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

[Gabriele Pannocchia]

> Might it make sense to add a check to qp and warn in the event the  
> problem is non-convex?

I think so. Although, checking all(eig(H)>0) could be somehow
conservative because if we have equality constraints, it would be more
appropriate to consider the projection Hessian, i.e.:

Z'*H*Z where Z = null(A)

Since the equality constraints will always be active any solution is in
the form:
x = x0 + Z*p
where X0 is a solution of A*x=b. If Z'*H*Z is positive definite, than it
is a convex QP.

I also think that a warning could be sent when maximum number of
iterations is reached. However, the users knows about that by looking at
INFO.info. 
John, what do you think about prompting this warning?

Gabriele