[
https://issues.apache.org/jira/browse/OWB-1416?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17650004#comment-17650004
]
Mark Struberg commented on OWB-1416:
------------------------------------
I agree with Romain that a) the spec is not clear and b) our approach is much
more safe.
Consider some code accessing the Contextual Reference via Instance<T> or
BeanManager.getContextualReference() etc. In that case there is not even an
injection point. Does this mean Weld will serve broken CDI beans out to user
code? Sounds really wrong to me.
> Possible misintepretation of spec regarding Unproxyable bean types
> ------------------------------------------------------------------
>
> Key: OWB-1416
> URL: https://issues.apache.org/jira/browse/OWB-1416
> Project: OpenWebBeans
> Issue Type: Bug
> Reporter: Alexander Larsen
> Priority: Major
>
> OWB seems to throw an exception for all unproxyable normal scoped beans. I
> think that this might be incorrect.
> The
> [specification|https://docs.jboss.org/cdi/spec/2.0/cdi-spec.html#unproxyable]
> says the "A bean type must be proxyable if an injection point resolves to a
> bean", not that all the types of the bean must be proxyable. In other words,
> as long as the bean is a legal bean, and all injection point resolving to
> this bean is a proxyable type - no exception should be thrown.
> In the part about [contextual
> references|https://docs.jboss.org/cdi/spec/2.0/cdi-spec.html#contextual_reference],
> there is further indications that unproxyable types should be allowed in the
> set of types for the bean. It's only when you try to get a reference(injected
> or by bean manager) to an unproxyable type, and the bean must be proxied
> (normal scoped, intercepted or decorated) an exception should thrown.
> Also, the [Weld user guide suggests introducing an interface as a solution to
> having an unproxyable
> bean|https://docs.jboss.org/weld/reference/latest/en-US/html_single/#_client_proxies].
> The current OWB implementation makes a pattern of having an interface and
> (one or more) implementation class with final fields/methods somewhat
> difficult :)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
