https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119948
--- Comment #18 from kargls at comcast dot net --- On 5/2/25 01:06, pault at gcc dot gnu.org wrote: > I got to thinking for once. I think that your change to primary.cc has cast > the > net a bit too wide and risks turning other kinds of interfaces into variables. > How about the more specific: > > diff --git a/gcc/fortran/primary.cc b/gcc/fortran/primary.cc > index 161d4c26964..85539612504 100644 > --- a/gcc/fortran/primary.cc > +++ b/gcc/fortran/primary.cc > @@ -4391,7 +4391,7 @@ match_variable (gfc_expr **result, int equiv_flag, int > host_flag) > case FL_PROCEDURE: > /* Check for a nonrecursive function result variable. */ > if (sym->attr.function > - && !sym->attr.external > + && (!sym->attr.external || sym->abr_modproc_decl) > && sym->result == sym > && (gfc_is_function_return_value (sym, gfc_current_ns) > || (sym->attr.entry > > which does the right thing and regtests OK? I must admit that I have not unraveled the logic in the if-stmt. The comment mentions nonrecursive, yet I see no test involving sym->attr.recursive. At one time, a recursive procedure needed to have an explicit interface with the RECURSIVE prefix, so the !sym->attr.external would seem to be the test. However, F2018 has changed Fortran where Procedures that are not declared with an asterisk type-param-value, including ELEMENTAL procedures, can be invoked recursively by default; the RECURSIVE keyword is advisory only. The NON_RECURSIVE keyword specifies that a procedure is not recursive. That said, your change is fine with me.
