How to fix static variable does not reset on test method when using either Test.startTest() nor System.runAs(User){}?

Question

Today I found out something weird about the way static variables act when used while running a test method and I wanted to know if anyone knows why this is the case.
I have a piece of code that creates a log record once an Incident record is created, and once when its status field value is changed to 'Closed'. In order to avoid redundancy I have a static boolean as a flag for the logic to run only once per context. Now, since incident creation and status change to 'Close' always happen on separate context in real life, this wasn't a problem. However, once I tried to write a test method to test my logic I realized that my static boolean flag doesn't reset even though using both test.startTest() and System.runAs(User) to start a new context. Eventually I went with this solution of resetting the flag's value manually. But I'm afraid I'm losing the ability to actually keep track of the flag behavior by doing so.

Here is my test method:

@IsTest
static void setClosingIncidentsStateTrackers(){
    //Select Required Records
    User                                incidentClient                  =   [SELECT Id FROM User WHERE Username = '[email protected]' LIMIT 1];
    BMCServiceDesk__Category__c         category                        =   [SELECT Id FROM BMCServiceDesk__Category__c WHERE Name = 'TestCategory'];
    BMCServiceDesk__BMC_BaseElement__c  service                         =   [SELECT Id FROM BMCServiceDesk__BMC_BaseElement__c WHERE Name = 'TestService'];
    BMCServiceDesk__BMC_BaseElement__c  serviceOffering                 =   [SELECT Id FROM BMCServiceDesk__BMC_BaseElement__c WHERE Name = 'TestServiceOffering'];
    BMCServiceDesk__Urgency__c          highUrgency                     =   [SELECT Id FROM BMCServiceDesk__Urgency__c WHERE Name = 'גבוה'];
    BMCServiceDesk__Impact__c           siteImpact                      =   [SELECT Id FROM BMCServiceDesk__Impact__c WHERE Name = 'אתר'];
    Problem_Source__c                   problemSource                   =   [SELECT Id FROM Problem_Source__c WHERE Name = 'בעיית הגדרות'];

    List<BMCServiceDesk__Incident__c> incidents = new List<BMCServiceDesk__Incident__c>();
    for(integer i = 0; i < batchSize; i++){
        BMCServiceDesk__Incident__c incident = new BMCServiceDesk__Incident__c(
            BMCServiceDesk__FKClient__c             =   incidentClient.ID,
            BMCServiceDesk__FKCategory__c           =   category.ID,
            BMCServiceDesk__FKServiceOffering__c    =   serviceOffering.ID,
            BMCServiceDesk__FKBusinessService__c    =   service.ID,
            BMCServiceDesk__FKStatus__c             =   awaiting_for_handling
        );
        incidents.add(incident);
    }
    insert incidents;

    for(BMCServiceDesk__Incident__c incident : incidents){
        incident.BMCServiceDesk__FKStatus__c = in_work;
    }       
    update incidents;

    test.startTest();
    System.runAs(new User(Id = UserInfo.getUserId())){
        for(BMCServiceDesk__Incident__c incident : incidents){
            incident.BMCServiceDesk__FKStatus__c    = handled;
            incident.BMCServiceDesk__FKBMC_BaseElement__c = service.ID;
            incident.BMCServiceDesk__ClosureCategory__c = 'נפתר';
            incident.BMCServiceDesk__incidentResolution__c = 'נפתר';
            incident.BMCServiceDesk__state__c       = false;
        }
        System.debug(LoggingLevel.ERROR, '*** /// Closing Incidents /// ***');
        //MNR_Incident_TriggerHandler.incidentTrackersWereAlreadyCreatedInThisContext = false; <-- This is the manual reset of the static boolean flag
        update incidents;
    }
    test.stopTest();

    List<Incident_State_Tracker__c> ists = [SELECT Id, Incident__c, State__c FROM Incident_State_Tracker__c];

    System.debug(LoggingLevel.ERROR, '*** // ' + ists.size() + ' // ***');
}

I'd like to know if you guys encounter this issue and know how to force clean new context inside a test method for resetting static variables.

Thank you very much in advance,
Lior.

Answer 1

I think there are two linked misunderstandings here.

I realized that my static boolean flag doesn't reset even though using both test.startTest() and System.runAs(User) to start a new context.

Test.startTest() and System.runAs() both manipulate aspects of the transaction context: Test.startTest() resets your governor limits, and System.runAs() alters the running user. Crucially, however, neither starts a new transaction. And static variables are scoped to the transaction.

When running unit tests, each test method executes in a single transaction. You can't start a new transaction inside a unit test. That's part of why you'll sometimes see unusual behavior in Asynchronous Apex when it's forced to execute inside a unit test method, synchronously, with Test.stopTest(): it's executing in a "dirty" transaction context, instead of its own isolated transaction as it expects. (This is the culprit behind many CalloutExceptions when running tests).

The other is

In order to avoid redundancy I have a static boolean as a flag for the logic to run only once per context.

This is almost never what you really want. You certainly want to ensure your code functions correctly when recursion occurs, but you don't want your whole trigger, or even one trigger event, to run exactly once per transaction. Among other subtle and difficult-to-debug maladies, this causes your trigger to silently fail if more than 200 records are touched in a transaction, or if any number of records (even 1) are both inserted and subsequently updated in a single transaction.

An answer by sfdcfox nicely covers a better recursion-guard pattern. In other circumstances, you may need to alter trigger logic to take no action when not required by state (best), or to track records already affected via a Set<Id>, rather than a static Boolean.