Reading the logs is an obvious suggestion. CGI::Carp misses some 500 stuff. Try adding this which is pretty robust:

#!/usr/bin/perl

BEGIN {
    $|++;
    print "Content-type: text/html\n\n";
    # this may also help as it will capture STDERR in the browser
    open STDERR, ">&STDOUT";
    use CGI::Carp 'fatalsToBrowser';
}

print STDERR "<p>Foo</p>";
print STDOUT "<p>Bar</p>";

# catches almost all errors one way or another
# if ( $syntax_error ) {

END { print STDERR "Exiting!" }
[download]

cheers

tachyon

s&&rsenoyhcatreve&&&s&n.+t&"$'$`$\"$\&"&ee&&y&srve&&d&&print