TC 12.53 End Events change between browsers

Is this a feature or a bug? If this is a feature, then it would be most helpful if it was described in the Understand End Events section in online help.

1. I create asimple TC 12.53 script using the Firefox browser. Action makes a single call to the foo() func in the Local_Lib, passing the Boolean arg, navToHP, which is set to true. The in the foo() function a single Navigation step to is executed and the script ends.

2. I run the script in the TC editor (continuing to use Firefox) and confirm that the HP website is invoked, there are no errors and that the Navigate End Event is set to Automatic:Document Loaded, as in:

3. Now I exit the TC editor, which returns me to vuGen. I then select the IE browser (or Chromium) and click the Develop Script button to invoke the TC editor. I then inspect the above Navigate End Event but find that it has changed to Not Yet Set:

This is the case for both End Event settings in the simple script. So does this mean that you have to run a script in its entirety (including error branches if you have them) in all three browsers under the TC editor if you later plan to use all browsers during your load testing? Or is this a bug?

The attached script is provided in a state that shows this issue (at least on my Win 10 desktop).

 ? Please Advise, Terry Horwath

P.S. I would have attached the script in ZIP format, but I can't believe this site does not allow that file type to be attached! How are you supposed to show problems and share friggin' code?

  • I continue this evaulation and now run the script under the IE browser in the TC editor. But it selects different Navigate End Event than was automatically selected for Firefox:

    (I also confirm that after exiting the TC editor and re-entering that the End Events remain constant but different for both IE and Firefox).

    I then run 3 iterations of the script under vuGen for both IE and Firefox, getting substantially different page load times when using the different browsers:

    (yes, I edited output.txt to prune statements...)

    I do not see these significantly different load times by manually loading the HP page in IE and Firefox, thus I have to conclude the TC protocol/toolset is causing the difference. Am I doing something wrong? Is Firefox the preferred and more reliable browser from a performance testing and response time measurement perspective?

    ? Please Advise, Terry Horwath

    P.S. And when I edit the IE version End Event to Document Loaded it seems to load the HP page pretty quick but then the step hangs for few mins before timing out:



  • Hi Terry,

    First this is a feature. Starting 12.53 TruClient assigns different automatic end event to each browser. If the user changes the end event to a manually selected event it will override the automatically selected value for all browsers. This feature was developed since we observed different behavior in different browsers (e.g. one browser triggers a specific event while another one does not causing the step to hang until step timeout).

    Now, the difference in transaction response times between the browsers is caused by the different end events. I have mentioned that in the explanation about end events in the first webinar. The fact the page load time seems to be the same when loading it outside TruClient cannot serve as a reference since you are not measuring the time it takes to download all resources while TruClient IE is waiting for them since the end event was set to "Network completed"

    Hope it clarifies it a bit.

    We will see if there is a need to elaborate the explanation in the documentation.




  • re: "Now, the difference in transaction response times between the browsers is caused by the different end events."

    Well I see this as a major problem using this protocol. I say this because a customer who "sees" no difference performing a workflow from one browser to the next is going to consider "LoadRunner as a false prophet..." when it renders unrealistic transaction response times (as compared to what a user sees interactively running a single user workflow).

    So is Firefox the "preferred" browser for realistic response times?

    I have no idea how this significant difference in browsers response times could be explained away to a customer for a load test that is supposed to measure end-to-end browser response times (rather than just server response times).

    re: "We will see if there is a need to elaborate the explanation in the documentation"

    Provide a link to this posting.

  • Hi Terry,

    I already provided a link to this posting and our documentation team is handling that. They have decided to add an explanation in the TruClient Fundementals section.

    I agree the behavior is not clear and may be percieved as a major issue. 

    1. I do not consider TruClient Firefox as the preferred browser because of more accurate response time measurements. I like it since it is a more customizable browser which allows developers and testers a more flkexible working environment. It might change soon since Mozilla have made a very significant architectural change. It might affect my personal opinion regarding it :-)

    2. In casses like the one you have encountered in TruClient IE, in order to align the response time to the user percived response time, you can check if manually setting the end event Firefox assigned works also in TruClient IE. If it doesn't you can always change the end event to use Object Exist