I know you didn't ask, but some of the code used here is not very secure. Other parts can be written better (if you plan to make the script portable)

Here is an updated version:

#!/usr/bin/perl
use strict;         # amen
use warnings 'all'; # or use perl -w
use File::Basename;

$| = 1;

# :raw for brain-dead OSes (or use binmode).
open my $in, '<:raw', $ARGV[0];        # use $in not IN
open my $out, '>:raw', "differences";

while (<$in>) {
    chomp;
    ($file,$problem,@machines) = split(/,/, $_);

    ($name,$path,$suffix) = fileparse($file,"");
    print "$name -- $path -- $suffix\n";

    if($problem eq "different") {
        # get the file from the remote machine
        foreach $machine (@machines) {
            next if ($machine eq "");

            # consider using File::Temp here
            $tempfile = "/tmp/" . $name . "_" . $machine;

            # should consider using File::Remote here
            system("rcp", "$machine:$file", $tempfile) == 0
                or die "Execution (rcp) failed: $?";

            # why not use Text::Diff ?
            print "comparing $name from $machine\n";
            system("diff", $file, $tempfile) == 0
                or die "Execution (diff) failed: $?";

            print "(k)eep or (i)gnore this difference?\n";
            $resp = <STDIN>;
            print ">>$resp<<\n";
            if($resp eq "k") {
                print $out "$machine $file\n";
            }
        }
    }
}
[download]

Executing stuff is always insecure. You should either consider using taint mode (-T) or use Perl modules to perform the same task in most cases (in all cases in the example above)

Documents for these modules are here: File::Remote, File::Temp and Text::Diff. They are more or less simple to use, a lot safer and sometimes faster. (since they don't fork() and exec())

Update: Use chomp instead of chop.

I've been meaning to reply to this for a while. Most of the perl I write has to be backwards compatible to early 4.0 era versions of perl. Also, the code needs to be able to run on every unix platform we support. So its just not feasible to use CPAN modules for everything, becasue garanteed, it'll break somewhere. Coincidentaly, thats why I also have to write in pure sh shell....

---

I hate my sig

$resp = <>; is reading the next line in the next file in @ARGV, not from STDIN. You probably mean $resp = <STDIN>; there.

See "I/O Operators" in perlop to see exactly what <> is doing.

To clarify holo's response, <> will read from STDIN only when all the files on the commandline have been exausted. In scripts without arguments, this will always read from STDIN. But since your script takes arguments, it won't DWIM (at least, not right away...)


Who is Kayser Söze?

<> will not read from STDIN if there are arguments; not even when it runs out of files. Proof:

$ echo "one" > file1; echo "two" > file2
$ echo "three" | perl -e 'print while <>' file1 file2
one
two
[download]

This script requires at least one argument to work so I assume that unless @ARGV is cleared, <> will never read from STDIN.