SVD Bug Continue (Bug #2027)

For floats SVD returns wron results ( as off [[http://code.opencv.org/issues/1448]] )

It seems that this is not totally resolved.
See the file in the attach.

If the matrix is CV_32FC1 sometimes the SVD results are incorrect. In this example the matrix U has some values with the wrong sign, and also it has NaN and inf in some elements. The matrix Vt also has wrong sign in some elements.

If we change all the matrix to CV_64FC1 then the results are accurate.

Opencv - Matrix CV_32FC1

Q=
[0.97855008, 1.2548437, 0.1352066;
  0.55914873, 0.71705663, 0.078065962;
  0, 0, 0]
U=
[0.86822981, -0.49616224, -inf;
  0.49616224, 0.86822981, -nan;
  0, 0, -nan]
W=
[1.839398; 0.00069751492; 1.4012985e-45]
Vt=
[0.61271936, 0.78572941, 0.084877647;
  -0.071619853, -0.05175164, 0.9960885;
  0.78704864, -0.61640161, 0.024564598]

Matlab/Octave

Q =

   0.97855   1.25484   0.13521
   0.55915   0.71706   0.07807
   0.00000   0.00000   0.00000

U =

  -0.86823  -0.49616   0.00000
  -0.49616   0.86823   0.00000
   0.00000   0.00000   1.00000

S =

Diagonal Matrix

   1.83940         0         0
         0   0.00070         0
         0         0   0.00000

V =

  -0.612719  -0.071656   0.787045
  -0.785729  -0.051724  -0.616404
  -0.084878   0.996087   0.024610

This happens with the most recent build (booth branch 2.4.1 and trunk 2.4.9 )
Os: Ubuntu 11.10