https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62300
--- Comment #4 from Dave Malcolm <dmalcolm at redhat dot com> ---
Sorry about this.
The issue is here:
3042 insn = as_a <rtx_insn *> (
3043 gen_extend_insn (op0, t, promoted_nominal_mode,
3044 data->passed_mode, unsignedp));
which I mistakenly introduced in r214698.
The result of the gen_extend_insn is a SET, not an insn:
(set (reg/v:DI 151 [ ptrres ])
(zero_extend:DI (mem/f/c:SI (reg/f:SI 77 virtual-incoming-args) [8 ptrres+0
S4 A64])))
hence the as_a <rtx_insn *> fails.
Am attaching a patch which appears to fix it.
