In stomp.js, the Frame.prototype.toString function sets the content-length to the number of characters in the body. This is only valid if there is one byte per character.

I added the following function to calculate the number of bytes in the body:

function countBytes(string) {
  var escapedStr = encodeURI(string);
  if (escapedStr.indexOf("%") != -1) {
    var count = escapedStr.split("%").length - 1;
    if (count == 0) count++;
    var tmp = escapedStr.length - (count * 3);
    return count + tmp;
  }
  else return escapedStr.length;
}

and changed

lines.push("content-length:" + ('' + this.body).length);

to

lines.push("content-length:" + countBytes('' + this.body));

Maybe there is a cleaner way of doing this. But this change worked for me.
There is possibly more issues of this type in stomp.js. The .length gives the number of characters and not the number of bytes in a string.