qp() in Octave 3.0.0 returns result egregiously violating input constraints

[Gabriele Pannocchia]

> In any case, we know for the example problem that feasible solutions
> exist, so according to the current spec for qp() it must return one of
> them, even if it's not the global optimum.  One could change the spec
> to say it may not return feasible solutions, but it's probably
> better/easier to find the root problem - I hope.

qp() must return only feasible solutions or INFO.info = 6, i.e. no
feasible solution exists. Just to sketch how it works:
1) If an initial point is provided and it is feasible, go to 3.
Otherwise, 
2) Call glpk to solve an LP that tells us either there is no feasible
solution or a feasible starting point.
3) The iterations start and at each iteration the current X remains
feasible because it is in the form:
	x(k) = x(k-1) + Z_k*p
where Z_k is a basis matrix for the null space of the active constraints
(i.e. equality constraints and inequality constraints currently active).
If a new constraint becomes active it is inserted into the active set
matrix...

If feasibility is lost during the iterations, then there is a bug. It is
my understanding from previous messages that it is not strictly related
to qp.m and __qp__.cc but to linear algebra libraries.
Most likely it is related to the computation of the null space of the
active set matrix.

However, please let me know if you feel that something should be done in
qp.m or __qp__.cc.

Gabriele