Unwanted simultaneous execution using Nightwatch2.0 and Cucumber

[pholland-hbc]

I have a fairly large library of cucumber functions that were running fine in NW 1.7 using nightwatch-api.
I am trying to migrate over to NW 2.0 but it appears that the cucumber commands are not fully completing before the subsequent command is launching.
Unfortunately our library (which we inherited from a third-party, no longer involved) did not appear to be aware of async/await and so most of the functionality was very reliant on the fact that NW 1.7 would fully complete one cucumber command before advancing to the next. To be able to get our library to function would require a fairly major re-write.

I think it would be very helpful to have cucumber complete one command prior to carrying on regardless of what is being executed. It is not very helpful to say:
Given I am on the website
When I login using valid credentials
Then the account page is displayed

if the check for account page is being executed before the previous steps have returned.
Perhaps we could make the cucumber test runner ensure the promise queue is empty prior to continuing?
Any solution would be greatly appreciated.

[beatfactor]

Can you provide an example? Are you trying to run cucumber directly from Nightwatch?

[pholland-hbc]

Yes. I am running cucumber directly from Nightwatch. The issue appears to be that our (poorly "async"ed) inherited code does two things that behave differently in MW2.0. Firstly, each step of the cucumber script will start to execute before the previous step completes. I have tried reworking the first few steps of a script and I can get them to work but the effort to fix all cucumber statements would be huge. If that was the only problem then I would start on that exercise.
Unfortunately, we have inherited another issue which was a modification to Nightwatch 1.7 that allowed us to use a function called "run" to execute cucumber statements from within another cucumber step:
run(Given I open "${site}");

This allowed the writing of more complex steps without needing to have all the mini steps in the feature file or repeated multiple times in different steps.

The problem becomes that whenever a "run" is called, it derails the "one at a time" execution of the cucumber steps (even if reworked to properly use async/await). I also tried adding async/await to the run function (the original did not have any async).

To rework the entire library to add async/await would be manageable but to replace all the "run"s with expanded code would be a huge undertaking.

Here is the code for the run command (it essentially intercepts the Given/When/Then commands and allows "run" as well). All our step definition files require the attached helper.js file instead of @cucumber. This does not appear to impact the scripts from running asynchronously as long as "run" is not called.

helper.js.txt
Nightwatch_2.0_code_and_output.txt

[pholland-hbc]

The "run" command had no issues in NW 1.7 and executed each step called (either by run or directly at the .feature file level) one at a time. When we run our full scripts they die almost immediately because the cucumber steps all start running without waiting for the previous step to complete.

[pholland-hbc]

I am currently running Nightwatch 2.0.7. I execute the script using: npx nightwatch --env cucumber-js

[pholland-hbc]

My feature file is:

Feature: Wishlist

# Passed-Completed
@WorkflowRegression
Scenario: ODE_135 | ODE_139 | ODE_140
    """
    Verify going to Wishlist page
    """
Given I am on the website
Then Signup for exclusive offer
And I click on "header wishlist link"

[beatfactor]

You can also try and use the Nightwatch programmatic api directly in cucumber, which I believe is how it was working in v1.7 and using the nightwatch-api package. There's an example project which does that here. I hope it helps.