[Bug fortran/85631] [8/9 Regression] Runtime error message array bound mismatch with nonzero optimization

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85631

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Blocks|                            |27766

--- Comment #3 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
The bouns checking is inserted in the wrong place, before
setting the descriptor:

$ cat b.f90
module x
contains
  subroutine sub(a)
    real, dimension(:,:), intent(in) :: a
    real, allocatable, dimension(:,:) :: var_1_mma
    var_1_mma = transpose(a)
  end subroutine sub
end module x

program main
  use x
  implicit none
  real, dimension(:,:), allocatable :: block_1

  block_1 = reshape([-2,3,-5,7], [2,2])
  call sub(block_1)
end program

$ grep mma b.f90.003t.original 
  struct array02_real(kind=4) var_1_mma;
      var_1_mma.data = 0B;
        D.3809 = (real(kind=4)[0:] * restrict) var_1_mma.data;
        D.3810 = var_1_mma.offset;
        D.3811 = var_1_mma.dim[0].lbound;
        D.3812 = var_1_mma.dim[0].ubound;
        D.3813 = var_1_mma.dim[1].lbound;
        D.3814 = var_1_mma.dim[1].ubound;
        if ((integer(kind=8)) (D.3811 <= D.3812 && var_1_mma.dim[0].lbound >
D.3811))
            _gfortran_runtime_error_at (&"At line 6 of file b.f90"[1]{lb: 1 sz:
1}, &"Index \'%ld\' of dimension 1 of array \'var_1_mma\' outside of expected
range (%ld:%ld)"[1]{lb: 1 sz: 1}, D.3811, var_1_mma.dim[0].lbound,
var_1_mma.dim[0].ubound);

...
            var_1_mma.dim[0].lbound = 1;
            var_1_mma.dim[0].ubound = NON_LVALUE_EXPR <D.3807>;
            var_1_mma.dim[0].stride = 1;
            var_1_mma.dim[1].lbound = 1;
            var_1_mma.dim[1].ubound = NON_LVALUE_EXPR <D.3808>;
            var_1_mma.dim[1].stride = NON_LVALUE_EXPR <D.3807>;

As a matter of fact, we are taking the variables from
the assignmemt, there should be no checking at all.

Hmm...


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=27766
[Bug 27766] [meta-bug] -fbounds-check related bugs