[PATCH] fortran: Fix up TYPE_ARG_TYPES of procs with scalar VALUE optional args [PR92305]

[Tobias Burnus]

Aha, there the problem was hidden! — I assumed something in the tree 
went wrong, but didn't spot where.

Does this also fix PR 92844?

Thanks for investigating and fixing the issue. – LGTM.

(I think one needs to re-check this when fixing some of the VALUE bugs, 
but that's independent of this patch.)

Tobias

On 2/1/20 12:52 AM, Jakub Jelinek wrote:
> Hi!
>
> The following patch fixes
> -FAIL: libgomp.fortran/use_device_addr-1.f90   -O0  execution test
> -FAIL: libgomp.fortran/use_device_addr-2.f90   -O0  execution test
> that has been FAILing for several months on powerpc64le-linux.
> The problem is in the Fortran FE, which adds the artificial arguments
> for scalar VALUE OPTIONAL dummy args only to DECL_ARGUMENTS where the
> current function can see them, but not to TYPE_ARG_TYPES; if those functions
> aren't varargs, this confuses calls.c to pass the remaining arguments
> (which aren't named (== not covered by TYPE_ARG_TYPES) and aren't varargs
> either) in a different spot from what the callee (which has proper
> DECL_ARGUMENTS for all args) expects.  For the artificial length arguments
> for character dummy args we already put them in both DECL_ARGUMENTS and
> TYPE_ARG_TYPES.
>
> Fixed thusly, bootstrapped/regtested on x86_64-linux, i686-linux,
> powerpc64le-linux and powerpc64-linux (the last one with both -m32/-m64).
> Ok for trunk?
>
> 2020-01-31  Jakub Jelinek  <jakub@redhat.com>
>
> 	PR fortran/92305
> 	* trans-types.c (gfc_get_function_type): Also push boolean_type_node
> 	types for non-character scalar VALUE optional dummy arguments.
> 	* trans-decl.c (create_function_arglist): Skip those in
> 	hidden_typelist.  Formatting fix.
>
> --- gcc/fortran/trans-types.c.jj	2020-01-12 11:54:36.000000000 +0100
> +++ gcc/fortran/trans-types.c	2020-01-31 21:26:34.199188677 +0100
> @@ -3098,6 +3098,16 @@ gfc_get_function_type (gfc_symbol * sym,
>   
>   	  vec_safe_push (typelist, type);
>   	}
> +      /* For noncharacter scalar intrinsic types, VALUE passes the value,
> +	 hence, the optional status cannot be transferred via a NULL pointer.
> +	 Thus, we will use a hidden argument in that case.  */
> +      else if (arg
> +	       && arg->attr.optional
> +	       && arg->attr.value
> +	       && !arg->attr.dimension
> +	       && arg->ts.type != BT_CLASS
> +	       && !gfc_bt_struct (arg->ts.type))
> +	vec_safe_push (typelist, boolean_type_node);
>       }
>   
>     if (!vec_safe_is_empty (typelist)
> --- gcc/fortran/trans-decl.c.jj	2020-01-30 09:34:43.207088430 +0100
> +++ gcc/fortran/trans-decl.c	2020-01-31 21:28:49.272197084 +0100
> @@ -2645,8 +2645,8 @@ create_function_arglist (gfc_symbol * sy
>   	      || f->sym->ts.u.cl->backend_decl == length)
>   	    {
>   	      if (POINTER_TYPE_P (len_type))
> -		f->sym->ts.u.cl->backend_decl =
> -			build_fold_indirect_ref_loc (input_location, length);
> +		f->sym->ts.u.cl->backend_decl
> +		  = build_fold_indirect_ref_loc (input_location, length);
>   	      else if (f->sym->ts.u.cl->backend_decl == NULL)
>   		gfc_create_string_length (f->sym);
>   
> @@ -2677,6 +2677,8 @@ create_function_arglist (gfc_symbol * sy
>             DECL_ARG_TYPE (tmp) = boolean_type_node;
>             TREE_READONLY (tmp) = 1;
>             gfc_finish_decl (tmp);
> +
> +	  hidden_typelist = TREE_CHAIN (hidden_typelist);
>   	}
>   
>         /* For non-constant length array arguments, make sure they use
>
> 	Jakub
>