Sorry for poor writing.
Because correct encoded string in my locale may looks like 'garbage' for most of you, I avoid to cut & paste exact script and output.
As you guess, "<UTF-8 string>" is a utf-8 encoded string in Japanese and output fine( because of combination of "use 'utf8'" and "binmode(STDERR , ':utf8')."
It may depend OS and locale, but for "ja_JP.UTF-8" locate on Ubuntu case, Perl generate language specific error messages.
Half of may question was ans wed by Anonymous Monk.
I just want to know how to get error string in C-locale
even if Perl-process start in non-C locale.