[ https://issues.jboss.org/browse/DROOLS-2592?page=com.atlassian.jira.plugi... ] Lance Leverich commented on DROOLS-2592: ---------------------------------------- After coming up with a script (see test.sh above) that # would execute the tests over and over # and would save... * Generated PMML * DRL and Java classes generated directly from the PMML * Java classes generated from the generated DRL * Output of the maven command I have concluded that this must be a race condition of some sort. If you compare the directories in the two attached zip files, you will find that the only differences in the generated files (excepting the output.log which is the maven output) are related to time stamps. The output.log files differ in the amount of time spent executing various tests, and of course in the results of the ScorecardReasonTest. I believe that failures caused by this kind of "race condition" are most likely due to some rules having the same salience, and then the consequences of those rules being applied in a semi-arbitrary manner. This presents me with the opportunity to more thoroughly go through the generated rules to look for when this kind of issue would be able to occur. Please let me know if you have any thoughts on my reasoning here, or suggestions as to how I might approach this more effectively. -- This message was sent by Atlassian JIRA (v7.5.0#75005)