[Bug fortran/104651] New: [OOP] CLASS with assumed-size/assumed-rank array

[Bug fortran/104651] New: [OOP] CLASS with assumed-size/assumed-rank array

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104651

            Bug ID: 104651
           Summary: [OOP] CLASS with assumed-size/assumed-rank array
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: rejects-valid
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
  Target Milestone: ---

The following looks valid - but gfortran complains:

    7 | call bar(A)
      |         1
Error: Rank mismatch in argument ‘x’ at (1) (rank-1 and rank-2)

    8 | call bar2(B)
      |          1
Error: Rank mismatch in argument ‘y’ at (1) (rank-2 and rank-1)


Additionally, I had expected that the array is passed just to class->_data, but
the dump shows that there is a full array descriptor, i.e. class->_data.data
for the dummy argument.


In trans-types.cc's gfc_get_derived_type, the code path taken is:

          if (c->attr.pointer || c->attr.allocatable || c->attr.pdt_array)
...
              field_type = gfc_build_array_type (field_type, c->as, akind,

and not:
          else
            field_type = gfc_get_nodesc_array_type (field_type, c->as,
                                                    PACKED_STATIC,
                                                    !c->attr.target);


implicit none (type, external)
type t
integer :: i
end type t
type(t) :: A(10,10), B(10)

call bar(A)
call bar2(B)
contains
subroutine bar(x)
  class(t) :: x(*)
  x(1)%i = x(2)%i
end
subroutine bar2(y)
  class(t) :: y(5,10)
  y(1,1)%i = y(2,2)%i
end
end

[Bug fortran/104651] [OOP] CLASS with assumed-size/assumed-rank array

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104651

kargl at gcc dot gnu.org changed:

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

--- Comment #1 from kargl at gcc dot gnu.org ---
(In reply to Tobias Burnus from comment #0)
> The following looks valid - but gfortran complains:
> 
>     7 | call bar(A)
>       |         1
> Error: Rank mismatch in argument ‘x’ at (1) (rank-1 and rank-2)
> 
>     8 | call bar2(B)
>       |          1
> Error: Rank mismatch in argument ‘y’ at (1) (rank-2 and rank-1)
> 
> 
> Additionally, I had expected that the array is passed just to class->_data,
> but the dump shows that there is a full array descriptor, i.e.
> class->_data.data for the dummy argument.
> 
> 
> In trans-types.cc's gfc_get_derived_type, the code path taken is:
> 
>           if (c->attr.pointer || c->attr.allocatable || c->attr.pdt_array)
> ...
>               field_type = gfc_build_array_type (field_type, c->as, akind,
> 
> and not:
>           else
>             field_type = gfc_get_nodesc_array_type (field_type, c->as,
>                                                     PACKED_STATIC,
>                                                     !c->attr.target);
> 
> 
> implicit none (type, external)
> type t
> integer :: i
> end type t
> type(t) :: A(10,10), B(10)
> 
> call bar(A)
> call bar2(B)
> contains
> subroutine bar(x)
>   class(t) :: x(*)
>   x(1)%i = x(2)%i
> end
> subroutine bar2(y)
>   class(t) :: y(5,10)
>   y(1,1)%i = y(2,2)%i
> end
> end

Error look correct to me.  Interfaces are resolved by type, kind, and rank;
sometimes referred to TKR.  Is there an exception for class?

[Bug fortran/104651] [OOP] CLASS with assumed-size/assumed-rank array

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104651

--- Comment #2 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Additionally, for the following (from
gcc/testsuite/gfortran.dg/finalize_15.f90),

  class(t1), allocatable :: x(:,:)
  call fin_test_1(x(::2,::3))

with

  subroutine fin_test_1(x)
    class(t1), intent(out) :: x(5,4)

I had expected a copy in/copy out - but in reality, fin_test_1 receives a
noncontiguous array.

[Bug fortran/104651] [OOP] CLASS with assumed-size/assumed-rank array

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104651

--- Comment #3 from Tobias Burnus <burnus at gcc dot gnu.org> ---
(In reply to kargl from comment #1)
> (In reply to Tobias Burnus from comment #0)
> > Error: Rank mismatch in argument ‘x’ at (1) (rank-1 and rank-2)
> > Error: Rank mismatch in argument ‘y’ at (1) (rank-2 and rank-1)
> >
> > subroutine bar(x)
> >   class(t) :: x(*)
> >
> > subroutine bar2(y)
> >   class(t) :: y(5,10)
> 
> Error look correct to me.  Interfaces are resolved by type, kind, and rank;
> sometimes referred to TKR.  Is there an exception for class?

Passing an array of any rank or even array element to ASSUMED-SIZE and
EXPLICIT-SIZE ARRAYS is a Fortran 66 feature which still exists. For those, the
simple (contiguous) byte stream ("storage sequence") is passed – and the
partition of that stream into array bounds/elements is determined by the
callee.

Example:
    integer A(10), B(10,10), C(10,10,10)
    call foo(A,A); call foo(B,B); call foo(C,C)
  contains
    subroutine foo(x,y)
      integer, intent(in) :: x(5:7,*), y(5,2)
    end
  end

This works just fine in gfortran. But for CLASS the like-wise code does not
work.

>From the standard (allocatable/pointers and deferred-shape arrays are in
section):


Fortran 2018, "15.5.2.4 Ordinary dummy variables":

"If a dummy argument is an assumed-shape array, the rank of the actual argument
shall be the same as the rank of the dummy argument ..."

As this restriction does not exist for assumed-size/explicit-size array
dummies, those are permitted. Some more wording is then in "15.5.2.11 Sequence
association":

"Sequence association only applies when the dummy argument is an explicit-shape
or assumed-size array. The rest of this subclause only applies in that case.
"An actual argument represents an element sequence if it is an array
expression, an array element designator, a default character scalar, or a
scalar of type character with the C character kind (18.2.2). [...]"

For details, see the two sections mentioned above as I did leave out some bits
related to this discussion.