You'd probably have to look at the JVM spec. I'd say the question would be
whether any instantiation of an array can be done at compile-time, or is it
always a run-time. If only at run-time (which my experimentation shows is
what javac seems to believe), then it wouldn't conform to the JLS and
wouldn't be legal. Another perspective is whether any array is ever a
"constant". I tend to doubt it. Again, the JLS would say that's illegal.
Rick
|---------+---------------------------->
| | Daniel Rall |
| | <dlr@finemaltcodi|
| | ng.com> |
| | Sent by: |
| | dlr@finemaltcodin|
| | g.com |
| | |
| | |
| | 02/22/2002 03:48 |
| | PM |
| | Please respond to|
| | rpc-dev |
| | |
|---------+---------------------------->
>----------------------------------------------------------------------------------------------------------------------|
|
|
| To: [EMAIL PROTECTED]
|
| cc:
|
| Subject: Re: Invalid static member of inner class
|
>----------------------------------------------------------------------------------------------------------------------|
I'm fairly certain that jikes just turns that into a constant. It's
odd that javac does not -- I wonder if there is a way to convince it
that it should be constant?
"EXT-Raiteri, Ashley L" <[EMAIL PROTECTED]> writes:
> Good call!
> Oneo of our guys here caught this as well, and I was wondering if anyone
> else would come up with it...
> ashley raiteri
>
> -----Original Message-----
> From: Rick Johnston [mailto:[EMAIL PROTECTED]]
> Sent: Friday, February 22, 2002 7:49 AM
> To: [EMAIL PROTECTED]
> Subject: Invalid static member of inner class
>
>
>
> In compiling the latest version of XmlRpcServer.java with javac (1.3.1),
I
> got a compiler error because of the static member 'EMPTY_BYTE_ARRAY' in
the
> inner 'Worker' class. Section 8.1.2 of the Java Language Spec 2nd ed.
does
> specifically state that "Inner classes may not declare static members,
> unless they are compile-time constant fields." Since EMPTY_BYTE_ARRAY is
> defined as "= new byte[0]", this is not a compile-time constant and is
> therefore invalid. Interestingly, the Jikes compiler does not complain
> about this (perhaps because it is also declared as final?).
Alternatively,
> if the Worker class was a Nested class but not an Inner class (i.e.,
Worker
> is 'static' - which doesn't seem appropriate), this would be legal.
>
> I plan to move EMPTY_BYTE_ARRAY back to the outer class in my copy. I
ask
> that someone checks this change into CVS as well.
