https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54262
--- Comment #3 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
> The code in comment 0 compiles with 4.8.5, but gives an error
>
> p = transfer (loc(x), p)
> 1
> Error: Assumed-type argument at (1) is not permitted as actual argument
> to the intrinsic loc
LOC is a GNU extension and it is a matter of choice to accept or reject it in
gfortran.
If the choice is to reject it, then this PR should be closed as INVALID.
With the following patch
--- ../_clean/gcc/fortran/intrinsic.c 2019-03-12 16:12:35.000000000 +0100
+++ gcc/fortran/intrinsic.c 2019-03-20 23:03:17.000000000 +0100
@@ -212,7 +212,8 @@ do_ts29113_check (gfc_intrinsic_sym *spe
&& specific->id != GFC_ISYM_SIZEOF
&& specific->id != GFC_ISYM_UBOUND
&& specific->id != GFC_ISYM_IS_CONTIGUOUS
- && specific->id != GFC_ISYM_C_LOC)
+ && specific->id != GFC_ISYM_C_LOC
+ && specific->id != GFC_ISYM_LOC)
{
gfc_error ("Assumed-type argument at %L is not permitted as actual"
" argument to the intrinsic %s", &a->expr->where,
the modified test with
type(*), target, CONTIGUOUS :: x(:) ! or x(:)
is accepted, but the original test is rejected with
5 | p = transfer (loc(x), p)
| 1
Error: Assumed-rank argument at (1) is only permitted as actual argument to
intrinsic inquiry functions
So far I did not find what to change to make specific->inquiry true for LOC.
Again as it is an extension, it is not an "intrinsic inquiry function"
as defined by the standard.
While looking at trans-intrinsic.c, I see in gfc_conv_intrinsic_loc
/* Create a temporary variable for loc return value. Without this,
we get an error an ICE in gcc/expr.c(expand_expr_addr_expr_1). */
temp_var = gfc_create_var (gfc_get_int_type (gfc_index_integer_kind), NULL);
gfc_add_modify (&se->pre, temp_var, se->expr);
se->expr = temp_var;
and in conv_isocbinding_function
/* TODO -- the following two lines shouldn't be necessary, but if
they're removed, a bug is exposed later in the code path.
This workaround was thus introduced, but will have to be
removed; please see PR 35150 for details about the issue. */
se->expr = convert (pvoid_type_node, se->expr);
se->expr = gfc_evaluate_now (se->expr, &se->pre);
These pieces of code look more as empirical hacks than as solid codes.
