HI @allanlewis - thanks for the report! We at Upsider do not use too much of matrix jobs, and this sounds like an oversight indeed. We will review this a bit more closely to see if we can fix this 👍

We use matrix jobs extensively where I work (see Test Render jobs in screenshot), and in a super simple test case I am not able to repro this issue. I'm going to test it further in other more complex scenarios to be sure.

As an aside: Thank you very much for making this action open source! It is extremely useful!

Perhaps this was fixed in v1.0.2? Based on the date I raised this, I was probably using v1.0.1. Maybe #28?

That could be! I'll continue testing this out on my pipelines and provide a report if I see the issue.

My team is considering using merge-gatekeeper for matrix jobs. Can anyone confirm the latest status? Is the tool working for matrix jobs?

Thanks all for providing insights! 🥰

I just did a quick test on my side, and it seems that the matrix job handling is working correctly given the recent updates. I have created a test PR in a separate repository, where you can see the matrix jobs are listed out correctly.
rytswd/merge-gatekeeper-test#1

We will make updates to our documentation and add some example test cases to make it clear in this repository as well.

Is this now closed, based on the most recent comments? I'll try out v1.2.0...

@allanlewis Yes, this should be sorted with the latest release (probably fixed from v1.1.1) 👍

Hi,
just want ro report, that I encountered "this" issue today. We have follwoing setup:

CI Workflow

• Job A defines how the matrix should look like
• Matrix B is build based on

Merge Gatekeeper is running with a check interval of 5 seconds and using @v1. I assume this should ensure that it is using latest.

Job A and merge gatekeeper where triggered. Job A calculated the matrix but the Matrix was not yet started and therefore did not exist in GitHub at all. Merge gatekeeper passed because it only saw job A which was successful. A second later GitHub builds up the matrix based on the output of job A.

This happens occasionally and I will increase our check interval to mitigate.

Possible Solution:
Merge Gatekeeper should not only check the state of all jobs it knows. It should also check the state of the parent workflow. Only if the Parent workflow state changes to successful it can stop checking for jobs and their states.

We've been seeing this from time to time as well. Impression is that it's very rare, we recently had it happen once, with about 50 runs that caught the failure (after it had been merged in when it was missed), before we fixed the error.

The matrix needs another job though, and it seems like while that other job is running, the matrix jobs are not listed by merge-gatekeeper. In the cases where failures weren't missed all the jobs were listed as pending at the moment that the needed job was listed as completed, while in the case where the failures were missed, the needed job was completed without the other jobs being listed as pending. Not sure if this is a different issue?