after hook not firing reliably

[searls]

In order for our tests to behave consistently, the after hooks for each suite need to execute consistently (so we can dump our database). What I'm finding is that when I run my suite with npm test, the after hooks (like this one) are not firing.

However if I grab an individual test like this one and set it to describe.only, then the after hook will fire just fine.

Any insight into this?

[searls]

Ah, so it would seem that this is a mocha issue, reopened as recently as 20 hours ago. Wow.

mochajs/mocha#534

[dasilvacontin]

@searls, is it possible that the (async) before hook is failing? Because yesterday I found out that it bugs the run.

[searls]

There is no evidence that one is failing. It seems happy right up to the point it calls done()

[dasilvacontin]

When it calls done inside the after? What happens then?

[searls]

The after is never invoked if I run the whole suite. (I was referring to the before hook to answer your question)

[dasilvacontin]

Hmm, not sure if it's related to the bug I found.

[searls]

(subtext: because the suite is erroring. when i run just a passing or failing subset of the suite, the after is invoked)

[dasilvacontin]

The suite is erroring? Where?

[searls]

Okay I think we need to back up. I'm currently implementing this repo, which in turn is using this repo's test suite as a guide to figuring out how I need to implement it.

Because my adapter is half-implemented, most of the tests are going to be failing most of the time. I suspect that whenever a failure occurs due to an error being thrown, the suite's after hook does not fire. That is to say, after doesn't fire at all unless I intentionally run just one test or suite that I know to be passing (or failing logically and not throwing an error)

[dasilvacontin]

I see. What I've confirmed so far (and I'm currently fixing and adding new tests so that it doesn't happen again) is that an async failure in either a before hook or a test causes the after hook not to fire. We probably stopped setting up the uncaught exception handler at some point. Looking into it.

[searls]

That sounds about right. Apologies for the vague and probably overly-terse descriptions on my end (when we hopped repos I wasn't sure if you were representing Mocha or Sails)

[dasilvacontin]

No worries! Glad to help!

[CWyrtzen]

Howdy folks, is this resolved now?

[dasilvacontin]

I've been very busy lately, so I personally haven't been able to further investigate/make progress.

[dmarcelino]

@searls, can you please try using the latest version of waterline-adapter-tests in master branch and filter out the broken tests using the example in #35? Let us know if that fixes this issue. Thanks

[dmarcelino]

Closing this as abandoned, please reopen if you are still experiencing this.