Skip to content

Conversation

@3405691582
Copy link
Collaborator

As configured, the same set of tests runs on pull request and on the merge queue. This is redundant, and flakes can cause large delays. Ideally, we should reduce the redundancy, either by pushing tests from pull request to merge, or vice versa. The consensus is on the former, to ensure that the code at HEAD is most likely to be passing tests.

As I understand it, the push GitHub workflow event triggers on the merge queue, so the new pull request workflow triggers on the pull_request event. This workflow does not run any of the build or end-to-end checks, which can take a long time, only the static and unit checks.

@3405691582 3405691582 requested a review from cjreynol November 14, 2025 17:59
Copy link
Collaborator

@cjreynol cjreynol left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The issue with this PR is we require all of the status checks to pass via our main ruleset, but Github (annoyingly) does not have a notion of separate status checks for PRs and the merge queue. I am totally on board with this separation of jobs, to be clear.

My understanding is the workaround for this is "running" the action, but have them be no-ops that "pass" automatically in the context you do not want them to run. I believe the discussion I linked in chat does it via conditionals on the event type, though I have seen others in the past that do it via declaring separate, identically named jobs.

One other point, if we are skipping certain jobs for PRs, can we add the workflow_dispatch event type to the trigger list for the skipped jobs? This should allow us to manually start action runs, which I imagine we will want for debugging eventual failures without needing to enqueue every time.

https://docs.github.com/en/actions/how-tos/manage-workflow-runs/manually-run-a-workflow#configuring-a-workflow-to-run-manually

@3405691582 3405691582 force-pushed the workflow branch 2 times, most recently from 834faf1 to d033ad7 Compare November 14, 2025 18:45
As configured, the same set of tests runs on pull request and on the
merge queue. This is redundant, and flakes can cause large delays.
Ideally, we should reduce the redundancy, either by pushing tests from
pull request to merge, or vice versa. The consensus is on the former, to
ensure that the code at HEAD is most likely to be passing tests.

As I understand it, the `push` GitHub workflow event triggers on the
merge queue, so the new pull request workflow triggers on the
`pull_request` event. This workflow does not run any of the build
or end-to-end checks, which can take a long time, only the static
and unit checks.
@3405691582
Copy link
Collaborator Author

The issue with this PR is we require all of the status checks to pass via our main ruleset,

I'm not sure I understand this point. Can you elaborate?

Adding workflow_dispatch is a good idea; done.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants