[jboss-jira] [JBoss JIRA] (JBRULES-3385) Rule not fired depending on the class loading order

[Manuel Reinaldo Falagan (JIRA)]

    [ https://issues.jboss.org/browse/JBRULES-3385?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12666644#comment-12666644 ] 

Manuel Reinaldo Falagan edited comment on JBRULES-3385 at 2/15/12 9:15 AM:
---------------------------------------------------------------------------

The data is arranged differently in the case that it works and in the case that it doesn't (within the itereator) but both seam ok.

When it works, the FieldIndexHashTableFullIterator.table, has a reference to all the tuples, but when it doesn’t, it has a reference to all except one. This one is linked to another tuple in the table which if I understand the code properly, it is correct.


                
      was (Author: manuelreinaldo):
    Further investigation shows that the problem is in the data in the iterator.
The FieldIndexHashTable$FullIterator used for the loop has one tuple less in the case that it fails.
I have put a breakpoint in the method LeftTupleIndexHashTable.getOrCreate() and this method is called for all the tuples in both cases, the one that works and the one which doesn't, but somewhere in between this and the creation of the iterator one tuple is lost. In fact, in the debuggin, the data in the AbstractHashTable is correct in both cases until the iterator is created.
                  
> Rule not fired depending on the class loading order
> ---------------------------------------------------
>
>                 Key: JBRULES-3385
>                 URL: https://issues.jboss.org/browse/JBRULES-3385
>             Project: Drools
>          Issue Type: Bug
>      Security Level: Public(Everyone can see) 
>    Affects Versions: 5.2.0.Final, 5.3.0.Final
>         Environment: SLES11 SP1 with java 1.6.0_14-b08 (32bit and 64bit)  or XP with jdk1.6.0_22 (32 or 64 bit)
>            Reporter: Manuel Reinaldo Falagan
>            Assignee: Mark Proctor
>            Priority: Blocker
>
> A very simple example "FailTest" triggers a rule only 15 times when it should trigger it 16 times.
> The same example with almost any change to the class model or to the order in loading the classes or the input data or the version of java it works.
> The Class FailTest contains only a main method calling the main method of the Class HelloWorldExample. Running directly the main method in the class HelloWorldExample does work.
> Any change fixes the problem in the example but it moves the error to other part. We have set a test testing more than 2000 million cases of rules/inputs and no matter how we write the rules or the code, we end up finding a case that fails, making the software not usable, as it is not predictable were it is going to fail.
> The case has been uner study for more than a year and we have tried AIX, Linux and XP with DROOLS 5.1, 5.2, 5.3 and the trial of Enterprise edition. The example provided has only been tested to fail for DROOLS 5.3.0 final and Enterprise edition for SLES11 SP1 with the SUN virtual machine (not the IBM one) and for XP but there is no combination for which DROOLS work for the 2000 million cases.
> The example provided is a 7kb example self-contained ( ftp://mpf:mpf@ftp.eumetsat.int/DROOLS_ERROR.zip ) 

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira