Hi, Jonathan Sauer wrote:
Hello,
Am I doing something wrong? You are missing two things:
* the \lccode of ü has to be nonzero for it to be a valid word constituent
Oh. I thought since hyphenation has been completely been revamped, the requirement of a non-zero \lccode has been removed. Especially
It actually was gone for a while (\lccodes were not in the new hyphenation codebase in the initial rewrite), but then I realised that that gives big problems with words followed by punctuation. So I had to reinsert the code in the hyphenation algorithm.
since I did not get an error message (IIRC, the original TeX complained about this, at least in \pattern).
I never bothered to restore the error (or at least not for \hyphenation), because it was not all that useful an error to begin with: it is only a potential error at this point, a following bit of code can still change the \lccodes if it wants to (even revert to zero). A better solution for this 'what is a word' problem would be nice. Perhaps one will be offered by the Google Summer of Code; one of the proposed projects is "Better unicode support".
* The font cmr10 doesn't have an ü glyph
Of course. But this should not affect the possible hyphenation points, or should it?
No, it shouldn't. Best wishes, Taco