http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51605

--- Comment #8 from Tobias Burnus <burnus at gcc dot gnu.org> 2011-12-19 
11:39:55 UTC ---
Created attachment 26139
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26139
Draft patch (for the rejects-valid part, not for the ICE)

Draft patch, which allows again
  ptr => associate-name
It also fixes an ICE ("class_ok" part, "(a)") and issues with coarrays and
select type. [Coarrays now fail after resolution with an ICE.]

Test cases (a) to (d):

type t
end type t
class(t), target :: p1               ! (a) Invalid (was ICEing before)
!class(t), allocatable, target :: p1 ! (b) Valid (unchanged by the patch)
!class(t), allocatable :: p1         ! (c) Correctly fails (ditto)
!class(t), pointer :: p1             ! (d) Valid (now accepted)
class(t), pointer :: p2

select type(p1)
  type is(t)
    p2 => p1
  class is(t)
    p2 => p1
end select
end

 * * *

Having fixed the rejects-valid issue, one again hits:
  internal compiler error: in gfc_trans_block_construct, at
fortran/trans-stmt.c:1215

1209      ns = code->ext.block.ns;
1211      sym = ns->proc_name;
1215      gcc_assert (!sym->tlink);

Here, "sym->name" is "eval_int_expr" and sym->tlink == sym.


The issue is seemingly the following code in an internal procedure:
      integer_or_logical: select type(symbol_ptr)
      ...
      end select integer_or_logical

and then another internal procedure:

  subroutine integer_or_logical( expr_str, flag)
    ...
  end subroutine integer_or_logical

  recursive subroutine eval_int_expr( int_expr, value)

The label "integer_or_logical" of the block some clashes with the same-named
subroutine - and thus marks the subroutine which follows somehow as
EXPR_BLOCK?!?

Reply via email to