Enables parallel test runs

[kettanaito]

Changes

• Removes --runInBand flag from unit and integration test commands.
• Removes the --max_old_space_size=8000 cap on the integration tests (previously added due to spawning WDS on each test and ts compiler causing a memory heap).

Motivation

Current CI takes a lot of time to run, decreasing the iteration speed for the changes. Integration tests alone take about 6m 30s to run, not taking the workspace attachment into account (ref).

History

The --runInBand flag was added to prevent some leaking state between tests. It might have been related to the old testing setup that we don't have anymore, or to #474. I didn't see any problems running parallel tests locally.

Benchmarks

• Before
• After (see CI for the latest commit)

[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 46a529e:

Sandbox	Source
MSW React	Configuration

[kettanaito]

CI fails spectacularly, reaching out of memory state, and taking a tremendous amount of time to run.

[fnmunhoz]

Hey @kettanaito I have experienced this problem before and I'm almost sure it's related to that:

https://discuss.circleci.com/t/memory-problems-with-jest-and-workers/10297/4

The CircleCI job exposes all the VM CPUs to Jest, but it actually can only use 2 in this case, which is the number of available CPUs to the job (via docker).

I would suggest playing with this option, starting with 2: --maxWorkers=2.

Jest recommends one worker per CPU core, minus 1 for the main thread, but it might be possible to spawn more in CircleCI, until the tests run consistently.

Hope it helps :)

[msutkowski]

Yup, circleci isn't very friendly with Jest depending on the instance size. We might want to consider switching CI to github actions for fun :)