@DavidCarlisle \ifcat A\scantextokens\expandafter{\Uchar`A}\message{YES}\else\message{NO}\fi

\def\goodUchar#1{\scantextokens\expandafter{\Uchar`#1}}

@egreg yes but that means I need to get scantextokens into xetex:( (and with that you can implement Uchar as \scantexttokens{^^^^^\someexpandablehexmacro#1}

@DavidCarlisle Does also your version of \Uchar always returns catcode 12?

@egreg yes I thought it was a bug (as I'd alreaqdy written a test that it failed) but then I saw luatex did same

@DavidCarlisle Oh, well! :)

@egreg thinking of suggesting that as is then also \Ucharcat 100 11 to make a token with charcode 100 mand catcode 11

@DavidCarlisle Which would be very nice.

@egreg need to understand how \Umathchar and friends are set up to parse multiple integers