nfrootsof1(bnf) vs. bnf.tu

[Jeroen Demeyer <jdemeyer@cage.ugent.be>]

Hello pari-users,

I somehow always assumed that the unit given by nfrootsof1(bnf) would be 
the same as bnf.tu

I now realize that this assumption is false. However, it does seem that 
PARI-2.9.1 has made this assumption much more false than it was in 
PARI-2.8.0.

With PARI-2.9.1 the result of nfrootsof1() seems random (it can be 
different with every call). With PARI-2.8.0, I could only find 
counter-examples to nfrootsof1(bnf) == bnf.tu in degree >= 16.

Do you consider this a bug? It would be convenient if nfrootsof1(bnf) 
would be the same as bnf.tu, but I don't know how much effort it would 
take to enforce that.

Cheers,
Jeroen.