[infinispan-dev] AS7-4007 missing Infinispan dependency for clustered JPA second level cache

Scott Marlow smarlow at redhat.com
Mon Mar 5 17:07:51 EST 2012


Galder,

I'm trying to identify the right fix for addressing AS7-4007.  I started 
a clustered second level cache test that currently fails since 
Infinispan isn't on the test application classpath.

The exception call stack that I'm getting is shown here 
http://pastie.org/3503803.

I would like to discuss how we can address AS7-4007 (either on email or 
IRC).

Initially, I worked around this by ensuring that the infinispan module 
is available in the test application classpath 
(https://github.com/jbossas/jboss-as/commit/31f3547960775dac88275447253fdae942f952b9#L0R40). 
  The hack was to export the Infinispan module when the Hibernate module 
is brought into an application.  This covered both JPA and Hibernate 
native applications, however it exports into every other module that 
references the Hibernate module (increasing the time that it takes to 
link the Hibernate module into other modules).  There are other problems 
with requiring Infinipan to be in the application classpath (see 
https://community.jboss.org/wiki/ModularSerialization).

I would like to better understand the alternatives (both short term and 
long term) to address AS7-4007 and issues like it.  Perhaps by using 
MarshallingConfiguration.setClassResolver() to pass a 
ModuleClassResolver to be used with Infinispan 
marshalling/unmarshalling.  Is that currently configurable with Infinispan?

Paul Ferraro, David Lloyd and myself chatted on irc briefly about this a 
few days ago (see http://pastie.org/3528553).

https://github.com/dmlloyd/jboss-marshalling/blob/master/api/src/main/java/org/jboss/marshalling/ModularClassResolver.java 
contains the source for the ModularClassResolver class that David 
referred to (as well as the ModularSerialization article linked above).

Scott


More information about the infinispan-dev mailing list