Re: [aspectj-dev] Declaring an inner anonymous class in an ITD method

[Simone Gianni <simoneg@xxxxxxxxxx>]

Hi Andy,
bug is filed : https://bugs.eclipse.org/bugs/show_bug.cgi?id=251326

You'll also find a patch that adds a testcase to aspectj test suite.

Simone


Andy Clement wrote:
> Hi,
>
> If AspectJ generated code that caused an IllegalAccessError then that
> is a bug, please open it.
> It should either say 'no you cant do that' at compile time or generate
> the right code.
> A testcase in the bug report would be great :)
>
> cheers,
> Andy.
>
> 2008/10/15 Simone Gianni <simoneg@xxxxxxxxxx <mailto:simoneg@xxxxxxxxxx>>
>
>     Hi all,
>     I have the following situation. I'm adding new methods to an existing
>     class using ITDs. These methods are installing some listeners, and
>     as it
>     is common they use anonymous inner classes, in the form of new
>     MouseListener() { ... }. AspectJ correctly creates these inner classes
>     in the package of the destination class. Unfortunately however, these
>     classes are package protected, and AspectJ generates ITD methods
>     in the
>     aspect class that call the TargetClass$1.<init>, causing an
>     IllegalAccessError.
>
>     Moving the aspect class to the target package resolves the problem,
>     cause anonymous inner classes are declared as package protected so
>     they
>     are accessible as long as the aspect is in the same package.
>
>     Is this a bug? I know this is a strange and less than usual situation,
>     but isn't this the role of the user? put as much fantasy and
>     creativity
>     as needed to find those usecases the original developers never dreamed
>     about? :)
>
>     If this is a bug, solutions could be :
>     - Declare anonymous inner classes declared inside an ITD method as
>     public to avoid this problem. I know this is not what standard javac
>     does, but will the JVM complain? This could be the simples solution.
>     - AspectJ correctly creates inner classes in the target package,
>     couldn't it also generate a "bridging class" in the target package and
>     wrap calls from the original ITD method (that resides in the aspect
>     class) to the inner classes via the bridging class?
>     - In both cases, this situation can be at least reported as a
>     warning or
>     an error at compile time.
>
>     Simone
>
>     P.S. Please note that another (tracing) aspect is also matching these
>     methods and using an around advice. I supposed, since the synthetic
>     methods that perform the call causing the error were named
>     aroundBody<number>, that the other aspect was playing a role here.
>     Anyway, removing the around aspect didn't change the situation.
>     The ITD
>     methods still reside in the aspect and still call the package
>     protected
>     <init> of the inner classes declared in the target class package, so
>     still raising the error.
>
>     --
>     Simone Gianni            CEO Semeru s.r.l.           Apache Committer
>     MALE human being programming a computer   http://www.simonegianni.it/
>
>     _______________________________________________
>     aspectj-dev mailing list
>     aspectj-dev@xxxxxxxxxxx <mailto:aspectj-dev@xxxxxxxxxxx>
>     https://dev.eclipse.org/mailman/listinfo/aspectj-dev
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> aspectj-dev mailing list
> aspectj-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/aspectj-dev
>   


-- 
Simone Gianni            CEO Semeru s.r.l.           Apache Committer
MALE human being programming a computer   http://www.simonegianni.it/