Fail to push to slack thread with failed step

[mamatt63]

Orb version:

circleci/slack@5.0.0

What happened:

Unable to have slack/notify adding a message to a previously created thread when event is set to fail.

After a 1st successful slack/notify call with event: always and thread_id: my-workflow, When calling slack/notify with event: fail a thread_id: my-workflow, it fails to .

Expected behavior:

The fail event is expected to be pushed in the specified thread.

Additional Information:

Here is the pseudo CI configuration to reproduce the issue:

jobs:
  start-slack-thread:
    steps:
      - slack/notify:
          event: always
          thread_id: my-thread
  a-job-that-fails:
    steps:
      - run: exit 1
      - slack/notify:
          event: fail
          thread_id: my-thread
      - slack/notify:
          event: pass
          thread_id: my-thread
  a-job-that-passes:
    steps:
      - run: exit 0
      - slack/notify:
          event: fail
          thread_id: my-thread
      - slack/notify:
          event: pass
          thread_id: my-thread
...
workflows:
  version: 2

  my-workflow:
    jobs:
      - start-slack-thread:
      - a-job-that-fails:
          requires:
            - start-slack-thread
      - a-job-that-passes:
          requires:
            - start-slack-thread

In the CircleCI, I've noticed that the cached /tmp/SLACK_THREAD_INFO/<channel_name> is correctly restored in the a-job-that-passes job. However in the a-job-that-fails job, the cache restore does not happen.

As a workaround, I forces the cache restore. So IMO, it's either an expected behavior to ignore the thread ID is case of failure and this behavior should be explained somewhere, or a bug.

  a-job-that-fails-and-notifies:
    steps:
      - restore_cache:
           keys:
             - cache-my-thread-<< pipeline.id >>
      - run: exit 1
      - slack/notify:
          event: fail
          thread_id: my-thread
      - slack/notify:
          event: pass
          thread_id: my-thread

I hope it helps, thanks!

[marboledacci]

@mamatt63 It looks like the restore_cache doesn't have a when parameter as the save_cache has.
https://circleci.com/docs/configuration-reference/#restorecache

I will internally see if it is posible to add this functionality, for now the only option is the workaround you are using.

[AustinMaddox]

I guess this is still a problem. I just came across this issue trying to post a thread reply upon failures in circleci/slack@5.2.1 orb and it erroneously posted the Slack notification as a regular slack message instead of a thread reply.

Also, I don't quite understand the workaround. Is the restore_cache key a convention of some sort?

Is it supposed to be:

cache-{{put my thread id here?}}-<< pipeline.id >>

❓

[marboledacci]

@AustinMaddox The key cache-{{thread_id}}-<< pipeline.id >> is what the orb uses to save that status, so restoring that same cache manually prevents the problem