Patch: Java -vs- Jacks

[Alexandre Petit-Bianco]

Tom Tromey writes:

> * `\uufeff' is valid syntax and is equivalent to `\ufeff'.

Well, your patch fixes the \uu problem in general, right?

> * In some situations gcj didn't recognize \u000a and \u000d as newlines.
>   I cleaned up newline handling quite a bit (imnsho)

Yeah. What you did seems a much better way to match CR or LF or CRLF
or CRCRLF as one line terminator.

> This patch isn't quite complete.  I believe the character ranges in
> _JAVA_LETTER_OR_DIGIT_P are out of date.  I think this particular
> macro should be automatically generated by running a script over the
> Unicode tables.  However, I haven't done this work.  Instead I am
> going to submit a PR for this.

It might be a good idea if you can extract the same information than
what Sun specified in 20.5.18 (20.5.18 is what the lexer should comply
with if I read the lexer documentation well.)

> Ok to commit?

Have you tried the XML xp module, especially
/com/jclark/xml/tok/Encoding.java?  I'll try it and let you know,
hopefully shortly enough.

> I think a couple are related to the "package out of date" problem
> that you recently patched, but I'm not certain.

I don't understand what you're referring to. Do you means the
bytecode/source file timestamp discrepancies that we sometimes get
wrong because gcj repeatedly call jc1 on single files?

./A