[antlr-interest] pushed antlr build with tree rewriting and composite grammars (java)

[Terence Parr]

> It seems that ANTLR is less happy about me trying to parse 'fragments'
> now.
>
> i.e. if I set up the input "test", and then invoke
>
>  MyParser.identifier()
>
> the generated parser now says
>
>  NoViableAltException(-1!=[()* loopback of ..blah..]
>
> whereas this seemed to work with the earlier ANTLR releases, at least
> for some rules.
>
>
> Is this sort of usage just not intended to work?
>
> In the cases where I had problems in the past, I'd just append some
> other token to the language fragment I was attempting to parse.
>
> Is there a 'correct' way to fix this?  e.g. putting EOF all over the
> place?

fragment rules are meant to be called only from other lexer rules; any  
other usage is "off label" ;)

Ter