The problem was that $secondhalflen is actually less than
string-length($secondhalfword) in case the number of chars in word is
even. This leads to $firsthalflen being calculated too long, resulting
in a duplicated character.

known to not completely break in Solaris

closes #25

https://bz.apache.org/bugzilla/show_bug.cgi?id=60172

suggested patch by Jeffrey Adamson

this is the combined patch or #18 which couldn't be applied via `git am`