Offset in the header may be wrong if the source contains multibyte
characters.  Fix it by encoding the string to utf-8 before calculating
the offset.

For examaple:

>>> s='Exiting¡ ¬'
>>> len(s)
10
>>> len(s.encode('utf-8'))
12

Signed-off-by: Tzung-Bi Shih <penvirus@gmail.com>