[Octave-bug-tracker] [bug #60357] Failing test for `qz` on Windows with LAPACK 3.9.1

[Markus Mützel]

Update of bug #60357 (project octave):

                Category:             Interpreter => Libraries              
                 Summary: Failing tests on Windows => Failing test for `qz` on
Windows with LAPACK 3.9.1

    _______________________________________________________

Follow-up Comment #1:

With a clean build from hg id d57c1d781093 and MXE Octave hg id 60a500c74ac7,
the error with "zip" disappeared.
A bunch of BISTs erroneously were marked as regressions. Should be fixed with
this change:
https://hg.savannah.gnu.org/hgweb/octave/rev/526a5ba817ca

The BIST in qz.cc-tst still fails.
Running the beginning of the BIST:

>> a = [ 10  1  2;
        1  2 -1;
        1  1  2];
>> b = reshape (1:9,3,3);
>> [aa, bb, q, z, v, w, lambda] = qz (a, b)
aa =

   4.4159   6.8422  -4.4675
        0  -4.6396   2.0919
        0        0   2.1964

bb =

    0.0000   -0.9650   -4.8929
         0    1.0902   10.8977
         0         0   11.8398

q =

   0.924500  -0.369800   0.092450
  -0.377574  -0.921698   0.088946
   0.052319  -0.117137  -0.991737

z =

   0.4082   0.8731  -0.2667
  -0.8165   0.2186  -0.5344
   0.4082  -0.4359  -0.8021

v =

   0.500000   0.856123   0.056123
  -1.000000   1.000000  -1.000000
   0.500000  -0.950851  -0.417518

w =

   0.500000  -0.424485   0.052755
  -1.000000  -0.811212  -0.118113
   0.500000   1.000000  -1.000000

lambda =

   3.4924e+15
  -4.2557e+00
   1.8551e-01


The same on Ubuntu 20.10 with the same hg id:

>> a = [ 10  1  2;
        1  2 -1;
        1  1  2];
>> b = reshape (1:9,3,3);
>> [aa, bb, q, z, v, w, lambda] = qz (a, b)
aa =

  -5.2673   0.0102  -8.0910
        0   1.9024   0.0690
        0        0   4.4907

bb =

    1.2377    9.8871   -5.7329
         0   10.2551   -6.9050
         0         0         0

q =

  -8.9548e-01  -4.4507e-01   5.3287e-03
   1.7735e-01  -3.6775e-01  -9.1286e-01
   4.0825e-01  -8.1650e-01   4.0825e-01

z =

   0.5272   0.2294   0.8182
   0.6158  -0.7666  -0.1818
  -0.5855  -0.5997   0.5455

v =

   0.856123   0.056123   0.500000
   1.000000  -1.000000  -1.000000
  -0.950851  -0.417518   0.500000

w =

  -0.424485   0.052755   0.500000
  -0.811212  -0.118113  -1.000000
   1.000000  -1.000000   0.500000

lambda =

  -4.2557
   0.1855

>> version -lapack
ans = Linear Algebra PACKage Version 3.9.0



That looks quite different to me.

If I drop in the "liblapack.dll" from the build from 5th of April (LAPACK
version 3.9.0), the results on Windows and Linux match.

Is this a bug in LAPACK 3.9.1? Or an error in our build rule in MXE Octave? Or
do we use those functions incorrectly?

CC'ing John D.


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?60357>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/