https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88945
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2019-01-21
Target Milestone|9.0 |---
Ever confirmed|0 |1
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed. remove_bb dumping BB contents is fishy ... but the real culprit is
dump_generic_node doing
op0 = array_ref_low_bound (node);
op1 = array_ref_element_size (node);
if (!integer_zerop (op0)
|| TREE_OPERAND (node, 2)
|| TREE_OPERAND (node, 3))
{
pp_string (pp, "{lb: ");
dump_generic_node (pp, op0, spc, flags, false);
pp_string (pp, " sz: ");
dump_generic_node (pp, op1, spc, flags, false);
pp_right_brace (pp);
}
break;
here operand 3 is a released SSA name. fold_convert_loc deals with
ERROR_MARK TREE_TYPE but not NULL (as it is for SSA names).
I agree it's annoying but not sure to what extent we want to fix this...
(eventually using error_mark_node for released SSA names is an acceptable fix)
