[wildfly-dev] Hacking on a fix for the JPA/CDI (WFLY-2387) injection problem but now hit a new maven cycle between wildfly-jpa + wildfly-weld when WeldPortableExtensions is used from wildfly-jpa...

[Scott Marlow]

If I have the WildFly JPA subsystem depend on org.jboss.as.weld 
(org.jboss.as.weld static module), that introduces a cycle that maven 
complains about [1], since wildfly-weld depends on wildfly-jpa already.

I pushed my test branch to github [2], which builds on the proposed 
upgrade to use Hibernate ORM 5.1 in WF 11.

Should we switch to use the "Capabilities" hammer to expose the 
org.jboss.as.weld.deployment.WeldPortableExtensions class that the JPA 
subsystem needs access to?  Or should we work around this minor 
difficulty by exposing WeldPortableExtensions via an additional module 
that the wildfly-jpa can access?

Scott

[1] The projects in the reactor contain a cyclic reference:

Edge between 
'Vertex{label='org.wildfly:wildfly-weld:11.0.0.Alpha1-SNAPSHOT  '}' and 
'Vertex{label='org.wildfly:wildfly-jpa:11.0.0.Alpha1-SNAPSHOT'}' 
introduces to cycle in the graph 
org.wildfly:wildfly-jpa:11.0.0.Alpha1-SNAPSHOT --> 
org.wildfly:wildfly-weld:11.0.0.Alpha1-SNAPSHOT --> 
org.wildfly:wildfly-jpa:11.0.0.Alpha1-SNAPSHOT

[2] 
https://github.com/scottmarlow/wildfly/tree/WFLY-2387_ORM51_CDI_INJECTION

[3] 
https://docs.jboss.org/author/display/WFLY10/Working+with+WildFly+Capabilities