mudflapth not working

[Stuart Clarke <sjclarke at tpg dot com dot au>]

Hi all,

First submission to this list, so go easy on me.  I've the following
file (temp.cc):

#include <stdio.h>
#include <unistd.h>
#include <pthread.h>


int foo_static[] = {0, 0};


void* run(void* arg)
{
    int foo[] = {0, 0};
    printf("foo[2]: %d\r\n", foo[2]);
    printf("foo_static[2]: %d\r\n", foo_static[2]);
}


int main(void)
{
    pthread_t thread;
    pthread_attr_t attributes;
    pthread_attr_init(&attributes);
    pthread_create(&thread, &attributes, &run, 0);

    for (;;) {
        sleep(1);
    }

    return 0;
}

And I conduct the following transaction with my computer:

beef@nono:~$ set | fgrep MUDFLAP
MUDFLAP_OPTIONS='-mode-check --viol-segv'
beef@nono:~$ g++ -fmudflapth -fmudflap temp.cc -lmudflapth -lpthread
beef@nono:~$ ./a.out
foo[2]: -1209174832
*******
mudflap violation 1 (check/read): time=1151571511.357805 ptr=0x80c9168
size=12
pc=0xb7ec5a30 location=`temp.cc:13 (run)'
      /usr/lib/libmudflapth.so.0(__mf_check+0x50) [0xb7ec5a30]
      ./a.out(_Z3runPv+0x1f0) [0x8048be2]
      /usr/lib/libmudflapth.so.0 [0xb7ed3a53]
Nearby object 1: checked region begins 0B into and ends 4B after
mudflap object 0x80ca2e8: name=`temp.cc:6 int foo_static [2]'
bounds=[0x80c9168,0x80c916f] size=8 area=static check=3r/0w liveness=3
alloc time=1151571511.288349 pc=0xb7ec53f0 thread=3083183808
number of nearby objects: 1
Segmentation fault
beef@nono:~$

As you can see, only the indexing of foo_static caused a violation, not
the indexing of foo.  If foo were in main()'s stack, however, it would
have caused a violation.

Versions of stuff as follows:

beef@nono:~$ gcc --version
gcc (GCC) 4.1.2 20060613 (prerelease) (Debian 4.1.1-5)
Copyright (C) 2006 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is
NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR
PURPOSE.

beef@nono:~$ uname -a
Linux nono 2.6.15.6 #2 Thu Mar 9 17:34:11 EST 2006 i686 GNU/Linux
beef@nono:~$

I've tried to compile 4.1.0 from a GNU mirror, but with the same result.

Can anyone shed some light on this?

Thanks,

Stuart

-- 
Stuart Clarke <sjclarke@tpg.com.au>