"...but it never seems to come back on the screen inverted."

At first glance, I'd guess the problem is the $/ = CRLF, which is done in the server, but not the client. It might just work if you were to comment that out... You'd have to do that on both ends if you were going to do it at all.

$/ is documented on the perlvar page.

jettero identified the problem correctly. However, setting $/ = CRLF on the client side won't help, as it's the input record separator, yet the \r\n (CRLF) needs to be sent on output, i.e. at the end of the $msg_out string in the statement print $socket $msg_out; .

Anyway, the easiest way to get this demo to work is to remove the $/ = CRLF; in the server code. Alternatively, you could add $msg_out =~ s/\n/\r\n/; right before the print $socket $msg_out; . (On unix, the string returned by STDIN->getline will be terminated by just \n, not \r\n.)

With this modification, you should see (assuming the server is running):

$ ./tcp_echo_cli2.pl localhost 2007
foobar
raboof
[download]

BTW, a good way to debug things like these yourself is to insert simple print ... statements to find out exactly where things are not working as expected. (In this case, you'd have seen that it doesn't get past the while (<$session>) { line in the server...)

I'm also under the impression that changing $/ might not ever help with this since the \n is converted on the way in from a file handle. The thing I don't know is if it converts "\x0d\x0a" to a \n for sockets. I suspect it does not.

-Paul

... changing $/ might not ever help with this since the \n is converted on the way in from a file handle.

Automatic line ending translations would happen if the crlf PerlIO layer were in effect. On Windows, this layer is by default pushed on the PerlIO layer stack for every filehandle (and, I suppose, for sockets, too — not 100% sure, though¹).  However, as the OP is trying this on Ubuntu, there is no crlf layer (unless explicitly requested), and thus no CR/LF translations.

___

¹ ...sorry, can't verify without rebooting into Windows, which would be too much fuss at the moment :)

Maybe someone else is willing to just run the example on Windows, and put a

print join(" ", PerlIO::get_layers($session)), "\n";
[download]

right after the my $session = $sock->accept; in the server code. This would show which layers are in effect.

Try to work your network programs so that they will continue to work if the client has the wrong line terminator, but always send "\r\n" as your own line terminator. This is pretty typical of the network ASCII protocols including HTTP.

This means that you probably want to strip any returns with tr/\r//d or such, in both your clients and your servers, before doing any other fussy parsing work (even just reversing the characters in the meat of the string).

You can play with $/ if you want to make chomp() smarter, for example, but it seems more flexible to specifically and explicitly work with newlines handling in your own code.