[hibernate-dev] 5.1 tentative release date

Scott Marlow smarlow at redhat.com
Fri Jan 8 22:00:51 EST 2016


I'll create a jira for the Javassist to be part of 5.1.

Should we also look at changing Hibernate to not require Javassist 
classes be on the deployment classpath?  This might require cloning some 
Javassist runtime classes so that we don't get CNFE on 
javassist.util.proxy.ProxyObject (and whatever else is required by 
enhanced entity classes).

[1] contains one of the CNFE's that I see with WildFly during deployment 
time (only if WildFly is hacked to not inject Javassist into the 
application classpath).  I am seeing 
org.hibernate.proxy.pojo.javassist.JavassistProxyFactory enhance the 
entity class with references to Javassist classes (see disassembled 
bytecode output [2]).  The generated class extends 
javassist.util.proxy.ProxyObject + javassist.util.proxy.MethodHandler. + 
javassist.util.proxy.RuntimeSupport + javassist.util.proxy.SerializedProxy.

I'm sure that there are other Javassist classes that we probably also 
generate bytecode to depend on, in other places in Hibernate.

I have no idea exactly how to resolve this.  I'm not sure if it would 
entail cloning the above javassist runtime classes into javassist.  Or 
separating them into a different (Javassist) library, so at least the 
application doesn't include the other Javassist classes.

Sanne had some ideas that he mentioned [3].  By only exposing the needed 
classloaders to the deployment, I think he meant the above idea of 
separating Javassist into different jars.  Or something like that.

Jason Greene also liked Sanne's suggestion of not requiring applications 
to have Javassist on their classpath, as applications might also include 
their own copy of Javassist because they want to generate some bytecode 
also.

What do others think about the idea of not requiring Javassist to be on 
the Hibernate application classpath?  Again, I'm not sure if this only a 
problem on WildFly.  If it is, I'm not sure why. :)

Scott

[1] 
https://gist.github.com/scottmarlow/4e23e62962101b740a4a#file-gistfile1-txt-L1

[2] https://gist.github.com/scottmarlow/dc7ebfea654984f84e2e

[3] https://github.com/wildfly/wildfly/pull/8474#issuecomment-162698801

On 01/08/2016 02:49 PM, Steve Ebersole wrote:
> I don't see a Jira to upgrade Javassist as part of 5.1...
>
>
> On Fri, Jan 8, 2016 at 1:35 PM Scott Marlow <smarlow at redhat.com
> <mailto:smarlow at redhat.com>> wrote:
>
>     Should we upgrade to javassist latest in 5.1 still?
>
>     On 01/08/2016 10:08 AM, Steve Ebersole wrote:
>      > Just a heads up that I tentatively set Jan 27th as the release
>     date for
>      > 5.1.  Please let me know if that does not work for anyone.  Also
>     please
>      > keep that date in mind if there is anything you want to get into 5.1.
>      > _______________________________________________
>      > hibernate-dev mailing list
>      > hibernate-dev at lists.jboss.org <mailto:hibernate-dev at lists.jboss.org>
>      > https://lists.jboss.org/mailman/listinfo/hibernate-dev
>      >
>


More information about the hibernate-dev mailing list