JDK16 Support

[jansupol]

Signed-off-by: jansupol jan.supol@oracle.com

[jansupol]

Required CQs for:

• org.ops4j.pax.url:pax-url-mvn:1.3.7
• org.apache.felix:org.apache.felix.framework:7.0.0
• org.hibernate.validator:hibernate-validator:6.2.0.Final
• org.hibernate.validator:hibernate-validator-cdi:6.2.0.Final
• org.mockito:mockito-core:3.9.0
• org.jboss.weld.se:weld-se-core:3.1.7.SP1
• org.jboss.weld.servlet:weld-servlet:3.1.7.SP1
• org.jboss.arquillian.container:arquillian-weld-embedded:2.1.0.Final

[jbescos]

I wrote a small comment to double check.

[jbescos]

Should we synchronize this or are we sure only one thread can invoke it the first time?. Something like:

    /* package */ Cache<Class<?>, Boolean> getJaxRsResourceCache() {
        /* CDI injection hack */
        if (jaxRsResourceCache == null) {
           synchronize(this) {
                if (jaxRsResourceCache == null) {
                      jaxRsResourceCache = new Cache<>(clazz -> Resource.from(clazz) != null);
                }
           }
        }
        return jaxRsResourceCache;
    }

[jansupol]

Nice catch. But invoking this multiple times would not break anything, there would be two instances of the cache & Resource.from would be invoked twice instead of having it used from the cache. I am not sure the double-checked locking is worth it in this case.

[jbescos]

I don't know all the details about how is going to be used, but imagine that there is a singleton that invokes that method and it saves it as a field. So next time it wants to use the cache, it reads it from the field (instead of getting it from getJaxRsResourceCache()).

It will not harm to have that lock because it will only lock at the beginning and we get rid of unnecessary extra memory usage.

[jansupol]

It happens that all the if (jaxRsResourceCache == null) { block is unnecessary, so we could have dropped it