Split out from  issue 774573 .

Yunlian, Luis said you may be a good person to look at this?

---

Looks like there's glibc on reef (and maybe other boards) cause FORTIFY to SIGKILL programs instead of aborting them. From discussion on said issue, it sounds like this may keep us from getting crash reports?

Repro (run from inside my chroot):

$ cat foo.c
int main() {
        const char buf[] = "Hello, world!";
        char cs[4];
        __builtin___memcpy_chk(cs, buf, sizeof(buf), sizeof(cs));
}
$ clang -O2 foo.c -w
$ gdb -ex r ./a.out
[[snip]]
*** buffer overflow detected ***: /tmp/a.out terminated
======= Backtrace: =========
/lib64/libc.so.6(+0x73b12)[0x7ffff7aa5b12]
/lib64/libc.so.6(__fortify_fail+0x47)[0x7ffff7b35487]
/lib64/libc.so.6(+0x101540)[0x7ffff7b33540]
/tmp/a.out[0x40056a]
/lib64/libc.so.6(__libc_start_main+0xf0)[0x7ffff7a52730]
/tmp/a.out[0x400489]
======= Memory map: ========
[[snip]]
Program received signal SIGABRT, Aborted.
0x00007ffff7a65cf2 in raise () from /lib64/libc.so.6

$ sudo cp a.out /build/reef
$ gdb-reef /a.out
(reef-gdb) r
Starting program: /a.out
*** buffer overflow detected ***: a.out terminated; report to <http://crbug.com/new>

Program terminated with signal SIGKILL, Killed.
The program no longer exists.
(reef-gdb)


Doing some initial greping around, FORTIFY lands us in __chk_fail, which calls __hardened_gentoo_fail. If ENABLE_SSP_SMASH_DUMPS_CORE is truthy, we'll abort(). Otherwise, we'll SIGKILL ourselves. SSP_SMASH_DUMPS_CORE controls whether ENABLE_SSP_SMASH_DUMPS_CORE is truthy.