Karim Belabas on Fri, 26 Jul 2013 18:17:15 +0200


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: Bug in bnfcertify?


* Georgi Guninski [2013-07-26 16:50]:
> Can bnfcertify return 0 even after interventions with debugger?
> 
> Last time I checked the source couldn't see codepath returning 0/raising
> exception, even if doing stuff with debugger.
> 
> What will bnfcertify return in the unlikely event GRH fails?

See ??bnfcertify:

  [...] It is correct if and only if the answer is 1.  If it is incorrect, the
  program may output some error message, or loop indefinitely.

Current implementation will not return if either GRH fails *OR* the regulator 
and class group were incorrectly computed: we go on trying to certify primes
that simply can't be certified. 

N.B. "Certifying a prime" means one of two things depending on the computation
stage : proving that all P | p belong to the subgroup of the class group
defined by the known generators, or proving that p does not divide the index of
known units in the full unit group. Both stages can lead to an infinite loop.

Cheers,

    K.B.

P.S. The part about the error message is currently a little misleading. The
only exception that can be raised is an OVERFLOW if our initial upper bound for
(Tentative Regulator) / (True Regulator) is too large (> 2^64, on 64-bit archs)

--
Karim Belabas, IMB (UMR 5251)  Tel: (+33) (0)5 40 00 26 17
Universite Bordeaux 1          Fax: (+33) (0)5 40 00 69 50
351, cours de la Liberation    http://www.math.u-bordeaux1.fr/~kbelabas/
F-33405 Talence (France)       http://pari.math.u-bordeaux1.fr/  [PARI/GP]
`