Hi all! I ended up reformulating the code as a test script: https://github.com/kjetilk/p5-html-html5-parser/blob/master/t/rt-96399.t This is the way I thought it should be, but with Greg's patch, the test intended to check the UNICODE char still fails for me... It may well be that the test is wrong, please poke at it.... :-) Cheers, Kjetil