First, when you are debugging something like this, print the actual return code that Perl sees! You made an erroneous assumption that 256 would be -1! That is true in 8 bit 2's complement math. But that's not true here!

I personally would always prefer assigning a lexical var for the return code instead of using some "magic" Perl variable ($!). my $rc = system($cmd); #use $rc and perhaps $! also, but $rc is "the error code from system"

Ysth is completely correct - get the Perl process to change its directory, do not use a shell. That shell is a separate process.

But a main point here is that when things go "bad", print more stuff, like the return code from system(). Then you see that 256 != -1, etc.

#!usr/bin/perl -w
use strict;

my $cmd = "cd /cat/bat";

my $rc = system($cmd);
print "rc = $rc\n";

#PRINTS:
#The system cannot find the path specified.
#rc = 256

$cmd = 'cd C:\temp';
$rc = system($cmd);
print "rc = $rc\n";

#PRINTS:
#rc = 0
[download]

Update: I would also add that most of the time an O/S error code is "0", zero upon success and non-zero when it fails. As with all things with software, this is not true always, but it is often true. So instead of checking for a comparison with -1 (or 256), better is usually is, if ($x !=0) for the error condition.