Here is a pattern. I am working on a pretty large project to basically change my tests for a BIG UI change within our AUT. I work on the test project happily day after day. I get a couple of great runs in the green. I go to run it again over night while I am merrily sleeping. I come in to find the CLICK behavior is not working. Click not working is a bit mind bending, right? Of course click works. I try to run again while looking at an example test, and it works fine. Take a look at the (sorry large) MHT for System Smoke Tests / AddCommentCodeSmoke. In the log folder Step: Add and Check, a click is occuring. I can see evidence in the log that the click is occuring. The error occurs then on the next action because the expected screen is not there. My autowait timeout is set to 20000. So there should be plenty of time to load the page for the next event as it searches for that object.
This is happening in many tests. I am trying to imagine if increasing my piddly (100) delay between events would help?
Anyone seen t
What's different about the system where it's running overnight? Is it a different pc? Is there some other job dragging things down?
Same machine. Nothing else running. The machines that run this are dedicated to this purpose. It is not a speed issue. The screen that should load from click neverloads.
The nighttime thing was a red herring. I observed it not working when running during the day. The issue is not reproducable when I run a single test. So I cannot tell what the spy sees. All I can tell is what the log shows me. I have entered a case with support.
That's strange behavior.
To stop the test at night use this method:
Check object with spy after these actions.
I just talked to the developer. Holy Oy Vay. So. Here is what we THINK is happening. That button I am trying to click? It moves. When the page is written server side, it is in one place, then is moved to another lower in the DOM. Ok that stinks. But I am aiming at the correct button. BUT I am speculating that the event handler for that button is not ready yet. This is the same behavior for many, many, many (like many more manies) buttons within the app. I increased the delay between events to absurdly high to wait for the code to futz around client side to see if this does anything. I will give more info when I can.
I have the strong sense that this is the problem since all debug actions are not presenting the symptom. It is likely working locally because the local machines we use are rocket ship fast compared to the production where the tests are getting executed.