ISPN-8647 Initialize all loggers statically

[tristantarrant]

This prevents a potential deadlock in the Logger factory which synchronizes on the ClassLoader.

https://issues.jboss.org/browse/ISPN-8647

[rvansa]

@tristantarrant While I agree with making the loggers static (despite losing info about inheriting class), I don't think that the output is equivalent to the old behaviour. Simple test tells me:

import java.lang.invoke.MethodHandles;

public class LookupTest {
   static class A {
      static final Class<?> clazz = MethodHandles.lookup().lookupClass();

      public void foo() {
         System.out.println(clazz.getName());
      }
   }

   static class B extends A {
   }

   public static void main(String[] args) {
      new B().foo();
   }
}

and the output is LookupTest$A. If this is intended you don't have to use MethodHandles, you can simply state the class directly.

[wburns]

A few tweaks and it should be a bit cleaner.

[wburns]

Unfortunately before this was used to show trace messages from the appropriate implementation. I have had to fix this before and the way to do this is to instead add an abstract getLog method to the abstract class that is implemented by the extended classes to properly say where the messages are from. This way the extending class can still define the logger as static but the abstract class uses the appropriate logger.

[wburns]

Check out https://github.com/infinispan/infinispan/blob/master/core/src/main/java/org/infinispan/stream/impl/AbstractCacheStream.java#L153

[wburns]

This can use the getLog approach as well.

[wburns]

Actually this line can be completely removed (I missed removing it on a refactor). This is already using the getLog approach I mentioned on the off heap one.

[wburns]

This line can be removed - unused.

[tristantarrant]

I have used the getLog() approach where inheritance matters.

[wburns]

Integrated into master, thanks @tristantarrant !