How stacking works?

I don't understand how stacking works in rapid miner. 
How to train base classifier then test the base classifier. Then the output base classifier is fed to meta classifier. How to test the stacking model? How does it work? What are the steps?