[jboss-jira] [JBoss JIRA] (DROOLS-1326) Stateful working memory regression OCT-2016: nurserostering long_hint01.xml

[Geoffrey De Smet (JIRA)]

    [ https://issues.jboss.org/browse/DROOLS-1326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13306958#comment-13306958 ] 

Geoffrey De Smet commented on DROOLS-1326:
------------------------------------------

[~jlocker] go in SolveAllTurtleTest (the superclass) and play with the time spends... and this is where it sucks:
those turtle tests run 3 solvers (2 minutes, 5 minutes and 3 minutes for a total of 10 minutes):

{code}
        Solution_ planningProblem = readPlanningProblem();
        // Specifically use NON_INTRUSIVE_FULL_ASSERT instead of FULL_ASSERT to flush out bugs hidden by intrusiveness
        // 1) NON_INTRUSIVE_FULL_ASSERT ASSERT to find CH bugs (but covers little ground)
        planningProblem = buildAndSolve(solverFactory, EnvironmentMode.NON_INTRUSIVE_FULL_ASSERT, planningProblem, 2L);
        // 2) FAST_ASSERT to run past CH into LS to find easy bugs (but covers much ground)
        planningProblem = buildAndSolve(solverFactory, EnvironmentMode.FAST_ASSERT, planningProblem, 5L);
        // 3) NON_INTRUSIVE_FULL_ASSERT ASSERT to find LS bugs (but covers little ground)
        planningProblem = buildAndSolve(solverFactory, EnvironmentMode.NON_INTRUSIVE_FULL_ASSERT, planningProblem, 3L);
{code}

You need to mimic this output:

{code}
016-10-11 12:44:26,305 [main] INFO  Opened: data/nurserostering/unsolved/long_hint01.xml
2016-10-11 12:44:26,592 [main] INFO  Solving started: time spent (63), best score (-740init/0hard/-1580soft), environment mode (NON_INTRUSIVE_FULL_ASSERT), random (JDK with seed 0).
2016-10-11 12:46:26,542 [main] INFO  Construction Heuristic phase (0) ended: time spent (120013), best score (-337init/0hard/-580soft), score calculation speed (167/sec), step total (403).
2016-10-11 12:46:26,543 [main] INFO  Solving ended: time spent (120014), best score (-337init/0hard/-580soft), score calculation speed (167/sec), phase total (2), environment mode (NON_INTRUSIVE_FULL_ASSERT).
2016-10-11 12:46:26,882 [main] INFO  Solving started: time spent (97), best score (-337init/0hard/-580soft), environment mode (FAST_ASSERT), random (JDK with seed 0).
2016-10-11 12:46:28,889 [main] INFO  Construction Heuristic phase (0) ended: time spent (2104), best score (0hard/-175soft), score calculation speed (17127/sec), step total (337).
2016-10-11 12:51:26,785 [main] INFO  Local Search phase (1) ended: time spent (300000), best score (0hard/-74soft), score calculation speed (6848/sec), step total (1233).
2016-10-11 12:51:26,785 [main] INFO  Solving ended: time spent (300000), best score (0hard/-74soft), score calculation speed (6914/sec), phase total (2), environment mode (FAST_ASSERT).
2016-10-11 12:51:27,164 [main] INFO  Solving started: time spent (157), best score (0hard/-74soft), environment mode (NON_INTRUSIVE_FULL_ASSERT), random (JDK with seed 0).
2016-10-11 12:51:27,165 [main] INFO  Construction Heuristic phase (0) ended: time spent (158), best score (0hard/-74soft), score calculation speed (1000/sec), step total (0).
{code}

As this happens in the 3th solver, this isn't easy to reproduce, but it is possible: use stepCountLimit to get the exact same return state of the first solver and then the second solver and then just run the 3th solver for a long time (in case your cpu is slower than jenkins's), it should hit it.

> Stateful working memory regression OCT-2016: nurserostering long_hint01.xml
> ---------------------------------------------------------------------------
>
>                 Key: DROOLS-1326
>                 URL: https://issues.jboss.org/browse/DROOLS-1326
>             Project: Drools
>          Issue Type: Bug
>          Components: core engine
>    Affects Versions: 7.0.0.Beta2
>            Reporter: Geoffrey De Smet
>            Assignee: Mario Fusco
>            Priority: Blocker
>
> Something regressed in Drools master between 27-SEP and 11-OCT.
> See:
> https://kie-jenkins.rhev-ci-vms.eng.rdu2.redhat.com/job/optaplanner-turtleTests/org.optaplanner$optaplanner-examples/73/testReport/junit/org.optaplanner.examples.nurserostering.app/NurseRosteringSolveAllTurtleTest/runFastAndFullAssert_2__data_nurserostering_unsolved_long_hint01_xml_/
> I 'll ask [~jlocker] to isolate.



--
This message was sent by Atlassian JIRA
(v6.4.11#64026)