Looking at the code near the line on which the error was reported, we find this set method in 5.6.1:

# IPC/Msg.pm lines 61-79
sub set {
    my $self = shift;
    my $ds;

    if(@_ == 1) {
        $ds = shift;
    }
    else {
        croak 'Bad arg count' if @_ % 2;
        my %arg = @_;
        my $ds = $self->stat
                or return undef;
        my($key,$val);
        $ds->$key($val)
            while(($key,$val) = each %arg);
    }

    msgctl($$self,IPC_SET,$ds->pack);
}
[download]

There is a bug in this code: on line 11 (which is line 71 of the original file), my $ds creates a new $ds variable which goes out of scope at the end of the else block. That is why the final line has an undefined value for $ds when it tries to invoke $ds->pack.

Removing the redundant "my" will fix that problem by ensuring that (as intended) the outer $ds variable is the one used, and indeed in later perls you'll find that exactly this change has been made: replacing

  my $ds = $self->stat
[download]

  $ds = $self->stat
[download]

If you don't have permission to make this fix to the code directly, and you can't convince the system administrator to apply the fix for you, there are a couple of ways you can get around it. One way is subclassing to replace the buggy routine:

  {
    package IPC::Msg::Bugfix;
    our @ISA = qw/ IPC::Msg /;
    sub set {
      # corrected version of set here ...
    }
  }
  # in your code
  my $msg = new IPC::Msg::Bugfix('24h', IPC_CREAT);
  # and continue as before
[download]

An alternative is to recode to avoid the path with the bug in it:

  # avoiding buggy multiple-arg set()
  # $msg->set('qbytes' => 32768); 
  my $ds = $msg->stat or die "stat: $!";
  $ds->qbytes(32768);
  $msg->set($ds);
[download]

(Please note, I haven't tested any of this code.)

Hugo