https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123271
Bug ID: 123271
Summary: -ftrapv fails to detect overflow on ARM32 since GCC 10
Product: gcc
Version: 14.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: jiannmao at qq dot com
Target Milestone: ---
Starting from GCC 10, the -ftrapv flag no longer works correctly on ARM32
architecture. Programs compiled with -ftrapv fail to abort on signed integer
overflow and instead produce incorrect overflow results.
Target: ARM32 (tested on arm-linux-gnueabi, armv7-a)
Command: gcc -ftrapv test.c -o test
```
long sub(long a, long b)
{
return a - b;
}
int main(void)
{
printf("%x\n", sub(0x80000000, 10)); // should overflow and abort, but
get incorrectly prints: 7ffffff6
return 0;
}
```
The program should abort when overflow occurs (as it did in GCC 9 and
earlier).but it completes normally and returns an incorrect overflowed result:
0x7ffffff6
GCC 9 (CORRECT - calls __subvsi3):
```
sub:
push {r7, lr}
sub sp, sp, #8
add r7, sp, #0
str r0, [r7, #4]
str r1, [r7]
ldr r1, [r7]
ldr r0, [r7, #4]
bl __subvsi3 # Calls libgcc function that checks overflow
mov r3, r0
mov r0, r3
adds r7, r7, #8
mov sp, r7
pop {r7, pc}
```
GCC 10+ (INCORRECT - direct subtraction without checking):
```
sub:
push {r7}
sub sp, sp, #12
add r7, sp, #0
str r0, [r7, #4]
str r1, [r7]
ldr r2, [r7, #4]
ldr r3, [r7]
subs r3, r2, r3
mov r0, r3 # No overflow check after subs
adds r7, r7, #12
mov sp, r7
ldr r7, [sp], #4
bx lr
```
