https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126257
Bug ID: 126257
Summary: [16/17 Regression] wrong code at -O2/3 with _BitInt(2)
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: xxs_chy at outlook dot com
Target Milestone: ---
Reproducer: https://godbolt.org/z/59bcMqdP3
Testcase:
#include <stdint.h>
#include <stdio.h>
_BitInt(2) g5;
int64_t g14, __chk, g22;
int main()
{
int32_t ov0;
_BitInt(2) v2;
_Bool c3;
int64_t v6 = g14;
lbl_br3:
ov0 = v6;
if (ov0 != 4) v2 = -1;
v2 = v2 + v2;
g5 = v2;
c3 = g22 - 4;
if (!c3) goto lbl_br3;
__chk = g5;
printf("checksum=0x%016llx\n", (unsigned long long)__chk);
}
This case outputs checksum=0x0000000000000000 with -O2 and
checksum=0xfffffffffffffffe with -O0.
And UBSan seems to report a false positive, as the reported line is actually a
legal addition of -1 and -1:
example.c:14:8: runtime error: signed integer overflow: -1 * -2 cannot be
represented in type '_BitInt(2)'
