waitpid PID,FLAGS
--
PID: -1(stands for any child process) | pid ( > 0 )
FLAGS: 0 (for a blocking wait) | WNOHANG (return 0 or -1 if no dead children)

waitpid($pid, 0) # blocking wait
waitpid(-1, WNOHANG) # no blocking wait

Blocking wait
--
reap should be processed as fork sequence: fork_child1/reap1 - fork_child2/reap2
But in real world, later fork may end earlier. blocking to wait it finish is NOT a good idea
so we can use $SIG{CHLD} to reap (no blocking wait)

Non blocking wait
--
$SIG{CHLD} = 'IGNORE'; # Children reaped by system
$SIG{CHLD} = 'DEFAULT'; # System defined
$SIG{CHLD} = &REAPER; # do REAPER if SIGCHLD catched

$SIG{CHLD} = &REAPER; # do REAPER if SIGCHLD catched
--
when child exit, child to send SIGCHLD to its parent
parent follow child and get its return value ($?)

waitpid is called from a SIGCHLD handler, to reap the children as soon as they die. And use the WHOHANG flag to make waitpid immediately return 0 if there are no dead children. So that the parent process knows which and when a child die. A loop process is used to avoid the race situation that two or more child processes reach to reaper, for normally kernel keeps track of underlivered signals using a bit vector, one bit per signal.

sub REAPER
{
  my ($child_pid,$ex);
  while (($child_pid = waitpid(-1,WNOHANG) ) > 0 ) {
    $ex = $?;
    if( $ex ) {
      $ex = $ex/256;	#exit code of waitpid is a 16-bit value
			#high-order 8 bits:exit code from exit().
			#low-order 8 bits are zero if the process exited normally, 
			#or encode the signal number that killed the process, and whether it 
			#dumped core or not (and if it was signalled, the high-order bits are zero).
    }
    if( $child_pid == -1 ) {
      printf( "reaped a child's system command\n" );
    }
    else {
      printf( "PID $child_pid: " . ($ex ? " (exit $ex)." : '.')."\n" );
    }
  }#end while
  $SIG{CHLD} = \&REAPER;
  return;
}