ci: Utah images are never rechunked, because reusable-build skips chunkah on the testing stream

[hanthor]

scripts/configure-services.sh says "Utah rechunks with chunkah, through projectbluefin/actions' reusable-build". It does not. The published image (ghcr.io/projectbluefin/utah@sha256:0ffe7751…, embedded in today's ISO run) has no dev.hhd.rechunk.info or ostree.components annotations, and its layer list is the raw build: 32 small base layers, three chunkah layers inherited from the Hummingbird base image, then Utah's Containerfile steps as ordinary layers, including one 4 GB COPY (#130 covers that layer's content).

The cause is in projectbluefin/actions reusable-build.yml at the pinned eb4c546, lines 276–291:

      - name: Rechunk Image
        if: github.event_name != 'pull_request' && inputs.stream_name != 'testing'

Utah's build.yml passes stream_name: testing for every build, because Utah's only stream is testing (promotion to main/stable moves tags, it does not rebuild). So the skip meant for Bluefin's throwaway testing builds applies to every image Utah ever publishes, and the sbom upload is skipped by the same condition.

Options

1. Add an input to reusable-build.yml (e.g. rechunk: "true") that overrides the stream-based skip, and set it from Utah. This is the clean fix but lives in projectbluefin/actions.
2. Rename Utah's stream. Tags and the promotion workflows key on testing, so this touches sync-main-to-testing.yml, promote-testing-to-main.yml, post-testing-e2e.yml and scripts/resolve-e2e-inputs.py.

Either way, fix #130 first: rechunking a 4 GB directory of RPMs into layers would only spread the problem across more of them.

scripts/configure-services.sh lines 38–42 should stop claiming the image is rechunked until it is.

[hanthor]

I cannot push to projectbluefin/actions from this session, so the fix that matters (an input on reusable-build.yml overriding the stream_name != 'testing' skip on the apply-pkg-intervals, Rechunk Image and SBOM steps, lines 276–291 at eb4c546) needs someone with access there. The Utah-side half (configure-services.sh no longer claiming the image is rechunked) rides on the image-size PR. Do the repository removal in that PR first; rechunking 4 GB of RPMs into layers would only spread them.

---

Generated by Claude Code