Re: bnrL1: wrong values

[Igor Schein <igor@txc.com>]

On Wed, Mar 30, 2005 at 06:51:26PM +0200, Xavier-Francois Roblot wrote:
> On Tue, 2005-03-29 at 18:31 -0500, buromano@rcn.com wrote:
> > I'm getting the wrong values from bnrL1 in gp 2.2.9.  An example:
> >                        GP/PARI CALCULATOR Version 2.2.9 (alpha)
> >                    PowerPC running darwin (PPC kernel) 32-bit version
> >        compiled: Mar 29 2005, gcc-3.3 20030304 (Apple Computer, Inc. build 
> > 1495)
> >                     (readline v4.3 enabled, extended help available)
> > 
> > ? bnf = bnfinit(x^2 - 229);
> > ? bnr = bnrinit(bnf,1,1);
> > ? bnrL1(bnr)
> > %78 = [[2, 1.747618204302940723551535783 + 2.271919407 E-28*I], [2, 
> > 1.747618204302940723551535783 - 2.271919407 E-28*I], [1, 
> > -4.068697957776515962021319266]]
> > 
> > should be:
> >                        GP/PARI CALCULATOR Version 2.2.7 (alpha)
> >                   Power PC running darwin (PPC kernel) 32-bit version
> >        compiled: Mar 29 2005, gcc-3.3 20030304 (Apple Computer, Inc. build 
> > 1495)
> >                   (readline not compiled in, extended help available)
> > ? bnf = bnfinit(x^2 - 229);
> > ? bnr = bnrinit(bnf,1,1);
> > ? bnrL1(bnr)
> > [[2, 2.355454590844564506512740955 + 2.776790386 E-28*I], [2, 
> > 2.355454590844564506512740955 - 2.776790386 E-28*I], [1, 
> > -4.068697957776515962021319266]]
> > Am I missing something?
> 
> Hi, thanks for the bug report. I fixed the bug in the CVS version. It
> turns out the program was using a routine to compute values of
> L-functions specific for quadratic fields but only for the kind of
> extensions used to construct abelian extensions using Stark units (as in
> the functions quadhilbert or bnrstark).
> 
> Xavier

Well, we have a regression:

? bnrL1(bnrinit(bnfinit(y^2+6),1,1),0)
  *** bnrL1: exponent (expo) overflow

Thanks

Igor