[Octave-bug-tracker] [bug #40436] median no longer passes make check

[Mike Miller]

Follow-up Comment #26, bug #40436 (project octave):

I've traced this down to liboctave Array::nth_element, here's an example using
the call to nth_element that the failing median test case makes. It does look
like uninitialized memory, because it's random which element of this result is
wrong, and it doesn't happen for me every time.


octave-cli:1> b = rand (3, 4, 6, 5);
octave-cli:2> tmp = nth_element (b, 3:4, 3)
tmp =

ans(:,:,1,1) =

   0.72585   0.35493   0.89880   0.37920
   0.75690   0.62664   0.54848   0.12803
   0.38214   0.57314   0.23438   0.86337

ans(:,:,2,1) =

   0.90676   0.63536   0.92257   0.53978
   0.79579   0.73971   0.73077   0.68045
   0.67474   0.64598   0.59156   0.95027

ans(:,:,1,2) =

  -1.4411e+296    2.5586e-01    4.0583e-01    3.5271e-01
    4.8441e-01    3.8306e-01    4.3900e-01    2.4635e-01
    3.7203e-01    3.5613e-01    6.8720e-01  -7.4608e+232

ans(:,:,2,2) =

  -3.2079e+176    2.5827e-01    5.8781e-01    6.4392e-01
    6.6424e-01    5.7252e-01    5.2495e-01    3.6968e-01
    5.1522e-01    4.7913e-01    8.9407e-01  -1.3622e-138

ans(:,:,1,3) =

   0.46609   0.48197   0.47079   0.35431
   0.36301   0.31437   0.54804   0.62638
   0.40946   0.48275   0.61334   0.19517

ans(:,:,2,3) =

   0.70109   0.56286   0.69651   0.44498
   0.54481   0.39687   0.65335   0.75202
   0.83503   0.61924   0.90139   0.72982

ans(:,:,1,4) =

   0.51566   0.17927   0.60041   0.44344
   0.56499   0.00000   0.50164   0.21559
   0.21606   0.26750   0.39789   0.49834

ans(:,:,2,4) =

   0.55074   0.18725   0.61739   0.61690
   0.56823   0.00000   0.86750   0.34780
   0.34628   0.39065   0.41223   0.54986

ans(:,:,1,5) =

   0.22879   0.11877   0.34887   0.64783
   0.50557   0.43964   0.00000   0.44355
   0.59332   0.67293   0.55430   0.36468

ans(:,:,2,5) =

   0.56224   0.54101   0.51298   0.68682
   0.57307   0.51974   0.21241   0.61531
   0.59603   0.74638   0.82469   0.50562



    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?40436>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/