fix(CI): prevent the e2e test from running on 0.71

[kelset]

(new variation of #35667, read the last comment there for more info)

Summary

Because of the usual ✨dark magic✨ that we do 'cause RN is not a proper monorepo (usual link react-native-community/discussions-and-proposals#480), one of the side effects is that test_js and test_js_prev_lts insta-crash whenever they run in a -stable branch because of the run_e2e job, that runs verdaccio and verdaccio needs the workspace, but the workspace is not there because of the aforementioned ✨dark magic✨

This PR fixes that by not allowing said command to be run when we are in 0.XX-stable branches. (TIL that under the hood CircleCI wants Java regex)

In main we don't need it because of the work of @hoxyq (see 7f29357)

Changelog

[INTERNAL] [FIXED] - prevent the e2e test from running on 0.71

Test Plan

I used this PR for testing: #35665

here's a screenshot of when things work (no E2E command run):

here's how it looks when they don't (E2E runs and crashes):

You can also verify that the condition is correct by throwing this sample code in an online Java ide:

import java.util.regex.*;  

public class Main
{
    public static void main(String[] args) {
        System.out.println(Pattern.matches("0.[0-9]{2}-stable", "main"));//false  
        System.out.println(Pattern.matches("0.[0-9]{2}-stable", "0.71-stable"));//true
        System.out.println(Pattern.matches("0.[0-9]{2}-stable", "just-a-test-with-stable"));//false
        System.out.println(Pattern.matches("0.[0-9]{2}-stable", "just-a-test-with-stable-test"));//false  
    }
}

It's expected to fail here because the name of the branch is NOT 0.XX-stable ;)

[ryancat]

@kelset I saw some CI tests are failing with the same error message:

{"type":"error","data":"Cannot find the root of your workspace - are you sure you're currently in a workspace?"}

Doesn't this change suppose to fix that test?

[NickGerleman]

Suggested change

	# When not in 0.XX-stable branch, run e2e tests
	- when:
	condition:
	not:
	matches:
	pattern: "0.[0-9]{2}-stable"
	value: << pipeline.git.branch >>
	steps:
	- run_e2e:
	platform: js

From @ryancat's comments it looks like the test we were wanting to disable might still be running. This change originally came from the main branch to conditionally run on stable branches.

I wonder if << pipeline.git.branch >> might evaluate to the branch a PR comes from or some jazz like that. But if this is only to the single branch we could also just remove everything.

FYI @ryancat, since this change targets a "stable branch" instead of "main", we do not sync it to Phabricator, and this change can be committed directly in GitHub.

[kelset]

@NickGerleman the test is still running on this PR because the branch is kelset/filter-e2e-on-071 ;)

[cipolleschi]

The failure you see are correct. The pipeline.git.branch you see are happening because the PR is on a branch called kelset/filter... which does not satisfy the RegEx, of course. The tests are run, but the base branch is a stable, so they fail.

I think that the condition is missing a case: we don't want to run those tests if:

• the CI is running on 0.xx-stable
  or
• the PR is opened against an 0.xx-stable branch.

WDYT?

[cipolleschi]

We can merge this as it is, to cleanup the release workflow when we push a tag, do the cherry-picks and for the actual release jobs. All of these run directly on 0.XX-stable.

A more complete solution, will skip the e2e tests also for PRs opened against the stable branch. But next year we will revamp the E2E testing, so it makes sense to wait for that, probably.

[NickGerleman]

We can merge this as it is, to cleanup the release workflow when we push a tag, do the cherry-picks and for the actual release jobs. All of these run directly on 0.XX-stable.

A more complete solution, will skip the e2e tests also for PRs opened against the stable branch. But next year we will revamp the E2E testing, so it makes sense to wait for that, probably.

There is some helpful context in the comment @kelset links to. These tests are for project initialization (not the UI "e2e" testing they might imply), but also the underlying issue preventing them from running in stable branches has already been fixed in main.

It might not hurt to rename these scripts/tests to something different. The only reason I know what these scripts do is stumbling into them looking for project initialization tests for the TS change 😅.

[kelset]

It might not hurt to rename these scripts/tests to something different. The only reason I know what these scripts do is stumbling into them looking for project initialization tests for the TS change 😅.

totally, but again - this PR is just a fix for 0.71-stable branch, in main you can go ahead and do more substantial changes.