I just made a PR for #C04NLA7QHH9 to master (a habit of mine). Usually I just switch branch and it works, but this time the lint step kept failing since it used the

--release

param. Claude tells me: The

--release

step in the template's

linting.yml

is gated on

github.base_ref

, which comes from the event payload. Switching the base fires

pull_request: edited

, which isn't in the default trigger set, so no new run happens — and re-running the failed job replays the original payload, so it still sees

base_ref: master

and runs

--release

again. It then fails on

release: 2.2.0dev

, which a dev-version pipeline can never pass. Closing and reopening the PR fixes it (fresh

reopened

event,

--release

correctly skipped), but that's not an obvious move when the error says "release". Hit this on nf-core/phyloplace#72. The gating itself looks long-standing — the only recent change I found is #4121 adding

main

alongside

master

— so I suspect it's GitHub's behaviour around base changes rather than anything in tools. Anyone seen this before / worth an issue?

If I remember correctly, it can also get fixed after you push something, after you've changed the branch from main/master (even just an empty line)

Yeah, switching branch itself is not enough, you have to commit to re trigger the tests to run

Maybe it was always the case that I switched branch but always had the CHANGELOG number to commit, so I never saw this.

I've done this more times than I'd like to admit, the answer is always

git commit -m "Re-run CI" --allow-empty && git push