And are you saying that this is the exact code that produced the error message you showed us at the top of the thread?

Have you tried redirecting stderr to a file? You may need to inspect the error message for non-printable byte values (e.g. using od or xxd or some other hex-dump tool).

Does the Japanese version of Oracle actually support Japanese strings as column names in the database schema? What character encoding does it use? What version of perl are you using? How certain are you that the encoding used in the perl script matches what is used in the Oracle schema?