Repository navigation
Conversation
[scanner] CI root cause identified:
|
[scanner] CI red is systemic, two clusters: (1) all
|
|
[scanner] Root cause of the rebuild/srpm failures: RPM transaction conflict between ruby3.3/ruby3.4/ruby4.0 default-gems (file conflicts on /usr/share/gems/specifications/default/*). Unblocked by utah-packages#83 (exclude legacy ruby default-gems from Hummingbird repo) — after #83 merges, rebase this branch onto main and CI should clear. scanner agent (ACMM L5 — hold-gated mode)🐝 Hive Agent: — hive: agent=scanner backend=copilot model=kimi-k3 copilot=1.0.78 |
hanthor
left a comment
There was a problem hiding this comment.
Verified the checksum against the real upstream release. It is correct, and this is ready as far as I can tell.
The hash matches sigstore's published checksums exactly. Fetched cosign_checksums.txt from the v2.4.1 release:
$ curl -sSL https://github.com/sigstore/cosign/releases/download/v2.4.1/cosign_checksums.txt | grep cosign-linux-amd64
8b24b946dd5809c6bd93de08033bcf6bc0ed7d336b7785787c080f574b89249b cosign-linux-amd64
against the line this PR adds:
echo '8b24b946dd5809c6bd93de08033bcf6bc0ed7d336b7785787c080f574b89249b cosign' | sha256sum -c -
Identical. And the placement is right — after the curl, before install -m 0755 cosign /usr/local/bin/cosign, so an unverified binary never becomes executable. Worth having: this binary is what signs repomd.xml for the published repo that projectbluefin/utah consumes, and it was being fetched over the network and run with no integrity check at all.
The CI failures on this PR are entirely pre-existing on main. Worth stating plainly, because the check list looks alarming for a one-line change. This PR's run (34616966239) failed these builds:
dracut, git, tailscale, fish, openssh, mozjs140, samba, webkitgtk,
gjs, firefox, evolution-data-server, pipewire-libs-extra,
gstreamer1-plugins-bad-free
The latest scheduled run of unmodified main at the same commit this PR branches from (run 34571911239 @ 54cc8d1) failed:
gcc, dracut, evolution-ews, fish, tailscale, git, openssh,
pipewire-libs-extra, samba, evolution-data-server,
gstreamer1-plugins-bad-free, webkitgtk, mozjs140, gjs, firefox
Same set. Nothing here caused any of them. Spot-checking the causes confirms they are unrelated to cosign: dracut and git both die on the Hummingbird ruby3.3/3.4/4.0-default-gems file conflict (job 103175696635 — file /usr/bin/erb conflicts between attempted installs of ruby4.0-default-gems-4.0.6-37.3.hum1 and ruby3.3-default-gems-3.3.10-23.6.hum1), which #83 and #58 both address. fish dies in %check on three root-sensitive tests (job 103175696704).
The one failure on this PR's run that is not on the main baseline is srpm (zsh), and that is the pruned Packit digest, also unrelated:
docker: Error response from daemon: manifest for quay.io/packit/packit@sha256:149e6e06... not found: manifest unknown
##[error]Process completed with exit code 125.
#85 fixes that one.
Local gates are clean, identical to the unmodified-main baseline:
PR#72 fc=0 val=0 pytest=0 | validated 193 source RPMs | 72 passed in 0.80s
PR#72 quoting=0 runtime_contract=0 unittest=0 | validated runtime contract: 72 packages
And it does not conflict with #80, which is the other open PR touching rebuild-rpms.yml. I merged both into main and the added line survives intact:
$ git merge pr/72 && git merge pr/80
4 files changed, 642 insertions(+), 77 deletions(-)
$ grep -n "sha256sum -c" .github/workflows/rebuild-rpms.yml
375: echo '8b24b946...249b cosign' | sha256sum -c -
Two small suggestions, neither blocking:
- The version and the hash are now two separate literals that must be kept in step. A
COSIGN_VERSIONvariable used in both the URL and a comment above the hash would make the coupling visible, and would stop a future version bump from silently leaving a stale hash that then fails the build for a confusing reason. sigstore/cosign-installerdoes this (and signature verification of the release itself) as a pinned action, if you would rather not maintain the hash by hand. Given this repo already pins actions by SHA, that would fit. Not worth blocking on — the explicit check is perfectly good and arguably easier to audit.
The body closes #46, #66 and #68; I have not checked that all three are fully addressed by this one line, so worth a glance before closing them out.
Generated by Claude Code
|
[scanner] CI root cause: both failure classes here are systemic, not from this PR.
Once #83 and #87 merge and this branch is rebased, CI should validate this PRs actual changes. Note: the hive App token currently lacks the scanner agent (ACMM L5 — hold-gated mode)🐝 Hive Agent: — hive: agent=scanner backend=copilot model=kimi-k3 copilot=1.0.78 |
hanthor
left a comment
There was a problem hiding this comment.
Approving. Converting my earlier COMMENTED review to an approval now that this carries lgtm — nothing has changed on the branch since I verified it (still e813e37), so the verification in my review stands as-is:
- the hash matches sigstore's published
cosign_checksums.txtfor v2.4.1 byte for byte; - the check sits after the
curland beforeinstall -m 0755, so an unverified binary never becomes executable; - every CI failure on this PR is also present on unmodified
mainat the same base commit (54cc8d1), and the one extra (srpm (zsh), pruned Packit digest) is #85's; - local gates are identical to the
mainbaseline; - it does not conflict with #80, the other open PR touching
rebuild-rpms.yml.
The two suggestions in that review (a COSIGN_VERSION variable to couple the version and the hash, or sigstore/cosign-installer pinned by SHA) were non-blocking then and are non-blocking now — worth a follow-up, not a reason to hold a one-line integrity fix on the binary that signs repomd.xml for the repo projectbluefin/utah consumes.
One thing to check before it lands: the body closes #46, #66 and #68. I have not verified that this single line fully addresses all three, so worth a glance rather than letting them auto-close.
Generated by Claude Code
…e import, and contract (#71, #72, #56, #57) (#92) * test(quality): unit coverage for validate.py and recalculate_hummingbird_gaps.py * fix(ci): verify cosign binary sha256 checksum in rebuild-rpms Verify the downloaded cosign v2.4.1 binary against its release SHA-256 checksum prior to installation and keyless signing. Closes #46 Closes #66 Closes #68 * fix(tools): validate package and branch inputs in import_rawhide (#56) * fix(contract): exclude anaconda-live from runtime transaction (#57) --------- Co-authored-by: castrojo <castrojo@users.noreply.github.com>
Verify the downloaded cosign v2.4.1 binary against its release SHA-256 checksum prior to installation and keyless signing in
.github/workflows/rebuild-rpms.yml.Closes #46
Closes #66
Closes #68