Repository navigation
chore(deps): update dependency ubuntu to v26 - #176
mergeraptor[bot] wants to merge 1 commit into
Conversation
castrojo
left a comment
There was a problem hiding this comment.
Reviewed against the API and the run logs rather than the PR description. The headline: the advertised change is fine, the runner label is valid, one of the two failures is transient infrastructure — but the PR silently contains an OS-base bump that would turn main red, and that part is not fixable from inside this branch.
The ubuntu-26.04 bump itself is sound. The label is GA (GitHub announced it 2026-09-17, the day before this opened) and it demonstrably works here: on commit a136c92 the unit job ran on a real runner with "labels":["ubuntu-26.04"] and passed all six steps including just test-unit, and track-refs also succeeded on ubuntu-26.04. So this does not break CI by naming a nonexistent runner, which was my first concern.
The build-kernel exit 255 is transient upstream infrastructure, not the runner. I pulled the log for run 35329005749:
START gnome-build-meta.bst: Fetching from https://gitlab.gnome.org/GNOME/gnome-build-meta.git
FAILURE [d395112a] gnome-build-meta.bst: failed to fetch: unexpected http resp 503 for
https://gitlab.gnome.org/GNOME/gnome-build-meta.git/info/refs?service=git-upload-pack
A 503 from gitlab.gnome.org. Note check-release-version.py actually PASSED on that run ("release-version 26.08.0 matches the pinned FSDK point release"), which rules out the drift theory for this particular failure. Nothing to fix; it needs a re-run.
The blocking problem: two undisclosed junction bumps. This PR's title, body and Renovate table describe only an ubuntu runner bump, but it also carries:
elements/freedesktop-sdk.bst:26.08.0-0-gdb97cce->26.08.1-0-gb02b59felements/gnome-build-meta.bst:50.4-0-g8524013->50.4-5-gd78f498
An FSDK point release and five commits of gnome-build-meta — the base of the OS — inside a runner-label PR. Merging as-is lands FSDK 26.08.1 while project.conf still pins release-version: "26.08.0", and check-release-version.py is the first thing just validate runs and fails closed on drift. That is exactly why the build job died in four seconds at "Validate Element Graph" on a136c92: the build job checks out the branch tip at 09:24:41, after track-refs pushed 26.08.1 at 09:23:52, so it validated 26.08.1 against a declared 26.08.0.
Why there are no checks on the current head, and why I can't just strip the hunks. Head eba9f344 was pushed by github-actions[bot], not Renovate. build.yml's track-refs job fires on every pull_request event for renovate/* branches, re-resolves the junction track: globs, and pushes the result back onto the branch. The runs for that head exist but sit at conclusion: action_required with "total_count": 0 jobs — created, gated, never dispatched — which is why gh pr checks shows nothing and the combined status is an empty pending rollup. Any push I make to clean this up re-triggers track-refs and re-injects the same two hunks.
So this needs a decision rather than a patch, and it is the same decision @hanthor raised on #172 and #49:
- Accept the FSDK move deliberately — bump
project.confrelease-versionto 26.08.1 in this same change and retitle it to say so; or - Fix
track-refsfirst so junction tracking happens on its own scheduled branch rather than being injected into unrelated dependency PRs.dakotaalready does this withtrack-bst-sources.yml.
I have closed #172 because its advertised payload was a verified no-op, so stripping the injected hunks would have left nothing. This PR is different — the ubuntu bump is real and worth having — so I am leaving it open. But it cannot merge while a runner-label PR silently carries an OS-base bump that reddens main.
Not approving. Next move is a maintainer call on (1) vs (2); I think (2) is right and should be its own issue, since it currently affects this PR and #49 identically.
There was a problem hiding this comment.
HUMAN DECISION NEEDED — choose whether to accept FSDK 26.08.1 (requires bumping project.conf release-version) or land #180 first and reduce this to a clean runner bump.
Correctness review at head eba9f34:
- Blocker (verified):
elements/freedesktop-sdk.bst:8moves the junction tofreedesktop-sdk-26.08.1-0-gb02b59f, butproject.conf:29still declaresrelease-version: "26.08.0"..github/scripts/check-release-version.pyfails closed on exactly this drift, so merging turnsmain's build red at "Validate Element Graph". The check exists because drift would otherwise ship new tags whose assets keep the old version string andsystemd-sysupdatefleets silently stop updating — do not bypass it; resolve it. - The runner-label hunks themselves (
.github/workflows/build.yml:28,66,189,kernel.yml:30,unit-tests.yml:25) are correct:ubuntu-26.04is a GA label and theunitjob already ran green on it. - Note:
build.yml'strack-refsjob re-injects the junction hunks on any push to thisrenovate/*branch, so reverting them in-place won't stick — the two paths above are the only durable resolutions.
Not verified by this review: whether the 5 gnome-build-meta commits (elements/gnome-build-meta.bst:8, 50.4-0 → 50.4-5) are behaviorally safe; the PR's own build died at validation before exercising them.
— hive: agent=reviewer backend=copilot model=claude-fable-5 copilot=1.0.78
|
Read-only note (I can't approve or start CI here): this PR's three workflow runs ( |
|
@castrojo — I can't approve or start CI here (read access only). The three workflow runs on this PR are waiting for an admin to approve them; until that happens nothing about the runner bump or the two BuildStream ref bumps has been verified. Once CI is green I'd be happy to review the result, but the approval has to come from someone with write access. |
|
Blocked on #180, and structurally red until it lands.
#180 fixes the root cause: it moves junction tracking to a scheduled Also note #84, which fixes the Leaving open, not closing: the runner and action-digest hunks here are independently wanted. Please rebase after #180. |
There was a problem hiding this comment.
HUMAN DECISION NEEDED — choose whether to accept the FSDK 26.08.1 bump (and bump project.conf release-version to match) or land #180 first and strip the junction hunks back to a pure runner bump.
intent-alignment
The Renovate body claims only a ubuntu 24.04 → 26.04 runner update, but the diff also bumps elements/freedesktop-sdk.bst:8 (26.08.0-0-gdb97cce… → 26.08.1-0-gb02b59f…) and elements/gnome-build-meta.bst:8 (50.4-0-g8524013… → 50.4-5-gd78f498…). As merged, this is guaranteed to break main: project.conf:29 pins release-version: "26.08.0", and .github/scripts/check-release-version.py fails closed on exactly this drift (its docstring explains the consequence: assets published under the old version string, so systemd-sysupdate sees no change and the fleet silently stops updating). The maintainer note on the PR already documents this and that the hunks are re-injected by the track-refs job (build.yml:26-31), so they cannot be durably reverted on this branch. This cannot merge as-is; the resolution (bump release-version here vs. wait for #180) is a maintainer call.
No findings from: security, style, docs-currency.
— hive: agent=reviewer backend=copilot model=claude-fable-5 copilot=1.0.78
|
Two separate things here, and the runner bump is not the problem.
One gap for after the rebase:
|
|
Closing in favor of Renovate update PR #183 (and non-existent runner image ubuntu-26.04). |
eba9f34 to
0293f91
Compare
|
Closing: ubuntu-26.04 runner image is incompatible with BuildStream tar extraction of Flatcar rootfs (cannot change ownership to non-root uid/gids in rootless user namespace, failing flatcar/flatcar-usr.bst build with 'Invalid argument'). CI runners must remain pinned to ubuntu-24.04. |
This PR contains the following updates:
24.04→26.04Configuration
📅 Schedule: (UTC)
🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.
♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
🔕 Ignore: Close this PR and you won't be reminded about this update again.
This PR has been generated by Mend Renovate CLI.