeed, you have hit a bug.

Thing is, if the input number is a perfect power, then the logics of
Cado-nfs goes crazy, because the sqrt binary has some sanity checks
against that case, while the calling python code does not the proper
cleaning of the input.

The situation is actually a bit worse: if the input number is a perfect
power multiplied by a smooth cofactor (only primes less than 1000000),
then the same problem occurs (I didn't test, but this is what I guess
from the code).

I'll write an issue, with a few more details.

Regards,
Pierrick