This is the mail archive of the gcc-help@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Trying to understand inconsistent behaviour with -ftrapv (and an ICE)


Hi,

This simple program:

#include <stdio.h>
#include <limits.h>
int main(void) {
    volatile int i = INT_MAX + 1u;
    puts("so far so good");
    i = i * -1u;
    puts("never reached with 4.1");
    return 0;
}

when compiled with -ftrapv (using GCC 4.1.1) aborts during the
  i = i * -1u;

Compiling with -ftrapv with older GCC versions (3.3.6, 3.4.6 and 4.0.3)
gets both lines printed. If I understand correctly, both the addition
and the multiplication are done unsigned and so shouldn't cause the
program to abort, and the conversion to int by storing it in i isn't
supposed to cause the program to abort either (and doesn't the first
time), so am I missing something obvious, or might this be a bug?

Something I noticed while writing this mail: probably also interesting
is that if I don't declare i as volatile, and return !i to prevent it
from being optimised away, GCC 4.1.1 ICEs if compiled with -O[23s],
aborts (as with volatile) with -O0, but behaves correctly with -O1. I
can't imagine an ICE not being a bug, but is it the same problem as the
first one, or does it just happen to show up with similar code?


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]