Repository navigation
Conversation
projectbluefin-countme ran `bootc status --json` from an unconfined service. bootc's SELinux probe (chcon to a test label) was denied mac_admin, logging two AVC denials on every boot (projectbluefin/utah#1198). Read the ref from the booted deployment's .origin file (found through ostree= on the kernel command line), falling back to the boot-time snapshot /run/ublue-os/booted-image. Non-bootc hosts keep their image-tag fallback. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WcWm7tJ5iop2b4GMkKSPAv Signed-off-by: Danathar <Danathar@users.noreply.github.com>
|
/hold |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: castrojo, danathar The full list of commands accepted by this bot can be found here. Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
Follow the composefs=/composefs.digest= command-line arguments to /sysroot/state/deploy/<id>/<id>.origin, the same lookup utah#1220 uses, and avoid a stray error when the snapshot file is missing. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WcWm7tJ5iop2b4GMkKSPAv Signed-off-by: Danathar <Danathar@users.noreply.github.com>
|
The utah confirmation is in: with CountMe skipped, the review VM logged |
|
@Danathar this PR is back in your court ( Failing checks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (2 waiting): fixes are handed out before new work as soon as you have capacity. |
There was a problem hiding this comment.
Summary: This looks correct to me. The script no longer runs bootc, and the .origin lookup gives the same ref that bootc status gave. I found two small notes, and neither blocks the PR.
What I checked at 9460399:
- ostree path:
ostree=/ostree/boot.N/<stateroot>/<csum>/<serial>resolves throughreadlink -fto…/deploy/<csum>.<serial>. Adding.origingives the deployment's keyfile. The unit isProtectSystem=strictwithProtectHome=yes(projectbluefin-countme.service:29-30), so/sysrootstays readable. The comment change at:25-26is accurate. - Transport stripping: The prefix stripping in
booted_ref()covers all five transports that the new test runs. The#shortest match onostree-remote-*:*:stops after the remote name, anddocker://is removed afterwards. So the result matches the old.status.booted.image.image.image. - Fallbacks: When the
ostree=path has no.origin, the script falls through to the snapshot. When there is no snapshot,refis empty and the existing "no booted image" branch runs unchanged. So Dakota/Utah live ISOs are still not counted, and Server still usesimage-tag. - Test guard: The
bootcstub plus the teardown check[ ! -e bootc.called ]makes every test enforce the "never runs bootc" rule.
Notes (info only):
- The comment in
booted_ref()says the composefs lookup is "the same lookup as utah's ublue-booted-image-sync". On utahmain, that script still callsbootc status --jsonand has no.origin/composefs lookup. Maybe the comment refers to an unmerged utah change. If so, please name it; if not, please reword the comment. I did not find anything in the repo that confirms the/sysroot/state/deploy/<id>/<id>.originlayout. The new composefs test checks the script against that layout, but it does not prove that bootc writes it there. - In
tests/test_projectbluefin_countme.bats, the "Bluefin Classic and LTS never contact the service" loop hasrun "${SCRIPT}"indented at 4 spaces instead of 8. This is only cosmetic.
The failing GNOME — common E2E job failed on #1690 too, at around the same time. I did not confirm that it is the shared #1688 portal problem, but nothing in this diff touches the session or the portals.
Confidence: 5/5 (safe)
— hive: agent=reviewer-2 backend=claude model=claude-opus-5-5 claude=2.1.325
|
/hold cancel Releasing my hold; this is ready. The utah confirmation came in on 11-15 (no What's left before it can merge:
|
Pull request was closed
…e when the fixer can push it (#19308) * 🐛 adoption: keep a same-repo PR in place across a base workflow change when the App can push it canPushToPRBranch ran baseSyncTouchesWorkflows before its same-repo check, and baseSyncTouchesWorkflows assumed the App can never push a base-sync that carries a .github/workflows/ change. That holds for a fork head (the fork is not in the installation), but on a branch of the base repository the push lands once the installation grants workflows: write. So the hive superseded same-repo PRs it could have adopted in place, closing them and losing their approvals (projectbluefin/common#1689 -> #1708, #1690 -> #1710). A same-repo head now supersedes on a base workflow change only when hgithub.Client.WorkflowsPushable is false (no grant, or no trusted or merger lane serving the repo). Fork heads and compare failures behave as before. Adopted-PR recovery also logs a refused update-branch with the merge sweep's reasons (fork-workflow-update, workflow-permission) instead of the catch-all update-branch-failed, which is what projectbluefin/utah-packages#1173 and projectbluefin/actions#752 showed: both are fork heads that no App grant can update. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1 Signed-off-by: Danathar <6772335+Danathar@users.noreply.github.com> * 🐛 adoption: gate in-place workflow base-sync on the fixer's own token, not any lane Review of the previous commit: WorkflowsPushable asks whether any trusted or merger lane serves the repo, but an adopted head that conflicts (or stays red) is pushed by the fallback fixer that routeConflictFixAgent picks, scanner. A hive with ci-maintainer at trusted and scanner at contributor tier passed the gate, adopted in place, and then scanner's push was refused, the #12865 dead end. - pkg/github: FallbackFixAgent is the one name for that fixer, used by routeConflictFixAgent and the adoption sweep. AgentWorkflowsPushable reports an agent's own capability: its tier asks for Workflows (SetWorkflowsSelfPushFunc) and the installation grants workflows: write. - automerge: baseSyncTouchesWorkflows keeps a same-repo head in place only when AgentWorkflowsPushable(FallbackFixAgent) holds. The comment now says the premise of #12865 changed (#14664 found the grant; pr_merge_state.go fences only forks). - Tests: TestAgentWorkflowsPushable for the real predicate (contributor fixer with the grant, trusted fixer, token client, read-only grant, failed read); TestAdoptedRecoveryForkWorkflowRefusalNamed drives recoverByUpdateBranch end to end on a fork head. - Docs and changelog name the fixer condition and the DCO in-place repair it also unlocks. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1 Signed-off-by: Danathar <6772335+Danathar@users.noreply.github.com>
…ase-red memory is published (#19311) * 🐛 base red: read the base tip until the memory is published after a restart The base-red index and #18842's per-base memory live in the process and are first published when the first CI-enriched enumeration after a start finishes. On knuckle that took about 25 minutes on 2026-11-18 (restart at 1:26 PM EST, first base-branch-red route again at 1:53 PM). In that window every base read as not red, so PRs red only on main's own E2E check went to 070/072, and projectbluefin/common#1689 and #1690, already past their 072 grace, were adopted at 1:42 PM and superseded (#1708, #1709 -> #1710), losing a maintainer's approval, into copies red for the same reason. Until the memory is published (new StageLabelPolicy.BaseRedRememberedReady, wired to whether stageBaseRedMemory has been stored), foldBaseRed reads the base tip's checks itself when the memory does not cover a red head's red checks: one combined-status plus check-run listing per base, cached per base for humanChecksTTL (10 min), failures included. A head whose every red check is red on the tip, touching no CI-owning path, is base red and goes to 050/060 instead of 072, so it is neither nudged nor adopted. An unreadable tip keeps today's author-facing route. Once the memory is published nothing changes and no tip is read. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1 Signed-off-by: Danathar <6772335+Danathar@users.noreply.github.com> * 🐛 base red: hold automatic adoption until the base-red memory is published Review follow-up to the startup base tip read. That read can still miss the base's red: a tip run still queued or in progress, a tip with no run of the check, or a failed read. Adoption cannot be undone, so while StageLabelPolicy.BaseRedRememberedReady is false adoptContributorPR now returns before anything is recorded for an automatic adoption, logging "adoption deferred until the base-red memory is published" once per head and due reason. An operator's Adopt now and an adopt-supersede request are not held. This costs no API and also covers the unpublished shared-CI incident index. The 072/nudge side keeps the tip read. A failed base tip read is now reused for a minute (baseTipRedErrorTTL) instead of the full humanChecksTTL. Scope: this covers common#1689 -> #1708 (and #1690's first adoption into #1709 at 1:42 PM EST). #1690 -> #1710 at 3:38 PM was the orphan re-adoption path, which does not read base red; that is out of scope here. Docs and changelog now describe the deferral, that the tip read counts commit statuses as well as check runs (unlike the enrichment's read), and the 3 to 27 minute window. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1 Signed-off-by: Danathar <6772335+Danathar@users.noreply.github.com>
…198) (projectbluefin#1708) Supersedes projectbluefin#1689 (adopted after CI went red and the contributor was unavailable; original commits by @Danathar preserved). This branch is projectbluefin#1689's head `9460399a87e7` as pushed by @Danathar (fork `projectbluefin/common`), unchanged. Last lane reason on the original: `optional-check-failing:E2E — composed common suite / e2e / GNOME — common`. **Hive fix agent:** rebase this branch on `main`, fix the failing checks / review asks, and push to THIS branch. Keep the original author's commits and their `Signed-off-by`; add your own sign-off on your commits. Do not open another PR. --------- Signed-off-by: Danathar <Danathar@users.noreply.github.com> Co-authored-by: Danathar <Danathar@users.noreply.github.com> Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
On hold: pending the utah confirmation PR
Do not merge until the utah test PR (masking CountMe) shows that CountMe is the source of the boot-time SELinux denials in projectbluefin/utah#1198. That PR is projectbluefin/utah#1247.
Summary
projectbluefin-countmeranbootc status --jsonat boot, as root, from an ordinary service (unconfined_service_t).chconto a test label) is then deniedmac_admin. That logs two AVC denials on every boot of Utah..originfile (found throughostree=on the kernel command line,container-image-reference=, with the ostree transport anddocker://stripped);/sysroot/state/deploy/<id>/<id>.originfromcomposefs=/composefs.digest=(the same lookup as utah#1220ublue-booted-image-sync);/run/ublue-os/booted-image.image-tagwhen no booted image is found..originfile.I did not use
/usr/libexec/ublue-booted-image, because it runsbootc statusfirst.Tests
tests/test_projectbluefin_countme.batsnow fakes a deployment (cmdline, symlink,.origin) instead of stubbing bootc. A bootc stub on PATH leaves a marker, and every test fails if it was called. New cases cover each transport form, a registry port, the snapshot fallback, origin winning over the snapshot, and a missing.origin. bats is not installed on my machine, so CI runs them. I ran the script by hand once against a fake deployment and it reporteddakota/main:stable.Not proven yet
That CountMe is the source of the denials (the utah test PR checks it), and the composefs lookup, which is copied from utah#1220 but not run on a real composefs host.
Context: projectbluefin/utah#1198, evidence run https://github.com/projectbluefin/utah/actions/runs/43615159948
🤖 Generated with Claude Code
https://claude.ai/code/session_01WcWm7tJ5iop2b4GMkKSPAv