Bug 192193

strace-4.5.14-0.EL4.x86_64.rpm

$ cat helloWorld.c
#include <stdio.h>

int main ( void )
{
    printf( "Hello World\n" );
    return 0;
}
$ gcc -g -m32 -o helloWorld helloWorld.c
$ strace -c ./helloWorld
[ Process PID=18962 runs in 32 bit mode. ]
Hello World
Process 18962 detached
% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
  3.96 1275605287168.1374389535023 4196074544597155       304           sendto
  4.51 1451698946383.1460288880979 4257185943319299       341       342 recv
  4.19 1348619729142.1425929142603 4049913378592020       333       334 send
Segmentation fault

The bt under ElectricFence:
2752                    if (counts == NULL || counts[i].calls == 0)
(gdb) bt
#0  0x0000000000405a41 in call_summary (outf=0x3d8e72e680) at syscall.c:2752
#1  0x0000000000403362 in main (argc=18945, argv=0x0) at strace.c:676
#2  0x0000003d8e51c4bb in __libc_start_main (main=0x402c00 <main>, argc=3,
   ubp_av=0x7fbffff6f8, init=0x413500 <__libc_csu_init>, fini=Variable "fini" is
not available.
)
   at ../sysdeps/generic/libc-start.c:209
(gdb) p i
$3 = 254

A good explanation of the problem from Bryan Mason:
"
Not only do x86_64 and i386 personalities have different numbers of syscalls
(i386 has 343 and x86_64 has 254), but the syscalls have different numbers.  For
example execve is syscall number 11 in i386 but it's syscall number 59 on x86_64.
"

The crash occurs when trying to access syscalls > 254.

The work-around is obviously to use the 32-bit strace in the meanwhile.