yet another nfgaloisconj() bug

[Igor Schein <igor@txc.com>]

Hi,

well, I *think* it's a bug, at least:

? a=Pol([1,1,-20,-22,189,552,364,-1514,-3979,-3317,3500,20391,26003]);
? nfgaloisconj(nfinit(a),0)
[x, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]

Looks like it's wrong, which can be verified by nfgaloisconj(a,4).

This issue goes back at least as far as 2.0.11

Igor

P.S. It appears nfgaloisconj(,0) doesn't depend on working precision,
so something else might be going here.  Maybe because poldegree(a)>11?
But again, it doesn't look like files from galois distribution are
being read at all.