NVIDIA
Disable workflows for a forked repository
![github-actions[bot] avatar](https://avatars.githubusercontent.com/in/15368?v=4 "Published by a pub.dev verified publisher"){kind=small}
Thanks Ben for reviewing this PR.
Here are the scheduled workflows
clean_up gh_registry integration_tests nvqc_regression_tests
Here are the ones which run on a PR in local repository (as can be seen from this PR)
CLA bot Basic content checks
Do we need all of these to be disabled or just the ones that automatically run on a schedule?
Just to be on the safer side, thought of disabling all. In case, if some tests later get added in the schedule workflows list.
Also, I know I sometimes run the integration_tests.yml from my account, and I'd like to be able to continue to do that as needed. Is there any way to only disable the scheduled runs in forks and retain the ability to run them on demand?
Yes, there is a way to run any tests in the local PR by removing the condition for forked repo.
Also, I know I sometimes run the integration_tests.yml from my account, and I'd like to be able to continue to do that as needed. Is there any way to only disable the scheduled runs in forks and retain the ability to run them on demand?
Yes, there is a way to run any tests in the local PR by removing the condition for forked repo.
Maybe I'm in the minority here, but I'd still like to be able to manually run the nightly integration tests from my repo (screenshot below) without having to make my main branch diverge from the upstream main branch by committing a change to revert integration_tests.yml back to how it is right now.
E.g.,
Would it be possible to retain this capability by adding structuring the if check like this?
if: ${{ github.event_name == 'workflow_dispatch' || ! github.event.repository.fork }})
I believe this would only need to be done for the ones that have this:
on:
workflow_dispatch:
We should also get @bettinaheim to weigh in here.
Does this check prevent manually running the yml file?
The original issue was logged because these scripts are "on by default". So if you have a fork, you have to go through them all one by one and disable them, or set up email filters to send all the failures to your deleted folder or both. For a fork, these scripts should all be "off by default" since many/most won't have access to the runners, etc. anyway.
Perhaps relevant: https://github.com/orgs/community/discussions/26704#discussioncomment-5268498
That thread makes it look like users have to manually accept workflows for new forks, at least as of March 2023.
FWIW, I picked 3 random forks that were opened in the last few months, and they all contain 0 Actions runs:
- https://github.com/InderpalSuthar/cuda-quantum/actions
- https://github.com/prateekchawla168/cuda-quantum/actions
- https://github.com/mletras89/cuda-quantum/actions
So maybe GitHub's changes supersede the need for this change?
@bmhowe23 Should I proceed with the changes as per the suggestions made here?
@bmhowe23 Should I proceed with the changes as per the suggestions made here?
Perhaps we should discuss at the team meeting. I think that the original issue (#1852) isn't relevant anymore due to GitHub's changes mentioned above. For example, the 3 freshly forked repos I mentioned above have 0 Actions runs, so it seems like the underlying issue won't appear for any new forks.
This is @bettinaheim comment on this PR
"A repository can always disable workflows in the GUI:
Personally, I find it a bit impractical if one needs to make a code change to enable the workflows on a fork. If there is a strong preference for that and no complaints, however, I am good with it."