[Bug/Question]: Does not work when jobs are in the same file?

[jimduchek]

What happened?

Stripped down to the following test:

name: Testing Wait Other Jobs

on:
    push:
        branches:
            - main
            - staging
            - development
        paths:
            - ".github/workflows/wait_other_job_test.yml"  
    workflow_dispatch:

permissions:
    contents: read # Since v2
    checks: read
    actions: read

jobs:
    job1:
        runs-on: ubuntu-latest

        steps:
            - name: Test 1
              shell: bash
              run: |
                sleep 60

    job2:
        runs-on: ubuntu-latest

        steps:
            - name: Wait
              uses: kachick/wait-other-jobs@v2

I expect job2 NOT to finish first. But it does, the very first poll passes, despite job1 not being done. If job1 and job2 are in separate yml files, it works as expected. Behavior is no different if specifying @v1, though THAT does have the interesting log output:

Get own job_id
  {
    "ownJobIDs": [
      23740674348,
      23740674474
    ]
  }

It seems perhaps it thinks other jobs in the same file are also itself (and therefore shouldn't wait)?

It's actually probably best for me to split my jobs into separate files anyway, but if this is the case (not sure if GH maybe changed something very recently), several of the examples ought not to work either.

Version

v2.0.4

Permissions

GITHUB_TOKEN Permissions
Actions: read
Checks: read
Contents: read
Metadata: read

Parameters

{
    "triggeredCommitSha": "3d0694542d6540fcb92347237727284e13fcafbb",
    "runId": 8657670048,
    "repositoryInfo": {
      "owner": "XXX",
      "repo": "XXX"
    },
    "waitSecondsBeforeFirstPolling": 10,
    "minIntervalSeconds": [15],
    "retryMethod": "equal_intervals",
    "attemptLimits": 1000,
    "isEarlyExit": true,
    "isDryRun": false,
    "waitList": [],
    "skipList": []
  }

Relevant log output

Not really useful -- it waits 10 seconds, polls, decides to continue:

Wait about 10 seconds before first polling.
Polling 1: 2024-04-12T06:09:52.142Z
  all jobs passed

[kachick]

Thanks for this report!

It seems perhaps it thinks other jobs in the same file are also itself (and therefore shouldn't wait)?

I didn't expect the behavior, I'll check!

[kachick]

Fixed in #755 and released with v3.0.0 🙏

[jimduchek]

Excellent, this fixes that, but introduces two new issues that don't show in v2... isn't that the way it always goes :D

1. If "wait-list" is used, and that job doesn't run at all (in my case, the job I'm waiting for doesn't always need to run), we get: throw new Error("No targets found except wait-other-jobs itself"); I think perhaps the proper fix here, if possible, might be to add a third parameter to wait-list to indicate the job is optional, or (much easier) just an overall "jobs-might-not-exist" parameter that skips that Throw? Because if it's not optional, this new behavior is actually Not A Bug But A Feature, as it catches misspelled job names and such.
2. If "wait-list" is NOT used, and the job is a matrix, it deadlocks. Adding it's own workflowFile and jobName to skip-list does NOT fix it. So I guess it's somehow seeing matrix jobs as a 'different job'.

[kachick]

Thanks for your testing and reports! I have extracted the problems to #760 and #761.

I don't have enough time in this weekend, but should be resolved in next versions 🙏