https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72770
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
.original has:
case 85893463:;
.__tmp_CHARACTER_0_1 = x->_len;
__tmp_CHARACTER_0_1 = x->_data;
__tmp_CHARACTER_0_1.dtype = ((integer(kind=8)) SAVE_EXPR
<(sizetype) .__tmp_CHARACTER_0_1> << 6) + 49;
{
logical(kind=4) test.2;
character(kind=1)[0:][1:.__tmp_CHARACTER_0_1] * D.3488;
integer(kind=8) D.3489;
integer(kind=8) D.3490;
integer(kind=8) D.3491;
integer(kind=8) D.3494;
test.2 = 0;
D.3488 = (character(kind=1)[0:][1:.__tmp_CHARACTER_0_1] *)
__tmp_CHARACTER_0_1.data;
D.3489 = __tmp_CHARACTER_0_1.offset;
D.3490 = __tmp_CHARACTER_0_1.dim[0].lbound;
...
and the type of D.3488 misses a DECL_EXPR (or manual handling in the Fortran FE
like we do for some cases - simply copy x->_len to a decl and use that for
the upper bound).
