nfsubfields() bug - serious

[Igor Schein <igor@txc.com>]

Hi,

I came across this behavior, first visible in 2.0.15:

? nf=nfinit(x^4+1);
? \g6
   debug = 6
? nfsubfields(nf);

***** Entree dans subfields

pol = x^4 + 1
dpol = 256
divisors = [1, 2, 4]

*** Recherche des sous-corps de degre 2

avma = 4445376, lg(Z) = 3, lg(Y) = 1, lg(vbs) = 0
Z = [[0,   ***   bus error: bug in GP (please report).

The object somehow get corrupted before entering calc_block(),
and then a debugging printf statement inside calc_block() activated
at \g6 or higher triggers that.

Thanks
Igor