This is an update for
PDFBOX-420 filed by Takashi Komatsubara.
In this patch, if "Identity-H" is used as encoding of font and the font doesn't supply TO_UNICODE table, then encoding name is generated from CID information (Registry and Ordering). This idea is borrowed from pdfminer, an another PDF library written in Python. I don't see any test failures with this patch.
I have published this patch last year, and got some good feedbacks from Japanese users.