Re: yet another rnfkummer() posting

[Igor Schein <igor@txc.com>]

On Thu, Sep 05, 2002 at 01:17:45AM +0200, Karim BELABAS wrote:
> On Thu, 29 Aug 2002, Igor Schein wrote:
> > ? setrand(1);rnfkummer(bnrinit(bnfinit(quadpoly(-6580,y)),9,1),matdiagonal([3,1,1,1]));
> >   ***   not an Abelian extension in rnfnormgroup.
> >
> > It worked fine in 2.2.2.
> 
> I have been very busy fixing various things I had upset while cleaning up
> bnfinit & quadclassunit [clearing long standing internal TODO items (comments
> in the code) in hnfspec & real_random_form, replacing unnecessarily obscure
> code, removing spurious global variables; also the current version of
> quadclassunit is nearly twice faster for smallish discriminants, say ~ 10^6].
> 
> My (relatively) simple regression tests come clear now [and so do a few of
> the things you've reported, in particular the above bug]. Can you make a list
> of all surviving regressions from your archives ?
> 
> Thanks,

? bnfisnorm(bnfinit(polsubcyclo(7,3)),2);
  ***   inconsistent data in inverseimage.
? factorpadic(x^3-1,3,10,1);
  ***   not enough precalculated primes
? setrand(1);quadclassunit(-148-16!);
  ***   sorry, buchxxx couldn't deal with this field PLEASE REPORT!.
? setrand(2126042356);rnfkummer(bnrinit(bnfinit(quadpoly(3409,y),1),25,1),[5,3;0,1]);
  ***   division by zero in gdiv, gdivgs or ginv