[Octave-bug-tracker] [bug #58795] ode15i and ode15s fail for Windows 32bit

[Markus Mützel]

Follow-up Comment #48, bug #58795 (project octave):

You wrote in comment #35 that "sunindextype" is used to interface to
BLAS/LAPACK functions inside SUNDIALS. You concluded that this means that
"sunindextype" must match the integer type used by the BLAS/LAPACK libraries,
and must hence match "octave_f77_int_type".
I agree with that conclusion.

I think there is only one place where we have to check on conversion from
"octave_idx_type" to "octave_f77_int_type".
I'm attaching an updated patch.

The following lines look fishy wrt integer overflow (in particular
"m_num*m_num" --- "m_num" is of type "octave_f77_int_type"):

        // FIXME : one should not allocate space for a full Jacobian
        // when using a sparse format.  Consider allocating less space
        // then possibly using SUNSparseMatrixReallocate to increase it.
        m_sunJacMatrix = SUNSparseMatrix (m_num, m_num, m_num*m_num,
CSC_MAT);


The prototype of that function is "SUNMatrix SUNSparseMatrix(sunindextype M,
sunindextype N, sunindextype NNZ, int sparsetype)".

Is there something similarly straight-forward that we can do to avoid an
integer overflow?

(file #49737)
    _______________________________________________________

Additional Item Attachment:

File name: bug58795_ode15_integer_types_v2.patch Size:9 KB
   
<https://file.savannah.gnu.org/file/bug58795_ode15_integer_types_v2.patch?file_id=49737>



    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?58795>

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