Breaking: Spawn webserver in separate process

[molant]

Pull request checklist

Make sure you:

• Signed the Contributor License Agreement
• Followed the commit message guidelines

For non-trivial changes, please make sure you also:

• Added/Updated related documentation.
• Added/Updated related tests.

Short description of the change(s)

This currently spawns a new server per test. I was thinking about doing something smart and have just one per test file but ava spawns every test in parallel so I'm not sure that would change anything.

Also this PR is missing:

• Update the testing documentation on how the test server works
• Update the create-hint package

But it will be good if you can start taking a look into it.

Fix #1694

[molant]

BTW, I passed all the tests on Windows for the first time in a very long time 😊

[molant]

@sarvaje I've done your feedback

[molant]

• Update the create-hint package ➡ Doesn't need any manual update. We mark this as a breaking change so the release script updates everything.

@alrra, did you change the script so major version bumps of devDependencies are patches/minors?

[sarvaje]

@molant why are you ignoring connector-chrome in so many tests?

[molant]

why are you ignoring connector-chrome in so many tests?

Windows build keep failing on Windows and I'm testing things to make it work. I was expecting them to be more reliable by spawning less chrome instances :(

[sarvaje]

Windows build keep failing on Windows and I'm testing things to make it work. I was expecting them to be more reliable by spawning less chrome instances :(

But it is just for testing, or is it going to be permanent?

[molant]

But it is just for testing, or is it going to be permanent?

🤷‍♂️, I'm doing tests for now

[alrra]

@alrra, did you change the script so major version bumps of devDependencies are patches/minors?

Done.

[molant]

I think this is ready for merge. I've done a few changes since the last review so it will be cool if you could review it again:

• Fix a race condition with CDP that was causing the Windows failures (that's what I believe, Windows builds are now passing on Pipelines 🎉)
• Refactor requestResponse in jsdom to match the one in chrome (forgot about it last time, both files should look about the same now)
• group some awaits into await Promise.all()

Also this PR should be rebased instead of squashed. I can do it later today if I get enough 👍

[molant]

@antross can you please take another look? 🙏 🥺
Can't wait to get this merged

[antross]

Looks good to me with the updates assuming tests pass.

[sarvaje]

Any reason to have these thinks hardcoded and not in a files?

[molant]

Perf, that way we avoid reading from disk and it's not like we are going to change them.

[sarvaje]

Ok!

[molant]

Looks good to me with the updates assuming tests pass.

Tests pass! mac is failing because of #1772 which happens randomly but that's unrelated to this PR.