This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
ubsan: ia64: left shift of negative value
- From: Alan Modra <amodra at gmail dot com>
- To: binutils at sourceware dot org
- Date: Wed, 11 Dec 2019 11:15:50 +1030
- Subject: ubsan: ia64: left shift of negative value
Here, since val is signed:
*valuep = (val << scale);
* cpu-ia64-opc.c (ext_imms_scaled): Avoid undefined left shift
of negative values by using unsigned vars.
diff --git a/bfd/cpu-ia64-opc.c b/bfd/cpu-ia64-opc.c
index 84ee0e2e4a..8df90befe3 100644
--- a/bfd/cpu-ia64-opc.c
+++ b/bfd/cpu-ia64-opc.c
@@ -186,7 +186,7 @@ ext_imms_scaled (const struct ia64_operand *self, ia64_insn code,
ia64_insn *valuep, int scale)
{
int i, bits = 0, total = 0;
- BFD_HOST_64_BIT val = 0, sign;
+ BFD_HOST_U_64_BIT val = 0, sign;
for (i = 0; i < NELEMS (self->field) && self->field[i].bits; ++i)
{
@@ -196,10 +196,10 @@ ext_imms_scaled (const struct ia64_operand *self, ia64_insn code,
total += bits;
}
/* sign extend: */
- sign = (BFD_HOST_64_BIT) 1 << (total - 1);
+ sign = (BFD_HOST_U_64_BIT) 1 << (total - 1);
val = (val ^ sign) - sign;
- *valuep = (val << scale);
+ *valuep = val << scale;
return 0;
}
--
Alan Modra
Australia Development Lab, IBM