[weld-dev] javax.el.ExpressionFactory has final methods! CDI doesn't allow that
Martin Kouba
mkouba at redhat.com
Thu Oct 22 02:22:43 EDT 2015
Well, might not be nice but we cannot guarantee backward compatibility
for such cases. CDI defines a contract. If you break the contract then
the deployment fails. If we allow to break the contract then the
contract is useless and we could allow anything...
Anyway what's the use case here? Is there a normal-scoped producer?
I believe that having such a producer for a third-party type other than
interface is not a good practice. Is there a normal-scoped bean class
extending Hashtable? Not a good practice either. If so, the bean author
should reduce the bean types by means of @Typed and allow the bean
clients to only work with interfaces.
Martin
Dne 22.10.2015 v 00:28 Emily Jiang napsal(a):
> What about legacy applications running on new JDK7? The only workaround
> is that the customers have to update the app. Otherwise, their app won't
> start. This is not nice.
>
> On Wed, Oct 21, 2015 at 2:57 PM, Martin Kouba <mkouba at redhat.com
> <mailto:mkouba at redhat.com>> wrote:
>
> Dne 21.10.2015 v 13:54 Emily Jiang napsal(a):
>
> doh. You were right that the OWB-616 was for a different issue
> rather
> than the one I am interested.
>
> Do you have any suggestions about working around the unproxiable
> Hashtable issue?
>
>
> Well, the spec only requires the injection points to be validated,
> or rather the required bean types to be proxyable. So if you
> inject/use Map interface instead of Hashtable, the validation should
> pass.
>
>
> On Wed, Oct 21, 2015 at 12:38 PM, Martin Kouba
> <mkouba at redhat.com <mailto:mkouba at redhat.com>
> <mailto:mkouba at redhat.com <mailto:mkouba at redhat.com>>> wrote:
>
> Dne 21.10.2015 v 13:20 Emily Jiang napsal(a):
>
> Thanks Martin!
> The new final method on the Hashtable is
>
> final boolean initHashSeedAsNeeded(int capacity)
>
>
> I see. This package-private method was added in JDK7.
>
>
> The change went as part of the following bug fix:
>
> [1]
> http://bugs.java.com/bugdatabase/view_bug.do?bug_id=8006593
>
> As for the OWB-616 jira, I did not look at the actual
> fix, but I
> assume
> the fix by the following comments. I might be wrong here.
>
> <https://issues.apache.org/jira/browse/OWB-616#>
> Mark Struberg
>
> <https://issues.apache.org/jira/secure/ViewProfile.jspa?name=struberg>
> added a comment - 01/Oct/11 19:12
>
> I now changed the checks to allow private final and
> public/protected
> static final methods.
>
>
> Mark does not talk about non-static package-private
> methods, right?
> He talks about private final and public/protected static
> final =>
> should work in Weld too.
>
> If non-static package-private methods are allowed than the
> spec is
> violated. Unfortunately, it seems there's no tck test for this.
>
>
>
> Thanks
>
> Emily
>
>
> On Wed, Oct 21, 2015 at 11:50 AM, Martin Kouba
> <mkouba at redhat.com <mailto:mkouba at redhat.com>
> <mailto:mkouba at redhat.com <mailto:mkouba at redhat.com>>
> <mailto:mkouba at redhat.com <mailto:mkouba at redhat.com>
> <mailto:mkouba at redhat.com <mailto:mkouba at redhat.com>>>> wrote:
>
> Hi Emily,
>
> commments inline.
>
> Dne 21.10.2015 v 11:02 Emily Jiang napsal(a):
>
> CDI specification does not allow proxying a
> class with
> non-private final
> methods. The java.util.Hashtable class has a
> non-private final method added to the class in
> later
> versions
> of Java, so a CDI application that previously
> worked
> may break
> when updating the Java level.
>
>
> Just for the record: what's the name of the method?
>
>
> This issue was logged in CDI-527.
>
> OpenWebBeans fixed this via the jira
> (https://issues.apache.org/jira/browse/OWB-616).
>
>
> OWB-616 does not fix CDI-527 but CDI-159, i.e. a
> private final
> method does not cause a deployment problem. This
> works in
> Weld too.
> On the other hand, CDI-527 is still an open issue
> so we
> can't simply
> fix it. In theory, we could add a new feature of a
> non-portable
> mode. But non-portable mode is not intended to be
> commonly
> used.
> It's kind of a workaround.
>
>
> Can Weld fix this in the 2.3 or 2.2 trunk?
>
>
> --
> Thanks
> Emily
> =================
> Emily Jiang
> ejiang at apache.org <mailto:ejiang at apache.org>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>>>
> <mailto:ejiang at apache.org
> <mailto:ejiang at apache.org> <mailto:ejiang at apache.org
> <mailto:ejiang at apache.org>>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>>>>
>
>
> _______________________________________________
> weld-dev mailing list
> weld-dev at lists.jboss.org <mailto:weld-dev at lists.jboss.org>
> <mailto:weld-dev at lists.jboss.org <mailto:weld-dev at lists.jboss.org>>
> <mailto:weld-dev at lists.jboss.org
> <mailto:weld-dev at lists.jboss.org>
> <mailto:weld-dev at lists.jboss.org <mailto:weld-dev at lists.jboss.org>>>
> https://lists.jboss.org/mailman/listinfo/weld-dev
>
>
> --
> Martin Kouba
> Software Engineer
> Red Hat, Czech Republic
>
>
>
>
> --
> Thanks
> Emily
> =================
> Emily Jiang
> ejiang at apache.org <mailto:ejiang at apache.org>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>>>
>
>
> --
> Martin Kouba
> Software Engineer
> Red Hat, Czech Republic
>
>
>
>
> --
> Thanks
> Emily
> =================
> Emily Jiang
> ejiang at apache.org <mailto:ejiang at apache.org>
> <mailto:ejiang at apache.org <mailto:ejiang at apache.org>>
>
>
> --
> Martin Kouba
> Software Engineer
> Red Hat, Czech Republic
>
>
>
>
> --
> Thanks
> Emily
> =================
> Emily Jiang
> ejiang at apache.org <mailto:ejiang at apache.org>
--
Martin Kouba
Software Engineer
Red Hat, Czech Republic
More information about the weld-dev
mailing list