Still me, still working on my SMF stuff. And I found another problem. But it is likely not really with RecordClassGenerator, but with the characters valid in a Java name. Some of the variable in the SMF DSECT for record types 72, 74 and 113. The names end in a # (hash sign, pound sign, octothorpe). I don't know what could be done about this. Modifying the IBM macro is not an option!
OK, what I did was open up each Java program which did not compile and changed the # to a _. They then compiled. I'll figure out some way to automate this, I guess.
That's a good question, but I'm not really a very good Java programmer. You might consider mapping # to _hash_ and @ to _at_ or something similar to that. The only problem which might occur in any case would be a name collision.