Update JSDOM and create script to automate it

[CompuIves]

Fixes #4530

[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 2a07c1c:

Sandbox	Source
Notifications Test	Configuration
Code Sandbox bug with old JSDOM version	Issue #4530

[lbogdan]

Build for latest commit 2a07c1c is at https://pr4539.build.csb.dev/s/new.

[nickserv]

Thanks for working on this! My only code critique is that the .js file is still being committed even though it's not used, and this isn't deleting the old ones easier. It would probably be best to rebase this PR to only add the .min.js file (so the 5MB git object for .js isn't written to the repository) and remove the old JSDOM 4 files that aren't in use any more.

[nickserv]

Unfortunately my test from #4530 is still failing with the same issue, seems like old JSDOM is still used in this sandbox: https://pr4539.build.csb.dev/s/code-sandbox-bug-with-old-jsdom-version-9gcnp?file=/index.test.js

[CompuIves]

It would probably be best to rebase this PR to only add the .min.js file (so the 5MB git object for .js isn't written to the repository) and remove the old JSDOM 4 files that aren't in use any more.

I've decided to not remove JSDOM 4 yet, since some service workers need to request the file still. I'm planning to remove it in a month though! For the .min.js vs .js, that makes sense since we can generate one ourselves now. I'll remove the .js one!

Unfortunately my test from #4530 is still failing with the same issue, seems like old JSDOM is still used in this sandbox

Hmm, that is weird. I see the template is static, so the bundler shouldn't be running any tests there.

[nickserv]

That's still strange, if it's running tests why is it still using JSDOM (and still 4 not 12)? I could tell if it was using the real DOM because the tests would be passing. Do we need to test this without the static template and disable the test runs separately?

[marcosvega91]

Just for info I have created a repo with the code that is broken on our side

https://pr4539.build.csb.dev/s/cold-currying-p7r3g?file=/src/__tests__/index.js

[nickserv]

FYI: My sandbox was based on this so should reproduce the same issue. The bug in user-event is caused by a combination of a regression (which we decided not to fix) and a bug in JSDOM (which is in Code Sandbox's version, JSDOM 4). My sandbox shows how JSDOM incorrectly marks selectionStart and selectionEnd, which causes that regression in user-event.

[CompuIves]

Picking this one up again! Apologies for the time, I lost track of this one.

I now updated JSDOM to 16.3.0, and only committed the min version, but still, getting the same error. It's really weird.

[CompuIves]

Actually... The log does work...

[CompuIves]

Weird, this returns false:

Object.prototype.hasOwnProperty.call(element, 'selectionStart')

[CompuIves]

Ah, it turns out this was a bug in jest-matchers that have since been fixed. I've just updated and the report now seems to work correctly on my side!

[marcosvega91]

this test is working now, so on our side we are good too 😄