https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71077
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |ice-on-valid-code
Status|UNCONFIRMED |NEW
Last reconfirmed| |2016-05-12
CC| |ppalka at gcc dot gnu.org
Target Milestone|--- |7.0
Summary|gcc -lto raises ICE |[7 Regression] gcc -lto
| |raises ICE
Ever confirmed|0 |1
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.
#1 0x0000000000d9c921 in find_taken_edge_cond_expr (
bb=<basic_block 0x7ffff688f9c0 (38)>, val=<vector_cst 0x7ffff66bc320>)
at /space/rguenther/src/svn/trunk/gcc/tree-cfg.c:2230
2230 gcc_assert (TREE_CODE (val) == INTEGER_CST);
(gdb) l
2225 {
2226 edge true_edge, false_edge;
2227
2228 extract_true_false_edges_from_block (bb, &true_edge, &false_edge);
2229
2230 gcc_assert (TREE_CODE (val) == INTEGER_CST);
2231 return (integer_zerop (val) ? false_edge : true_edge);
2232 }
2233
2234 /* Given an INTEGER_CST VAL and the entry block BB to a SWITCH_EXPR
(gdb) p val
$1 = <vector_cst 0x7ffff66bc320>
simplify_control_stmt_condition doesn't handle vector equality tests in
GIMPLE_CONDs nor do I think it handles the combining step correctly
(failing to properly use constant_boolean_node instead of zero/one).
