Discussion related to cado-nfs

[Robert Balfour <rhb11931@gmail.com>]

Dear Emmanuel,

Thank you for fixing the sopt bug!

Here is a link to c100.sparse.bin: https://drive.google.com/file/d/1qroDYM_VsVhveP2YgI1Q8c8vbi0octbt/view?usp=sharing

Best regards,

Robert

On Tue, Aug 30, 2022 at 9:21 AM Emmanuel Thomé <Emmanuel.Thome@inria.fr> wrote:

Hi,

Thanks for your feedback.

On Mon, Aug 15, 2022 at 12:44:43AM +0100, Robert Balfour wrote:
> Dear all,
>
> I have two potential bugs to report.
>
> Bug 1: as of the latest revision of CADO, running "sopt" on a file of
> polynomials will run size-optimization repeatedly on the first polynomial
> in the file, instead of progressing through the file. This bug was
> introduced at some point since last November.

Ooops. I filed it as
https://gitlab.inria.fr/cado-nfs/cado-nfs/-/issues/30046 (which is now
fixed). This kind of bug should never have been let through with proper
testing, but as it turns out sopt had no testing...

> Bug 2: running the latest revision on RSA-100 with the default parameters
> (and -t 2, not sure if that is important) results in a crash at the end of
> lingen. This crash doesn't seem to occur on every number or matrix. The
> command line used was

This one sounds really annoying.

It might be related to #30041, but it's really just a random guess.

Could you please post the c100.sparse.bin file somewhere where I can
download it? Our tests here didn't yield a reproducer.

> The first obvious sign that something has gone wrong is the line
>
> Number of lucky columns: 0 (64 wanted)
>
> near the end of c100.bwc.stdout.1.

Yes indeed. This should really not happen, and the fact that the rest
crashes is just a consequence.

Best,

E.

>
> ./cado-nfs.py
> 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139
> workdir=/store/charybdis/cadojobs/tests/c100test_3ab6a478_tryagain -t 2
>
> The first obvious sign that something has gone wrong is the line
>
> Number of lucky columns: 0 (64 wanted)
>
> near the end of c100.bwc.stdout.1.
>
> This then produces
>
> ## [01;31mCheck for mksol [0m: Lingen output files missing
> ## [01;31mCheck for mksol [0m: can't find any F file, cannot infer the
> mksol max iteration at
> /store/charybdis/cado_3ab6a478/cado-nfs/build/tantalum/linalg/bwc/bwc.pl
> line 1499.
> ## [01;31mCheck for mksol [0m: Please run lingen first.
>
> in c100.bwc.stderr.1, at which point CADO quits.
>
> Strangely enough, if CADO is resumed then it finds the 64 lucky columns
> it needs and the factorization completes successfully.
>
> The relevant log files are attached: stdout.1 and stderr.1 are from the
> original run that crashed, stdout.2 and stderr.2 are from the resumption.
>
> Best regards,
> Robert