Help with #32747

[Jordi Gutiérrez Hermoso]

I've been spending some time with

     https://savannah.gnu.org/bugs/?32747

What I've been able to ascertain so far is that when the
SparseRep::SparseRep (octave_idx_type, octave_idx_type,
octave_idx_type) ctor in liboctave/Sparse.h:90 is called by
spallocate, there's unitialised memory in the SparseRep::r member
variable, but that shouldn't matter, because that memory shouldn't be
getting read. Further on inside Sparse<T>::assign (const idx_vector&,
const idx_vector&, const Sparse<T>&) near liboctave/Sparse.cc:2047,
during the S(5,1:2) call, the tmp variable that gets assigned to *this
is already all wrong, and I think it's at this point that the
uninitialised memory gets read. In the offending changeset that I
found with hg bisect, I see that there is a switch to use nnz instead
of nzmax, and I think there should have been a corresponding change
elsewhere in the aforementioend Sparse<T>::assign function that also
switches nzmax with nnz.

Anyways, that's just what I've been able to find out, and I'm
wondering if anyone else has further thoughts.

Thanks,
- Jordi G. H.