> -----Original Message----- > From: gcc-help-owner@xxxxxxxxxxx [mailto:gcc-help-owner@xxxxxxxxxxx] On > Behalf Of Andrew Haley > Sent: Dienstag, 3. Januar 2012 16:52 > To: gcc-help@xxxxxxxxxxx > Subject: Re: Questions before submitting a bug report > > On 01/03/2012 03:39 PM, Markus Henschel wrote: > >> -----Original Message----- > >> From: gcc-help-owner@xxxxxxxxxxx [mailto:gcc-help-owner@xxxxxxxxxxx] > >> On Behalf Of Andrew Haley > >> Sent: Dienstag, 3. Januar 2012 16:26 > >> To: gcc-help@xxxxxxxxxxx > >> Subject: Re: Questions before submitting a bug report > >> > >> On 01/03/2012 03:02 PM, Markus Henschel wrote: > >>> Is there some obvious flaw in my code? > >> > >> What error did you get? Not all CPUs can do __sync_add_and_fetch on > >> a long long. > > > > The resulting program crashes with a segmentation fault on the line > with __sync_add_and_fetch. This is the complete command line I used for > compiling: > > > > "gcc -fPIC -march=i586 -fvisibility=hidden -O3 mytest.c -o mytest" > > > > The error goes away if I remove "-fPIC", "-fvisibility=hidden" or if > I use some less aggressive optimization. It also disappears if I use a > long instead of a long long. > > I think you've found yourself a bug. I can't reproduce it, nor can I > see anything wrong with your code. It's be interesting to see what > instruction makes it crash. > > Andrew. Thanks for having a look at this. To satisfy your curiosity here is the asm output of gdb for an even simpler test case: int main(int argc, char * argv[]) { static long long foo; return __sync_add_and_fetch(&foo, foo); } I don't know if this is meaningful because I don't know much about assembler. ┌───────────────────────────────────────────────────────────────────────────┐ │0x8048350 <main> lea 0x4(%esp),%ecx │ │0x8048354 <main+4> and $0xfffffff0,%esp │ │0x8048357 <main+7> pushl -0x4(%ecx) │ │0x804835a <main+10> push %ebp │ │0x804835b <main+11> mov %esp,%ebp │ │0x804835d <main+13> push %edi │ │0x804835e <main+14> push %esi │ │0x804835f <main+15> push %ebx │ │0x8048360 <main+16> push %ecx │ │0x8048361 <main+17> call 0x8048366 <main+22> │ │0x8048366 <main+22> pop %ebx │ │0x8048367 <main+23> add $0x1206,%ebx │ │0x804836d <main+29> sub $0x8,%esp │ │0x8048370 <main+32> mov 0x2c(%ebx),%eax │ │0x8048376 <main+38> mov 0x30(%ebx),%edx │ │0x804837c <main+44> mov %eax,%esi │ │0x804837e <main+46> mov %edx,%edi │ │0x8048380 <main+48> add 0x2c(%ebx),%esi │ │0x8048386 <main+54> adc 0x30(%ebx),%edi │ │0x804838c <main+60> mov %esi,-0x18(%ebp) │ │0x804838f <main+63> mov %edi,-0x14(%ebp) │ │0x8048392 <main+66> mov %esi,%edi │ │0x8048394 <main+68> mov -0x14(%ebp),%ecx │ │0x8048397 <main+71> xchg %ebx,%edi │ >│0x8048399 <main+73> lock cmpxchg8b 0x2c(%ebx) │ │0x80483a1 <main+81> xchg %ebx,%edi │ │0x80483a3 <main+83> jne 0x804837c <main+44> │ │0x80483a5 <main+85> mov -0x18(%ebp),%eax │ │0x80483a8 <main+88> pop %edx │ └───────────────────────────────────────────────────────────────────────────┘ child process 15488 In: main Line: ?? PC: 0x8048399 (gdb) run Starting program: /home/ubuntu/mytest Program received signal SIGSEGV, Segmentation fault. 0x08048399 in main () If gcc 4.2 isn't maintained anymore I'll use a workaround and see that we can switch to a newer gcc release in the future. Thank you for your time. Markus
