[Bug fortran/68560] [6 Regression] The test gfortran.dg/shape_8.f90 now fails when compiled with -flto

[dominiq at lps dot ens.fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68560

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tkoenig at gcc dot gnu.org
          Component|lto                         |fortran

--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
For the test

program test
   implicit none
   integer :: c(2), d(2)
   real, allocatable :: x(:,:)

   allocate(x(2,5))
   c = shape(x)
   d = shape(x,kind=4)
end

the result of the compilation with -fdump-tree-original is

...
    {
      struct array2_real(kind=4) * D.3443;
      struct array1_integer(kind=4) parm.1;

      parm.1.dtype = 265;
      parm.1.dim[0].lbound = 1;
      parm.1.dim[0].ubound = 2;
      parm.1.dim[0].stride = 1;
      parm.1.data = (void *) &c[0];
      parm.1.offset = -1;
      D.3443 = &x;
      _gfortran_shape_4 (&parm.1, D.3443);
    }
    {
      integer(kind=4) * D.3453;
      static integer(kind=4) C.3452 = 4;
      struct array2_real(kind=4) * D.3451;
      struct array1_integer(kind=4) parm.2;

      parm.2.dtype = 265;
      parm.2.dim[0].lbound = 1;
      parm.2.dim[0].ubound = 2;
      parm.2.dim[0].stride = 1;
      parm.2.data = (void *) &d[0];
      parm.2.offset = -1;
      D.3451 = &x;
      D.3453 = &C.3452;
      _gfortran_shape_4 (&parm.2, D.3451, D.3453);
    }
...

and -Wlto-type-mismatch seems right to complain that

_gfortran_shape_4 (&parm.2, D.3451, D.3453);

does not match

      _gfortran_shape_4 (&parm.1, D.3443);

I have tried to understand why using the optional 'kind' argument generates a
third argument (not needed as it is reflected by _gfortran_shape_x for kind=x).

Looking at libgfortran/generated/shape_i4.c I see that shape_4 expects only two
arguments. 

Looking at gcc/fortran/trans-intrinsic.c, I see

...
    case GFC_ISYM_SHAPE:
    case GFC_ISYM_SPREAD:
    case GFC_ISYM_YN2:
      /* Ignore absent optional parameters.  */
      return 1;
...

Am I correct to understand that it is intended to prevents the emission of the
optional argument? If yes, why is it not working (I lost the scent after that)?