Re: More finalize woes

[Chris Gray]

On 5 Mar 2003, Tom Tromey wrote:

> >>>>> "Dalibor" == Dalibor Topic <address@hidden> writes:
> 
> >> java.util.zip.[De|In]flater both have an empty finalize method "for
> >> compatibility". Is this really necessary?
> 
> Dalibor> Yes, in order to have reflection on these methods
> Dalibor> working. A pretty good reason in my opinion.
> 
> I'm not so sure.  It seems to me that anybody using reflection has to
> be aware of, and understand, the Java binary compatibility rules.
> Implementing or not implementing finalize doesn't affect binary
> compatibility, so it should be a change we are free to make.

I think the main problem lies elsewhere. If you subclass a class which has 
a finalizer and your subclass needs to do some finalization of its own, 
then your finalizer needs to contain a call to super.finalize(). If the 
published API doesn't show a finalizer then you wouldn't include this 
statement. That's not an argument against trivial finalizers provided for 
"compatibility", but it it's a strong argument agains non-trivial 
finalizers where the public API has none: there's areal likelihood that 
someone will subclass your class and fail to invoke your finalizer.

Regards

-- 

Chris Gray

VM Architect, ACUNIA