https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71203
--- Comment #15 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
(In reply to Gerhard Steinmetz from comment #1)
> $ cat z4.f90
> program p
> integer :: i
> integer, parameter :: x(2) = 0
> integer, parameter :: y(*) = [(x(i:i), i=1,2)]
> end
>
The patch
Index: decl.c
===================================================================
--- decl.c (Revision 269443)
+++ decl.c (Arbeitskopie)
@@ -1983,8 +1983,14 @@ add_init_expr_to_sym (const char *name, gfc_expr *
return false;
}
- /* Shape should be present, we get an initialization expression. */
- gcc_assert (init->shape);
+ /* The shape may be NULL for EXPR_ARRAY, set it. */
+ if (init->shape == NULL)
+ {
+ gcc_assert (init->expr_type == EXPR_ARRAY);
+ init->shape = gfc_get_shape (1);
+ if (!gfc_array_size (init, &init->shape[0]))
+ gfc_internal_error ("bletchful state");
+ }
for (dim = 0; dim < sym->as->rank; ++dim)
{
fixes this, but a) causes some strange regression in the testsuite,
some tree dump patterns in assumed_type_2.f90 and no_arg_check_2.f90
and b) might not the right place to put in the shape.
