Repository navigation
chore(deps): update taiki-e/install-action digest to 84f5ac3, and gnome-build-meta to 50.4-5 - #49
mergeraptor[bot] wants to merge 7 commits into
Conversation
Edited/Blocked NotificationRenovate will not automatically rebase this PR, because it does not recognize the last commit author and assumes somebody else may have edited the PR. You can manually request rebase by checking the rebase/retry box above. |
🛑 Fix loop escalated — human attention neededThis PR has failed CI on 1 distinct fix attempts (new commits, still red). The hive has stopped dispatching further automated fixes for it. Failing checks: build Raw failure evidence (from check-run annotations): Remove the |
🛑 Fix loop escalated — human attention neededThis PR has failed CI on 1 distinct fix attempts (new commits, still red). The hive has stopped dispatching further automated fixes for it. Failing checks: build Raw failure evidence (from check-run annotations): Remove the |
7dcc062 to
110c070
Compare
Signed-off-by: sec-check[bot] <sec-check[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
hanthor
left a comment
There was a problem hiding this comment.
The digest bump is fine. The rest of this PR is not.
I verified the advertised change and it is correct:
$ git ls-remote https://github.com/taiki-e/install-action | grep 84f5ac3124727fb3d284d4d22ee9ab3654fd09a6
84f5ac3124727fb3d284d4d22ee9ab3654fd09a6 refs/tags/v2.87.7
The digest is a real published tag in the v2 line, so the # v2 comment is accurate. If this PR only touched taiki-e/install-action, it would be ready. It does not.
1. Undisclosed hunk: this PR also moves gnome-build-meta 5 commits past its tag
The title and body describe one thing: 7b8d471 → 84f5ac3. The diff contains a second, unmentioned change:
$ git diff main...pr49 -- elements/gnome-build-meta.bst
- ref: 50.4-0-g8524013485013dd5616bfff02dbeab274d5a4553
+ ref: 50.4-5-gd78f4981a3321e07ae4c62d4a7761de46ac8172b
That came from the track-refs job, not from Renovate. Upstream confirms what each ref is:
$ git ls-remote https://gitlab.gnome.org/GNOME/gnome-build-meta.git "refs/tags/50.4^{}" "refs/heads/gnome-50"
8524013485013dd5616bfff02dbeab274d5a4553 refs/tags/50.4^{}
d78f4981a3321e07ae4c62d4a7761de46ac8172b refs/heads/gnome-50
So main is pinned exactly to the 50.4 tag, and this PR moves it to the current tip of the gnome-50 branch. That is a real upstream change to the OS composition riding inside a PR labelled as a GitHub Action digest update.
2. That bump re-applies a change that was already reverted for breaking the build
This branch's own history says the gnome-build-meta move off 50.4-0 was reverted on purpose. Commit 02c22de / a82e952 on this same branch:
fix: keep server build on compatible GNOME metadata
Revert gnome-build-meta to the known-good 50.4-0 revision. The 50.4-1 update pulls an FSDK change whose linux-firmware composition overlaps runtime files, causing os-rootfs to fail before image generation.
track-refs has now silently pushed it forward again — past 50.4-1, all the way to 50.4-5 — undoing that revert with no discussion.
3. The current head has never been built by CI
mcp__github__pull_request_read(method="get_check_runs", pullNumber=49)
-> {"total_count":0,"check_runs":[]}
Zero checks on fbe321c. The workflow run list explains why:
| head_sha | commit | conclusion |
|---|---|---|
fbe321c (current head) |
chore: resolve BuildStream junction refs… | action_required |
9bf45ec |
ci: retry transient failing checks | success |
110c070 |
chore(deps): update taiki-e/install-action digest | failure |
Every commit authored by github-actions[bot] from the track-refs push lands in action_required and never executes — 7dcc0621, d320675f, 3e06afab, 8c9e8e33, 34d3ffc0 all show the same conclusion. The one green run on this branch (34528096011) was on 9bf45ec, which is the commit before the gnome-build-meta ref was rewritten. Nothing green has ever included the change in item 1.
Worth noting as a workflow bug in its own right: track-refs pushes with the workflow GITHUB_TOKEN, and pushes made with that token do not trigger new workflow runs. The job is structurally incapable of getting its own output tested.
4. Automerge is enabled on this state
The body says 🚦 Automerge: Enabled, and the PR carries both automerge and needs-human. An unbuilt, unreviewed upstream OS bump that was previously reverted for breaking os-rootfs is one green check away from landing on main unattended.
What I'd like to see
- Drop the
elements/gnome-build-meta.bsthunk from this PR — reset it to50.4-0-g8524013, matchingmain. The digest bump then stands alone and is mergeable. - Propose the gnome-build-meta move separately, where it can get a build that actually runs and where the earlier revert can be discussed.
- Until
track-refs's push can trigger CI, treat any commit it authors as untested regardless of what the checks column shows.
Baseline note: main at 7f376fa is fully green locally — 201 passed, 1 xfailed (pytest), 41/41 bats, and all three .github/scripts checkers exit 0. None of the above is pre-existing breakage.
Generated by Claude Code
|
Baseline triage: the same check is red on Tracked as one incident in #175. No action needed on this branch — deferring rather than retrying until the baseline is green. — hive: agent=scanner backend=copilot model=claude-opus-5 copilot=1.0.78 |
castrojo
left a comment
There was a problem hiding this comment.
Reviewed. Mechanically this is the healthiest of the three Renovate PRs, and part of the blocking review is now stale and should be withdrawn. But one objection stands and it is the one that matters.
Stale: the breakage concern no longer reproduces. @hanthor's review (2026-09-12, commit fbe321c) objected partly on the grounds that nothing green had ever included the gnome-build-meta hunk, and that a prior 50.4-1 bump had been reverted on this very branch for pulling an FSDK change whose linux-firmware composition overlapped runtime files and broke os-rootfs before image generation. That was a fair concern then. It is no longer true now: main has been merged in several times since, and the current head 39549f58 has build SUCCESS after a full 54-minute compile with the 50.4-5 hunk in the tree, plus unit SUCCESS and track-refs SUCCESS. The only red check is installer-test, which fails on main itself at tip 5ebfeae (run 34920225526) and is unrelated. So the os-rootfs breakage does not reproduce at 50.4-5.
Also worth noting: head 39549f58 is a merge of origin/main by a human, not a track-refs bot commit, which is why this PR has real CI at all — unlike #176 and the now-closed #172, whose bot-pushed heads sit at action_required with zero jobs.
Not stale: the title still does not describe the diff. The PR is titled as a taiki-e/install-action digest bump, and that part is correct and well-formed — 84f5ac3124727fb3d284d4d22ee9ab3654fd09a6 # v2 is a full 40-char SHA with an accurate version comment in all three call sites, complying with docs/skills/ci-tooling.md, and @hanthor separately confirmed the digest resolves to refs/tags/v2.87.7. But the diff also moves elements/gnome-build-meta.bst from 50.4-0-g8524013 (the 50.4 tag) to 50.4-5-gd78f498 (the tip of the gnome-50 branch), which appears in no title, body, or Renovate table.
That is the live objection, and it is not something I can resolve by editing the branch: track-refs re-injects the hunk on the next pull_request event, so reverting that one line is not durable.
What I think should happen. The honest option is to accept the gnome move explicitly — amend the title and body to state that this bumps gnome-build-meta to the tip of gnome-50 alongside the action digest — now that a full green build covers it. Silently reverting the line would be undone by CI on the next push, and holding the PR indefinitely gains nothing since the underlying behaviour is a workflow design problem, not an author mistake.
@hanthor: items (3) and (4) of your review are overtaken by events — there is now a green build including the hunk. Could you confirm whether item (1) is satisfied by a retitle, or whether you want the gnome bump split out regardless? You are the blocking reviewer and I would rather not override a CHANGES_REQUESTED on a judgement call that is yours.
Housekeeping: the needs-human label is left over from the 2026-09-08/09 hive escalations, which referenced a build: exit code 255 that has since resolved. Worth clearing.
Not approving while the CHANGES_REQUESTED stands. The next move is @hanthor's, not the author's.
There was a problem hiding this comment.
Reviewed (correctness) — no blockers at head 39549f5.
installer-test / Install and boot in QEMUfails at this head, but it fails identically on main tip (5ebfeae, PR #157 merge) — pre-existing, not caused by this PR. Don't misattribute it to the gnome-build-meta bump.- Verified
84f5ac3124727fb3d284d4d22ee9ab3654fd09a6resolves torefs/tags/v2.87.7; all three call sites updated consistently (.github/workflows/build.yml:40,78; .github/workflows/unit-tests.yml:42). - gnome-build-meta
50.4-0→50.4-5(elements/gnome-build-meta.bst:8) is covered by a green 54-minbuildat this head, consistent with the maintainer triage note.
— hive: agent=reviewer backend=copilot model=claude-fable-5 copilot=1.0.78
|
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. |
|
Tracked as a single repository incident in #218. Failure mode is Please don't retry this check per-PR — it will stay red until #218 is resolved. — hive: agent=scanner backend=copilot model=claude-opus-5 copilot=1.0.78 |
Shared CI — not a defect in this PR
Four open PRs (#49, #80, #208, #211) carry this identical failure, which makes it one Already fixed on
|
This PR contains the following updates:
7b8d471→84f5ac3Configuration
📅 Schedule: (UTC)
🚦 Automerge: Enabled.
♻ 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.
Maintainer note (castrojo, PR triage sweep). I retitled this PR to disclose a change it was silently carrying.
Besides the advertised
taiki-e/install-actiondigest bump, the diff also moveselements/gnome-build-meta.bstfrom50.4-0-g8524013(the 50.4 tag) to50.4-5-gd78f498(tip of thegnome-50branch). That appeared in no title, body, or Renovate table. It is not the author's doing:build.yml'strack-refsjob fires on everypull_requestevent forrenovate/*branches and pushes re-resolved junction refs onto the branch, so the hunk is re-injected on any push. Reverting the line would not stick. PR #180 moves that tracking onto a scheduled workflow with its own PR, which is the durable fix.I have retitled rather than reverted because the gnome bump now has real evidence behind it: head
39549f5hasbuildSUCCESS after a full 54-minute compile including that hunk, plusunitandtrack-refsgreen. @hanthor's review predated that and reasonably objected that nothing green had ever covered it, and that an earlier50.4-1bump had been reverted on this branch for breakingos-rootfs. That breakage does not reproduce at50.4-5.The digest itself is confirmed good:
84f5ac3124727fb3d284d4d22ee9ab3654fd09a6resolves torefs/tags/v2.87.7, full 40-char SHA with an accurate# v2comment in all three call sites, complying withdocs/skills/ci-tooling.md.No code was changed. Title and this note only.