Vladimir, I repeat again: there are cases of DRLVM failure before xml file creation. They can occure along with successful test passing of another tests. I'm not sure for CLASSLIB testing, but I can imagine the situation when JVM is failed if the name of the test to be executed contains "CrashIt" substring. So we will have unnoticed CLASSLIB's test failure. So, to my mind, empty xml file analysis can be used for CLASSLIB testing (although and without 100% confidence in results), but it can't be used for DRLVM testing (again - because of the possible absence of failure indicator - empty xml file).
Alexander
