Re: javacc help needed

[Paul Cager <paul-debian@home.paulcager.org>]

Michael Koch wrote:
> Hello,
> On Mon, Nov 05, 2007 at 10:49:29AM +0900, Sanghyeon Seo wrote:
>> In the original modifed file, ParseException extends CSSException
>> which extends RuntimeException, so it is an unchecked exception. But
>> unmodifed ParseException extends Exception by default, which makes it
>> a checked exception, but methods throwing ParseException doesn't
>> declare to throw it, resulting in compile error.
> 
> Thats close to the worst things library developers can do...
> 
> This makes me really scared about this. Depending on such a library in a
> distro is like depending on dynamite to not explode.

I would say it's not _entirely_ the library developers' fault. JavaCC
has always had a rather inconsistent use of Exceptions - ParseException
(thrown for syntactic errors) extends Exception, but TokenMgrError
(thrown for lexical errors) extends Error. It's not surprising users get
in a pickle. Having said that, I would certainly think it would be
better to trap the (unchecked) TokenMgrError and convert it to a
(checked) ParseException.

(And sometimes it is quite appropriate for a ParseException to be
treated as a RuntimeError - e.g. if you are using the parser on
known-good input. I've not checked if that's the case for flute).

The next release of JavaCC should make things a little more logical:
https://javacc.dev.java.net/issues/show_bug.cgi?id=12


Thanks,
Paul.