https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92017
Tobias Burnus <burnus at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2019-10-11
CC| |burnus at gcc dot gnu.org
Ever confirmed|0 |1
--- Comment #1 from Tobias Burnus <burnus at gcc dot gnu.org> ---
> Starting with gfortran-8
Which is not surprising as gfortran-7 rejects CHARACTER
to minloc/maxloc.
The problem is that the character length is missing:
The first argument to minloc/maxloc is named "array" in the Fortran
standard. And also in gfc_simplify_minmaxloc); there:
array->expr_type == EXPR_ARRAY and ts.type == BT_CHARACTER
but array->ts.u.cl == NULL – which causes the problem.
Looking at resolve.c's fixup_charlen(), that can be fixed by the following,
but my feeling is that it should be fixed earlier - as I am sure more locations
in simplify.c have the same issue.
--- a/gcc/fortran/simplify.c
+++ b/gcc/fortran/simplify.c
@@ -5427,4 +5427,7 @@ gfc_simplify_minmaxloc (gfc_expr *array, gfc_expr
gcc_unreachable();
+ if (array->expr_type == EXPR_ARRAY && array->ts.type == BT_CHARACTER)
+ gfc_resolve_character_array_constructor (array);
+
extremum = gfc_get_constant_expr (array->ts.type, array->ts.kind,
&array->where);
init_result_expr (extremum, init_val, array);
