https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90972
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |ASSIGNED
Last reconfirmed| |2019-06-24
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
Target Milestone|--- |9.2
Summary|[9 Regression] ICE in |[9/10 Regression] ICE in
|fold_convert_loc, at |fold_convert_loc, at
|fold-const.c:2429 |fold-const.c:2429
Ever confirmed|0 |1
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.
#3 0x00000000011a188f in vect_get_vec_def_for_operand (
op=<string_cst 0x7ffff6a45210>, stmt_vinfo=0x269f460, vectype=<tree 0x0>)
at /space/rguenther/src/svn/gcc-9-branch/gcc/tree-vect-stmts.c:1608
1608 return vect_init_vector (stmt_vinfo, op, vector_type, NULL);
(gdb) p op
$1 = <string_cst 0x7ffff6a45210>
(gdb) p debug_gimple_stmt (stmt_vinfo->stmt)
# .MEM_23 = VDEF <.MEM_12>
MEM[(char * {ref-all})p_11] = "\x06";
and
1483 else if (CONSTANT_CLASS_P (val))
1484 val = fold_convert (TREE_TYPE (type), val);
which is too simplistic for STRING_CSTs. I think we want to use
VIEW_CONVERT_EXPRs here. Testing patch.
