diff --git a/.github/workflows/publish-template-image.yml b/.github/workflows/publish-template-image.yml index 9f3c5d7..572d15c 100644 --- a/.github/workflows/publish-template-image.yml +++ b/.github/workflows/publish-template-image.yml @@ -100,7 +100,54 @@ jobs: - name: Set up Docker Buildx uses: docker/setup-buildx-action@v3 - - name: Build & push template image to GHCR + - name: Build template image (load for smoke test, do not push yet) + # Build into the runner's local docker first so the smoke test can + # actually boot the image. We push :latest + :sha-* only AFTER the + # smoke test passes — this is the gate that prevents broken images + # from poisoning :latest. Background: 2026-04-27 outage where the + # template's adapter.py imported a symbol (RuntimeCapabilities) + # that the published runtime didn't yet export. The old smoke + # test only inspected the entrypoint string, so the broken image + # shipped to GHCR and every workspace provision hung. + uses: docker/build-push-action@v6 + with: + context: . + file: ./Dockerfile + platforms: linux/amd64 + load: true + push: false + tags: ${{ steps.tags.outputs.image }}:sha-${{ steps.tags.outputs.sha }} + cache-from: type=gha + cache-to: type=gha,mode=max + labels: | + org.opencontainers.image.source=https://github.com/${{ github.repository }} + org.opencontainers.image.revision=${{ github.sha }} + org.opencontainers.image.description=Molecule AI workspace template — ${{ steps.tags.outputs.runtime }} runtime + + - name: Smoke test — boot image and import adapter.py + # The real boot test. Runs `python -c "import adapter"` inside the + # image, which exercises: + # - adapter.py exists at /app/ + # - all `from molecule_runtime...` imports resolve against the + # pip-installed runtime version (catches the version skew + # class of bug — symbol added to runtime but PyPI not yet + # republished, or template pinned to old runtime, etc.) + # - no syntax errors in adapter.py + # We bypass the gosu/agent entrypoint with --entrypoint sh because + # we don't need workspace permissions for an import check. + shell: bash + env: + IMAGE: ${{ steps.tags.outputs.image }}:sha-${{ steps.tags.outputs.sha }} + run: | + set -eu + docker run --rm --entrypoint sh "${IMAGE}" -c \ + "cd /app && python3 -c 'import adapter; print(\"adapter imports cleanly:\", adapter.__name__)'" + echo "::notice::✓ ${IMAGE} adapter.py imports cleanly against installed runtime" + + - name: Push image to GHCR (post-smoke) + # Now that the smoke test passed, push both tags. build-push-action + # reuses the cached build from the load step above, so this is fast + # — it's effectively a layer push, not a rebuild. uses: docker/build-push-action@v6 with: context: . @@ -116,20 +163,3 @@ jobs: org.opencontainers.image.source=https://github.com/${{ github.repository }} org.opencontainers.image.revision=${{ github.sha }} org.opencontainers.image.description=Molecule AI workspace template — ${{ steps.tags.outputs.runtime }} runtime - - - name: Smoke test the pushed image - # Pull the tag we just pushed and verify the entrypoint is set. - # Catches "image pushed but binary missing" regressions without a - # full end-to-end provision test. We don't `docker run` — most - # templates need platform env (WORKSPACE_ID, PLATFORM_URL, etc.) - # to actually boot, so inspection is the right layer here. - shell: bash - env: - IMAGE: ${{ steps.tags.outputs.image }}:sha-${{ steps.tags.outputs.sha }} - run: | - set -eu - docker pull "${IMAGE}" - docker inspect "${IMAGE}" --format '{{.Config.Entrypoint}} {{.Config.Cmd}}' \ - | tee /dev/stderr \ - | grep -qE '.' || { echo "::error::Image has empty entrypoint+cmd"; exit 1; } - echo "::notice::✓ ${IMAGE} pulled and entrypoint verified"