works for me

text-to-html.js

function convert(text) {
  var tokens = text.split('\n'), i, token, src = [];
  for (var i=0; i < tokens.length; i++) {
    var token = tokens[i];
    token && token.charAt(0).match(/\w/) ? (function () {
      src.push('<p>' + token + '</p>');
    }()) : (function () {
      token && token.match(/^\s*<(a|b|i|strong|em|cite)([ \=\'\"\w\-\/\.\:]+)*>/i) ?
        (function() {
          src.push('<p>' + token + '</p>');
        }()) :
        src.push(token);
    }());
  }
  text = src.join('\n');

  var r = /<pre><code>([\s\S]+?)(?=<\/code><\/pre>)/g;
  text = text.replace(r, function (m, code) {
    return '<pre><code>' + code.replace(/<p>([\s\S]+?)<\/p>/g, '$1');
  });
  return text;
}

There are 3 immediately-invoking-functions that seams useless. Inlining them would improve performance.
Also code is not DRY.