workflows/pr: block merging PRs when jobs have been cancelled

[wolfgangwalther]

This currently happens, for still unknown reasons, for the "check cherry picks" job. The job gets cancelled by GHA mid-way. This should be the same as an error, because an important check didn't run: Merging should be blocked and auto-merge should not succeed.

This only fixes the symptoms of #433053 (comment), will look into the actual cause next.

Things done

• Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.

---

Add a 👍 reaction to pull requests you find important.

[philiptaron]

I suspect we'll discover this isn't what we need, but it's worth the experiment.

[nixpkgs-ci]

Successfully created backport PR for release-25.05:

• [Backport release-25.05] workflows/pr: block merging PRs when jobs have been cancelled #433080

[wolfgangwalther]

I suspect we'll discover this isn't what we need, but it's worth the experiment.

What makes you believe this?

It seems it indeed doesnt work, but I don't know why. In https://github.com/NixOS/nixpkgs/actions/runs/16909083566?pr=433053, the check cherry picks job is cancelled due to timeout, but the "no PR failures" job is still skipped. It should not be, according to the new condition.

[wolfgangwalther]

I suspect we'll discover this isn't what we need, but it's worth the experiment.

What makes you believe this?

Reading the docs carefully:

cancelled

Returns true if the workflow was canceled.

and

failure

Returns true when any previous step of a job fails. If you have a chain of dependent jobs, failure() returns true if any ancestor job fails.

OK, that's clearly a difference. I had assumed cancelled() to work the same way as failure()... but it does not.

[philiptaron]

It's a probabilistic belief. GHA's semantics are knowable but not obvious at every turn, and this turn -- what happens on timeout? what happens on cancel? what happens on re-run? -- are likely to be no different.

I'm approving because I buy your intent in the PR headmatter and this looks like it will accomplish it. I just have low faith in GHA and given that we can't read the source, we'll have to muddle through with experiments in production.

Experiments I am very glad you're running.