I believe this also fixes some other bugs regarding /i matching and Unicode properties, but RT has been down for a while, preventing me from finding out. This patch is also the basis for a number of other bug fixes that should follow in short order. The problem is that the existing algorithm is defective, not catching a number of possible matches, as explained in the commit messages and code comments. This patch works by creating, at first need, a hash that quickly gives all the code points that fold to the one being examined. It then knows all the possibilities, and tries them, instead of giving up early. The hash is created from the probably already-read-into-memory folding swash.Thread Next