gcj and jikes

Jeff Sturm jsturm@one-point.com
Mon May 5 18:50:00 GMT 2003


On Mon, 5 May 2003, [iso-8859-1] Joerg Brunsmann wrote:
>
> > As a layman, I certainly find the Jikes source code to be far
> > more readable and understandable than the parse.y in the GCJ
> > front-end, but that could simply be because I haven't had any
> > experience in writing parsers beyond my undergraduate days.
>
> You're not alone. That's exactly one of the reasons why it does
> make sense to switch the front end.

It's been months now since I've looked at parse.y, but from my
recollection the actual grammar is quite small compared to the code doing
error handling and semantic analysis.  Those latter two may be some
of the most difficult code in the frontend.

Using a non-yacc parser should help with error recovery, and the semantic
code may be rewritten eventually to support the tree-ssa efforts.
Regardless, this is a big job, and unless a lot of volunteer time becomes
available, adopting the jikes frontend in some capacity makes a lot of
sense to me.

Jeff




More information about the Java mailing list