[Bug fortran/89384] [9 Regression] CONTIGUOUS dummy argument in BIND(C) interface incorrect when actual is non-contiguous

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89384

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pault at gcc dot gnu.org

--- Comment #2 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
I did a bit more of bisecting, and it seems that r267881 is the one.

With that revision, the tree dump shows

ctg (struct array01_real(kind=4) & restrict x)
{
  integer(kind=8) ubound.0;
  integer(kind=8) stride.1;
  integer(kind=8) offset.2;
  integer(kind=8) size.3;
  real(kind=4)[0:D.2275] * restrict x.0;

[...]

      _gfortran_gfc_desc_to_cfi_desc (&cfi.11, &parm.10);
      ctg (cfi.11);
      _gfortran_cfi_desc_to_gfc_desc (&parm.10, &cfi.11);

whereas at r267880 we get

ctg (struct array01_real(kind=4) & restrict x)
{
  integer(kind=8) ubound.0;
  integer(kind=8) stride.1;

[...]

      D.2290 = _gfortran_internal_pack (&parm.10);
      parm.10.data = D.2290;
      ctg (&parm.10);
      parm.10.data = origptr.11;
      if ((real(kind=4)[0:] *) parm.10.data != (real(kind=4)[0:] *) D.2290)
        {
          _gfortran_internal_unpack (&parm.10, D.2290);
          __builtin_free (D.2290);
        }
    }

so it seems the pack/unpack is missing.  I'm also fairly
sure that the previous code would not have worked together with C
in the absence of ISO_Fortran_binding.h :-)