clean up decoding error reporting by counting lines instead of reading single characters to find the error position