When attempting to move to Log4J 2.0, I've noticed a number of hard deps on log4j classes.

SampleConfigFilesCorrectnessTest - this class makes use of a custom appender to analyse what a user is being warned of when a config file is parsed. Why are we using Log4J for this? Our own logging interface should be mocked and messages captured directly.

RehashStressTest and NucleotideCache - seems like a bug, I presume the author intended to use org.infinispan.logging.Log.

CompressedFileAppender and ThreadNameFilter- can this be written in a way that works with Log4J 1.x as well as 2.x? Or have the SPIs changed that much?