fix(waitFor): handle odd timing issue with fake timers

[kentcdodds]

What: fix(waitFor): handle odd timing issue with fake timers

Why: It's super hard to explain and I honestly don't really understand the issue completely myself. All I know is, before moving the checkCallback function call above the promise flush line, I had issues with tests impacting each other. After this change, that issue goes away.

I also added a comment explaining why we shouldn't use jest.advanceTimersToNextTimer along with a test to demonstrate the issue.

How: Change the order of operations

Checklist:

- [ ] Documentation added to the docs site N/A

• Tests
  - [ ] Typescript definitions updated N/A
• Ready to be merged

[codesandbox-ci]

This pull request is automatically built and testable in CodeSandbox.

To see build info of the built libraries, click here or the icon next to each commit SHA.

Latest deployment of this branch, based on commit 739afef:

Sandbox	Source
eloquent-thompson-fg3fr	Configuration

[codecov]

Codecov Report

Merging #667 into master will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master      #667   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           24        24           
  Lines          593       593           
  Branches       148       148           
=========================================
  Hits           593       593           

Impacted Files	Coverage Δ
src/wait-for.js	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2fe177f...739afef. Read the comment docs.

[kentcdodds]

🎉 This PR is included in version 7.17.1 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[eps1lon]

All I know is, before moving the checkCallback function call above the promise flush line, I had issues with tests impacting each other. After this change, that issue goes away.

@kentcdodds Do you remember which tests were impacting each other? It's totally fine if this only reproduces in a full test suite. But some repro would be nice since this change is quite odd considering we're scheduling more work from waitFor after the waitFor promise is resolved.

[kentcdodds]

If this change is causing issues, then let's change it back. I'm pretty sure this was in the bookshelf app, but I don't have time to reproduce things and the latest version of react-query eliminates all the problems that I experienced with those tests anyway so we're probably ok to revert this.

[eps1lon]

#1073 has a test that breaks with the behavior introduced in this PR (#667).

I can fix the problematic behavior with a less invasive fix. But I can also just revert this change. I'm more comfortable with just reverting this change.

In any case, I'll try both fixes in the bookshelf app so reduce likelyhood of unintended breakage. Thanks for the pointer!

[eps1lon]

See #1073 (comment) for conclusion of the thread started with #667 (comment)