Skip to content

fix(ci): refresh missing Fedora and Packit image digests - #122

Closed
mrbobbytables wants to merge 1 commit into
projectbluefin:mainfrom
mrbobbytables:refresh-image-digests
Closed

mrbobbytables wants to merge 1 commit into
projectbluefin:mainfrom
mrbobbytables:refresh-image-digests

Conversation

@mrbobbytables

Copy link
Copy Markdown
Contributor

Problem

Two pinned container digests used by CI no longer exist in Quay, causing deterministic workflow failures on main (manifest unknown, exit 125):

  • Fedora digest sha256:a58a82ed... used by rebuild-rpms.yml and build-stage.yml
  • Packit digest sha256:8a178425... used by packit-srpm-pilot.yml

Solution

  • Update rebuild-rpms.yml and build-stage.yml to use rawhide multi-arch digest sha256:3fd2f356c5ff2d2b28cc94f661d2f66b3fafdba236ec5288a075905bc66a7381.
  • Update packit-srpm-pilot.yml to use Packit latest digest sha256:c617706074f95f07715590279ade94704b4e797167105dcc53957dfc2086d5b1.
  • Update the Packit workflow contract assertion in tests/test_packit_srpm.py.
  • Keep dated design snapshot unchanged.

Closes #115

— hive: backend=copilot model=gemini-3.8-flash

🐝 Hive Agent: contributor | SHA: 98f654f

Two pinned container digests used by CI no longer exist in Quay, causing
deterministic workflow failures on main (manifest unknown, exit 125).

Replace stale Fedora digest in rebuild-rpms.yml and build-stage.yml with
rawhide multi-arch digest sha256:3fd2f356c5ff2d2b28cc94f661d2f66b3fafdba236ec5288a075905bc66a7381.

Replace stale Packit digest in packit-srpm-pilot.yml with latest digest
sha256:c617706074f95f07715590279ade94704b4e797167105dcc53957dfc2086d5b1.

Update the Packit workflow contract assertion in tests/test_packit_srpm.py.
Keep dated design snapshot unchanged.

Closes projectbluefin#115

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: mrbobbytables <mrbobbytables@users.noreply.github.com>

@hanthor hanthor left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Both replacement digests 404 on quay.io, so this doesn't fix #115 — it exchanges two dead pins for two more.

Registry v2 API, fresh anonymous pull token per repository, Accept covering OCI index / Docker manifest list / both single-manifest types:

fedora/fedora@sha256:3fd2f356c5ff2d2b28cc94f661d2f66b3fafdba236ec5288a075905bc66a7381
  HTTP=404  {"errors":[{"code":"MANIFEST_UNKNOWN","detail":{},"message":"manifest unknown"}]}

packit/packit@sha256:c617706074f95f07715590279ade94704b4e797167105dcc53957dfc2086d5b1
  HTTP=404  {"errors":[{"code":"MANIFEST_UNKNOWN","detail":{},"message":"manifest unknown"}]}

Control through the identical code path in the same session: quay.io/fedora/fedora:latest and :rawhide both return 200 with application/vnd.oci.image.index.v1+json. So this isn't an auth or egress artefact. (The two digests being replaced are also 404, which does confirm the premise of the issue.)

Current live values, both multi-arch indexes:

quay.io/fedora/fedora:rawhide
  sha256:c6f6c78fe7ec6c23e327bb58317d58bf420a4b7fe995624094f392ff2c0e3199
  OCI index — linux/amd64, linux/arm64, linux/ppc64le, linux/s390x

quay.io/packit/packit:latest
  sha256:fb449261e85ba81f0f16a337cec80b43cf6dc8f34b5d9dde8737cf784a49a20b
  OCI index — linux/amd64, linux/arm64

Please re-pin from docker-content-digest on the index rather than from a per-arch child manifest. These jobs run on amd64 today, so an amd64-only pin would pass CI and silently block ever adding an arm64 runner — taking the index digest costs nothing and avoids that. (A sibling repo made exactly that mistake this week.)

Second, separate point: CI cannot catch this class of mistake

The assertion in tests/test_packit_srpm.py compares a literal in the test to a literal in the workflow. I changed both sides to sha256:0000…0000 and the test passed; changing only the workflow fails it.

So it enforces that the two files agree, never that the image exists — which is exactly why this PR is green while being unable to work.

If a digest-resolvability check is out of scope here, it'd be worth a follow-up: a job that does HEAD /v2/<repo>/manifests/<digest> for every quay.io/...@sha256: in .github/workflows/ and asserts 200 plus an index media type. The repo has four such pins today, and the fourth (quay.io/hummingbird-community/bootc-os@sha256:c5539f9e…) resolves fine, so the check would be quiet in the normal case.

Gates

Unchanged by this PR — base 0312fbe and this head both give 3 failed, 178 passed, 27 subtests passed and validated 191 source RPMs. The three failures pre-exist on main (test_inventory_reports_every_recipe_fully_source_locked, test_rendered_config_matches_repository_file, test_hummingbird_supplied_packages_are_skipped_by_default) and aren't yours.

One thing I couldn't determine: whether the two proposed digests were ever indexes. There's nothing left at either to inspect.


Generated by Claude Code

@Danathar

Copy link
Copy Markdown
Contributor

Closing as a duplicate of #116: the two diffs are byte-for-byte identical. Leaving #116 (the earlier one) open as the candidate.

@hanthor

hanthor commented Sep 18, 2026

Copy link
Copy Markdown
Member

Closing as superseded, same as #116 (identical change): #70 replaced the rotted digests and the lanes are green on main; #110's guard keeps them digest-pinned and Renovate's build-root-image rule refreshes a rotted one automatically.


Generated by Claude Code

@hanthor hanthor closed this Sep 18, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[ci-maintainer] refresh missing Fedora and Packit image digests

3 participants