Bug 51868

Created attachment 27559 [details]
Example of jmeter.log

jmeter.log attached

2011/09/22 12:36:56 ERROR - jmeter.control.IfController: missing ; before statement (<cmd>#1) org.mozilla.javascript.EvaluatorException: missing ; before statement (<cmd>#1)

is caused by a syntax error in the JMX file

Try disabling them in turn until the problem disappears, and then double-check the variables don't contain spaces etc. - use the Debug sampler.

Please look at attachment. Error occured ONLY if option "Start next loop on error" set. Before error in sample all actions were successfull and all samplers after failed sampler must not be executed. But if ni any errors in loop occured, all work properly

Have you tried using the Debug Sampler to see what the variables are?

Have you tried finding out which If Controller is involved?

Have you tried enabling debug logging for the If Controller?

This will help narrow down the issue.

(In reply to comment #4)
> Have you tried using the Debug Sampler to see what the variables are?
  Yes. If no any errors in sample, variables calculated in BeanShell and used in IFC are valid  (see screenshot)

> Have you tried finding out which If Controller is involved?
  There is the only IFC in scenario. It work properly, if errors absent. And it must not be validated, if error occured at any sampler before IFC

> Have you tried enabling debug logging for the If Controller?
  In jmeter.properties I set everything to DEBUG. Now I see 2 errors here - one about recirculation (51865) and another - exception. IFC try to validate undefined value (getCondition() : [${spent_time}>5000]). Yes, it is not valid. But IFC must not be processed at all, because value calculating for IFC and IFC are at the end of loop. After error first sample of loop must be processed, is it?
Log attached
 
> This will help narrow down the issue.

Created attachment 27560 [details]
Variables

Created attachment 27561 [details]
Debug log

Hello,
Your second BeanShell Sampler has an error, it should be:
vars.put("spent_time",execution_time.toString());

look in jmeter.log you have:
2011/10/01 17:05:14 ERROR - jmeter.util.BeanShellInterpreter: Error invoking bsh method: eval	Sourced file: inline evaluation of: ``current_time = System.currentTimeMillis(); start_time = vars.get("start_time");  . . . '' : Error in method invocation: Method put( java.lang.String, long ) not found in class'org.apache.jmeter.threads.JMeterVariables' 
2011/10/01 17:05:16 WARN  - jmeter.protocol.java.sampler.BeanShellSampler: org.apache.jorphan.util.JMeterException: Error invoking bsh method: eval	Sourced file: inline evaluation of: ``current_time = System.currentTimeMillis(); start_time = vars.get("start_time");  . . . '' : Error in method invocation: Method put( java.lang.String, long ) not found in class'org.apache.jmeter.threads.JMeterVariables' 


Regarding the issue, this is my analysis, 
when error occurs in either HTTP Sampler or bean, as Start NExt loop is configured we enter this part of code of JMeterThread:

sam = controller.next(); // need perfom a until loop for special case (tc as parent)

while (sam != null && !sam.equals(firstSampler)) { // while the thread is NOT on the begining of the tree
    sam = controller.next();
}

next controller is IfController , I think we enter a known issue:
"The If Controller may cause an infinite loop if the condition is always false from the first iteration"


Anyway, Maybe IfController#evaluateCondition should throw an exception when condition has a evaluation error instead of returning false, it would stop test and let you know your condition is wrong.
Regards
Philippe

Created attachment 27658 [details]
Fix to bad handling of error in Condition

Sorry I think I read too fast
Part of my analysis is wrong.
I recap:
When an error occurs (enabled (fail) Load page (loop)) JMeterThread will enter this part of code:
sam = controller.next(); // need perfom a until loop for special case (tc as
parent)

while (sam != null && !sam.equals(firstSampler)) { // while the thread is NOT
on the begining of the tree
    sam = controller.next();
}

It means if will for IfController execute the condition which is wrong due to the fact that BeanShell sampler just before did not run, so spend_time won't be in variables.

That explains the logs.
Regards
Philippe

A fix to issue would be to flag we are in a restartNextLoop
try{
                                JMeterContextService.getContext().setIsWithinRestartNextLoop();
                                // Last not ok. start get the begining of the tree
                                sam = controller.next(); // need perfom a until loop for special case (tc as parent)
                                while (sam != null && !sam.equals(firstSampler)) { // while the thread is NOT on the begining of the tree
                                    sam = controller.next();
                                }
                                // At this point: begining tree, thus Last must Ok
                                threadContext.getVariables().put(LAST_SAMPLE_OK, TRUE);
                            } finally {
                                JMeterContextService.getContext().unsetIsWithinRestartNextLoop();                                
                            }

In IfController#next:
Test for flag and don't eval condition but what to return ?


What's your opinion ssebb ?

Regards
Philippe

Created attachment 27669 [details]
FIx to issue

Hello,
I implemented described solution that sets a flag when within restart next loop code in JMeterThread.
IfController tests this state and does not evaluate condition and return next sampler as if condition was true.

Regards
Philippe

Created attachment 27696 [details]
Fix to issue taking into account your note in issue 51866

Hello,
I think this fixes both 51868 and 51866.

ant test passed OK 
2 Test plans of 51866 and Test plan of 51868 also passed OK.
Regards
Philippe M.

Date: Mon Oct 31 10:50:11 2011
New Revision: 1195404

URL: http://svn.apache.org/viewvc?rev=1195404&view=rev
Log:
Fix to Start Next Loop broken feature, fixes following issues:
- Bug 51865 - Infinite loop inside thread group does not work properly if "Start next loop after a Sample error" option set
- Bug 51868 - A lot of exceptions in jmeter.log while using option "Start next loop" for thread
- Bug 51866 - Counter under loop doesn't work properly if "Start next loop on error" option set for thread group


Added:
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/threads/FindTestElementsUpToRootTraverser.java   (with props)
Modified:
   jakarta/jmeter/trunk/src/components/org/apache/jmeter/control/ForeachController.java
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/control/Controller.java
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/control/GenericController.java
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/control/IfController.java
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/control/LoopController.java
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/control/RunTime.java
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/control/WhileController.java
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/threads/AbstractThreadGroup.java
   jakarta/jmeter/trunk/src/core/org/apache/jmeter/threads/JMeterThread.java
   jakarta/jmeter/trunk/xdocs/changes.xml

Hello,
A Fix has been provided, it would be Nice of you to test it and submit a comment confirming it is fixed for you.

Regards

verified on r1213077

I change state cause with VERIFIED state , issue still appears as opened in weekly bug report.
Thanks for verification.

This issue has been migrated to GitHub: https://github.com/apache/jmeter/issues/2557