On Wed, 6 Mar 2002, Peter Donald wrote:

> > What happens at compile time to support inner classes is very tricky and
> > unexpected ( by most people ). The compiler removes 'private' and adds
> > parameters to methods  - the first one does have security implications
> > ( not an issue for ant ). I think it is far better to avoid the magic and
> > pass the parameters explicitely - and as a benefit the code will compile
> > and work with more compilers.
> 
> Yep. I generally have have a no-innerclasses policy unless they are static 
> and don't access any variables from parent class. Too many things that people 
> don't understand - I think your only the 4th person who has ever struck me as 
> knowing how inner classes really behave (and 2 of the others were JVM 
> implementors).

Well, thanks - but I never claimed I _understand_ how inner classes 
work :-) ( or why they were designed this way, or why people are using 
them ).

I assume jikes and gcj developers are in the same situation, not 
understanding very well how this is supposed to work ( since both still 
have problems dealing with them :-). That's why it's better to avoid
non static inner classes.

Costin






--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to