I think I understood at last. I wrongly guessed about the behavior of the regex-based implementation.
- "NU CP PO" matches the pattern. So "NU × CP × PO". - "CP PO" doesn't. Now default rule (LB31) comes; so "CP ÷ PO". These are what LineBreakTest.txt is saying and why the implementation notes are added on it, aren't they? Thank you all for your patience. I would have been still confused without your posts.
