[Octave-bug-tracker] [bug #56232] Octave crash when inverting an empty sparse matrix.

[Rik]

Follow-up Comment #58, bug #56232 (project octave):

@Marco: This is looking good.

On the question of inv(zeros(2)), I think we should retain current behavior
and Matlab compatibility and return a matrix of all Inf.

I ran 'make check' and got this one failing test


>>>>> processing /home/rik/wip/Projects_Mine/octave-dbg/test/diag-perm.tst
***** test
 x = diag (1:3);
 assert (inv (x), diag ([1 1/2 1/3]));
 x = diag (0:2);
 assert (inv (x), diag ([Inf 1 1/2]));
!!!!! test failed
ASSERT errors for:  assert (inv (x),diag ([Inf, 1, 1 / 2]))

  Location  |  Observed  |  Expected  |  Reason
   (2,2)         Inf           1         'Inf' mismatch
   (3,3)         Inf          0.5        'Inf' mismatch


In Matlab and Octave,


x = full (diag ([Inf, 1, 1/2]))
inv (x)
x =

       Inf   0.00000   0.00000
   0.00000   1.00000   0.00000
   0.00000   0.00000   0.50000

inv (x)
warning: matrix singular to machine precision
ans =

   0  -0  -0
   0   1  -0
   0   0   2


 
Should the implementation of diagonal matrices be more like sparse (Inf
whenever the element is non-zero) or should it be more like full matrices (the
result above which is 1./diagonal_matrix)?



This is simple to fix


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?56232>

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