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

[anonyme]

Follow-up Comment #21, bug #56232 (project octave):

Proposition for a simple fix.

1/ Rather than use Rik's patch in sparse-lu.cc around line 459, male a change
like


// If nz is zero then ridx is probably a nullptr. In this case
// the value of ridx is not important but UMFPACK needs a valid
// pointer in any case, so just set it to cidx that is always
// valid
const octave_idx_type *Ai = (a.cidx(nc+1) == 0 ? a.cidx() : a.ridx ());


This should fix the LU factorization problem

2/ In tinverse (line 718 in dsparse.cc) make the change


      octave_idx_type nz2 = nz == 0 ? 1 ; nz;


and the same in CSparse.cc. This should fix the segfault


I can't test it but this will probably work, even if inelegant

David


    _______________________________________________________

Reply to this item at:

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

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