We suspect that exclusive test coverage of com.ibm.wala.cast.python is coming from com.ibm.wala.cast.python.ml.tests. It would be great to have this code also covered by com.ibm.wala.cast.python.tests so that we are not dependent on com.ibm.wala.cast.python.ml.tests. In other words, we should see similar coverage metrics on com.ibm.wala.cast.python if we remove com.ibm.wala.cast.python.ml.tests.
Please compare the two runs to see exactly which code coverage we are losing in com.ibm.wala.cast.python between the two runs. If we can understand the code coverage we are losing, perhaps we can add some tests to get them even.
Originally posted by @khatchad in wala/WALA#1412 (reply in thread)