Problem
Nothing validates main after a merge.
.github/workflows/pr-test.yaml triggers on pull_request, schedule and workflow_dispatch. There is no push trigger, so a merge to main runs no validation at all.
On 2026-09-08 four PRs merged in succession — #533, #544, #446, #514 — one of which (#446) changes the base image and the apt strategy in Dockerfile_ODELIA. No run validated the result. Establishing whether main was healthy required dispatching the workflow by hand.
It was healthy: run 34236734953 came back with 25 steps passing and only step 23 failing. But that was luck confirmed after the fact, not a signal the project produced on its own.
The weekly schedule does not fill the gap
0 5 * * 0 is the only automatic check on main, and it has failed every time it has run — 2026-08-16, 08-23, 08-30, 09-06, all schedule / failure. The cause is a one-line omission in the dispatcher (#552), unrelated to the health of main.
A check that is always red carries no information. It has been red long enough that a genuine regression on main would have looked exactly the same as the previous four weeks.
Suggested fix
The second matters as much as the first: a post-merge run that nobody trusts is no better than no run.
Related
Problem
Nothing validates
mainafter a merge..github/workflows/pr-test.yamltriggers onpull_request,scheduleandworkflow_dispatch. There is nopushtrigger, so a merge tomainruns no validation at all.On 2026-09-08 four PRs merged in succession — #533, #544, #446, #514 — one of which (#446) changes the base image and the apt strategy in
Dockerfile_ODELIA. No run validated the result. Establishing whethermainwas healthy required dispatching the workflow by hand.It was healthy: run 34236734953 came back with 25 steps passing and only step 23 failing. But that was luck confirmed after the fact, not a signal the project produced on its own.
The weekly schedule does not fill the gap
0 5 * * 0is the only automatic check onmain, and it has failed every time it has run — 2026-08-16, 08-23, 08-30, 09-06, allschedule / failure. The cause is a one-line omission in the dispatcher (#552), unrelated to the health ofmain.A check that is always red carries no information. It has been red long enough that a genuine regression on
mainwould have looked exactly the same as the previous four weeks.Suggested fix
push: branches: [main]topr-test.yaml, so a merge is validated.The second matters as much as the first: a post-merge run that nobody trusts is no better than no run.
Related