??gaussred is wrong

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

Hello pari-dev,

In the ??gaussred help, it is said that the result will be 
upper-triangular but this is not true.  I suppose this happens when the 
given matrix is not (positive) definite.
The formula which is written in ??gaussred is correct though, if one 
replaces j>i by j!=i.  So I think the docs needs to clarified.

gp> qfgaussred([0,1;1,0])
%302 =
[1/2 1]

[-1 -1/2]