Here's two ways of doing it:

while () {
    $_ = substr($records, $p, index($records, "\n", $p) - $p);
    last if (!$_); $p += length($_) + 1;
    # do whatever with data now in $_
}

while ($records =~ /(.*?)\n/g) {
    # do whatever with data now in $1
}
[download]

I benchmarked both for 10,000 iterations on a 1,000 record string. The substr took 22 seconds vs 46 seconds for the regex. Both assume there's a \n on the end of the string.

Using split directly cuts it to 10 seconds, but requires about twice the memory:

for (split(/\n/, $records)) {
    # do whatever with data now in $_
}
[download]