system.privilege.taskport succeds but task_for_pid returns !=KERN_SUCESS

Hi all!

I wrote a small test application on Leopard 10.5.8 that does  
task_for_pid after authorization which always fail no matter what I  
do. I did a codesign and authorization itself goes fine:

[ivan]~/hoby/debuger > gcc -m64 syscall2.c -o sysc2 -framework  
Security 
                                                                       
[21:35]
[ivan]~/hoby/debuger > codesign -s gdb ./ 
sysc2 
                                                                                               [21 
:36]
[ivan]~/hoby/debuger  
 > 
                                                                                                                       [21 
:36]
[ivan]~/hoby/debuger  
 > 
                                                                                                                       [21 
:36]
[ivan]~/hoby/debuger > ./ 
sysc2 
                                                                                                               [21 
:36]
Jan 22 21:36:25 devil com.apple.SecurityServer[26]: Succeeded  
authorizing right system.privilege.taskport by client /Users/ivan/hoby/ 
debuger/sysc2 for authorization created by /Users/ivan/hoby/debuger/ 
sysc2.
task_for_pid failed.
[ivan]~/hoby/ 
debuger 
 > 
                                                                                                                       [21 
:36]
[ivan]~/hoby/debuger > ./ 
sysc2 
                                                                                                               [21 
:36]
task_for_pid failed.
[ivan]~/hoby/ 
debuger 
 > 
                                                                                                                       [21 
:36]
[ivan]~/hoby/debuger > sudo ./ 
sysc2 
                                                                                                          [21 
:36]
wait pid: 1111
RIP: 7fff5fc01010
[ivan]~/hoby/debuger>

I really started loosing hair over this so please help (running sudo ./ 
auth works fine and task_for_pid returns KERN_SUCCESS)

This is an example source code:

-----Code-----

#include <stdio.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/ptrace.h>
#include <assert.h>
#include <mach/mach.h>
#include <errno.h>
#include <stdlib.h>
#include <Security/Authorization.h>

int acquireTaskportRight() {
   AuthorizationRef authorization;
   OSStatus status = AuthorizationCreate (NULL,  
kAuthorizationEmptyEnvironment, kAuthorizationFlagDefaults,  
&authorization);
   if (status != 0) {
     fprintf(stderr, "Error creating authorization reference\n");
     return -1;
   }
   AuthorizationItem right = { "system.privilege.taskport", 0, 0 , 0 };
   AuthorizationItem items[] = { right };
   AuthorizationRights rights = { sizeof(items) / sizeof(items[0]),  
items };
   AuthorizationFlags flags = kAuthorizationFlagInteractionAllowed |  
kAuthorizationFlagExtendRights | kAuthorizationFlagPreAuthorize;

   status = AuthorizationCopyRights (authorization, &rights,  
kAuthorizationEmptyEnvironment, flags, NULL);
   if (status != 0) {
     fprintf(stderr, "Error authorizing current process with right to  
call task_for_pid\n");
     return -1;
   }
   return 0;
}


void check(int cond, char* msg) {
   if (!cond) {
     printf("%s\n", msg);
     exit(-1);
   }
}

int main() {
   pid_t child = fork();

   if (child == 0) {
     /* PT_TRACE_ME will stop the process after the execl is executed  
and allows the parent
        to take control. */
     check(!ptrace(PT_TRACE_ME, 0, 0, 0), "PT_TRACE_ME failed.");
     execl("./hello", "hello", NULL);
   } else {
     /* Get the task for this pid. Seems to require superuser  
privileges or some gid hack.
        Go Apple! */
     mach_port_t task;

     check(acquireTaskportRight()==0,"acquireTaskportRight failed");

     check(task_for_pid(mach_task_self(), child, &task) ==  
KERN_SUCCESS, "task_for_pid failed.");

     /* Get the list of threads in that process (we expect one thread  
exactly.) */
     thread_act_port_array_t threadList;
     mach_msg_type_number_t threadCount;
     check(task_threads(task, &threadList, &threadCount) ==  
KERN_SUCCESS, "task_threads failed.");
     check(threadCount == 1, "task has more than one thread.");


       /* Wait for updates from the child process we are tracing */
       int status;
       pid_t w = wait(&status);
       printf("wait pid: %d\n",w);


       /* Read the register state of the child via Mach (since we are  
single-stepping the child is guaranteed to be suspended at the moment.  
*/
       x86_thread_state64_t state;
       mach_msg_type_number_t stateCount = x86_THREAD_STATE64_COUNT;
       check(thread_get_state(threadList[0],
			     x86_THREAD_STATE64,
			     (thread_state_t)&state,
			     &stateCount) == KERN_SUCCESS, "thread_get_state failed.");

	printf("RIP: %lx\n",state.__rip);




       check(ptrace(PT_CONTINUE, child, (char*)1, 0) == 0,  
"PT_CONTINUE failed.");
     }

}

-----Code-----

Best Regards,
Ivan