It's a little difficult to help without knowing what exactly you've done that results in the error. Or more generally, what exactly you want to achieve. From your initial description I gather you want to make a redirection in case a specific substring is found in the URL. But I don't yet understand how the "need to pass @ARGV" issue is related to that...

As for the "Can't use an undefined value as a symbol reference", the error happens when the module is trying to use logfh the first time (line 765) — which is because you've passed <LOGFILE> (presumably), instead of *LOGFILE, i.e. it should be

    logfh => *LOGFILE,
[download]

<LOGFILE> reads lines from the file handle (which doesn't make sense here), while *LOGFILE refers to the file handle itself.   In case you find the glob syntax (*...) curious, you can also use a lexical file handle

my $fname = "/var/log/repeater.log";
open my $logfh, ">>", $fname or die "couldn't open '$fname': $!";
...
    logfh => $logfh,
[download]

Hi again. I've managed to start a proxy with your help. And now it even serves some requests ( it's some log entries):

[Sun Jan  8 10:38:53 2012] (1554) SOCKET: New connection from 127.0.0.
+1:49293
[Sun Jan  8 10:38:53 2012] (897) PROCESS: Forked child process 1554
[Sun Jan  8 10:39:01 2012] (1554) REQUEST: GET http://www.google.com/s
+orry/?continue=http://www.google.com/search%3Fq%3Dcar+rent
[Sun Jan  8 10:39:01 2012] (1554) REQUEST: Via: 0.9 domU-00-05-34-21-B
+5-AC.compute-1.internal (HTTP::Proxy/0.25)
[Sun Jan  8 10:39:01 2012] (1554) REQUEST: X-Forwarded-For: 127.0.0.1
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: 503 Service Unavailable
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: Cache-Control: no-cache, m
+ust-revalidate
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: Date: Sun, 08 Jan 2012 15:
+39:01 GMT
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: Via: 1.1 domU-00-05-34-21-
+B5-AC.compute-1.internal (HTTP::Proxy/0.25)
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: Server: HTTP server (unkno
+wn)
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: Content-Length: 2639
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: Content-Type: text/html; c
+harset=UTF-8
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: Expires: Mon, 01 Jan 1990 
+00:00:00 GMT
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: Title: http://www.google.c
+o.uk/search?q=car rent
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: X-Cache: MISS from domU-00
+-05-34-21-B5-AC.compute-1.internal
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: X-Cache-Lookup: MISS from 
+domU-00-05-34-21-B5-AC.compute-1.internal:3128
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: X-Frame-Options: SAMEORIGI
+N
[Sun Jan  8 10:39:01 2012] (1554) RESPONSE: X-XSS-Protection: 1; mode=
+block
[Sun Jan  8 10:39:01 2012] (1554) SOCKET: Connection closed by the pro
+xy
[Sun Jan  8 10:39:01 2012] (1554) PROCESS: Served 1 requests
[Sun Jan  8 10:39:02 2012] (897) PROCESS: Reaped child process 1554
[Sun Jan  8 10:39:02 2012] (897) PROCESS: 0 remaining kids:
[download]

But the link is not being transformed.

I guess you need to unescape the URI in $new_location, i.e.

use URI::Escape;

my $location = "http://www.google.com/sorry/?continue=http://www.googl
+e.com/search%3Fq%3Dperlmonks";

my $new_location = $location;
$new_location =~ s/.*\/sorry\/\?continue=(.*)/$1/;
$new_location = uri_unescape($new_location);

print "   $location\n=> $new_location\n";

__END__
   http://www.google.com/sorry/?continue=http://www.google.com/search%
+3Fq%3Dperlmonks
=> http://www.google.com/search?q=perlmonks
[download]

That said, I don't get any captchas or sorries from Google (is this still your squid's sorry thing?), so I couldn't test it as is... but when I submit an incorrect search, e.g.

    http://www.google.com/search?=perlmonks
[download]

(i.e. missing q, as in your OP), Google redirects me to http://www.google.com/webhp, so I used that response to test, and when I rewrite the Location to

    if ($location =~ m|/webhp$|) {
        my $new_location = "http://www.google.com/search?q=perlmonks";
        $headers->header( Location => $new_location );
    }
[download]

things work fine for me, while with the escaped http://www.google.com/search%3Fq%3Dperlmonks I do get a 404 Google custom error page.