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

[John W. Eaton]

On 27-Oct-2008, Gabriele Pannocchia wrote:

| what was OBJ in the three cases and the maximum constraint violation for
| the two cases based on glpk 4.29?

Here is what I see with the current development sources with glpk
4.29:

  octave:1> load example.dat
  octave:2> [x, OBJ, INFO, LAMBDA] = qp (x0, H, Q, A, B, LB, UB, A_LB, A_IN, A_UB);
  octave:3> OBJ
  OBJ =  3.7723e+14
  octave:4> max (abs (A*x - B))
  ans =  1.3009e-06
  octave:5> idx = x < LB; max (LB(idx) - x(idx))
  ans =  1.4049e+09
  octave:6> tmp = A_IN*x; idx = tmp < A_LB; max (A_LB(idx) - tmp(idx))
  ans =  1.9927e-06
  octave:7> tmp = A_IN*x; idx = tmp > A_UB; max (tmp(idx) - A_UB(idx))
  ans =  1.0487e+09

and here is what I see with 3.0.1 and glpk 4.29:

  octave:1> load example.dat
  octave:2> [x, OBJ, INFO, LAMBDA] = qp (x0, H, Q, A, B, LB, UB, A_LB, A_IN, A_UB);
  octave:3> OBJ
  OBJ =  3.7723e+14
  octave:4> max (abs (A*x - B))
  ans =  1.3009e-06
  octave:5> idx = x < LB; max (LB(idx) - x(idx))
  ans =  1.4049e+09
  octave:6> tmp = A_IN*x; idx = tmp < A_LB; max (A_LB(idx) - tmp(idx))
  ans =  1.9927e-06
  octave:7> tmp = A_IN*x; idx = tmp > A_UB; max (tmp(idx) - A_UB(idx))
  ans =  1.0487e+09

With the current development sources and glpk 4.32, I see this:

  octave:1> load example.dat
  octave:2> [x, OBJ, INFO, LAMBDA] = qp (x0, H, Q, A, B, LB, UB, A_LB, A_IN, A_UB);
  octave:3> OBJ
  OBJ =  2.8942e-06
  octave:4> max (abs (A*x - B))
  ans =  1.2060e-11
  octave:5> idx = x < LB; max (LB(idx) - x(idx))
  ans =  3.6306e-12
  octave:6> tmp = A_IN*x; idx = tmp < A_LB; max (A_LB(idx) - tmp(idx))
  ans =  1.3049e-11
  octave:7> tmp = A_IN*x; idx = tmp > A_UB; max (tmp(idx) - A_UB(idx))
  ans =  1.0958e-13

If there is something else you'd like to see, then please send the
expressions you like for me to evaluate.

jwe