nick, have you tried asking on pax wicket mailing list? those guys use
wicket with osgi all the time. perhaps they have a clean solution.

-igor

On Fri, Jun 20, 2008 at 6:38 AM, Nick Giles <[EMAIL PROTECTED]> wrote:
> We're using Wicket within an OSGi evironment (Equinox 3.3.0), and have run 
> into an issue with MarkupResourceStream. It is given a class the markup is 
> associated with, but it stores that class using only its name. When accessed, 
> it then tries to instantiate that class through Classes.resolveClass.
>
> Unfortunately, despite having our own IClassResolver, without giving in to 
> some bad OSGi practice, it is not permissible for the required class to be 
> visible, only the interfaces. This occurs because of the (helpful) classpath 
> segregation that OSGi encourages, such that a bundle only imports the 
> packages it actually needs, and is not allowed access to anything else.
>
> I have logged this as a bug 
> (https://issues.apache.org/jira/browse/WICKET-1681), but the conclusion there 
> was that there should be a workaround. Without either importing specific 
> implementation packages into the affected bundle (fragile, poor practice) or 
> allowing dynamic imports (evil hack to get around not knowing what classes 
> you'll load, effectively drops the benefits of OSGi's classloader 
> separation), I can't see a solution.
>
> Does anyone have any suggested solutions, or opinions on whether this should 
> be handled by Wicket or our client code?
>
> Thanks,
>
> Nick
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>

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

Reply via email to