Skip to content

[16.0-stable] Bump alpine for openssh update - #7048

Open
shjala wants to merge 2 commits into
lf-edge:16.0-stablefrom
shjala:backport/16.0-openssh-10.6p2
Open

shjala wants to merge 2 commits into
lf-edge:16.0-stablefrom
shjala:backport/16.0-openssh-10.6p2

Conversation

@shjala

@shjala shjala commented Nov 5, 2026 •

Copy link
Copy Markdown
Member

Description

Adapt the OpenSSH package-cache update from #7054 to 16.0-stable. Rebuild from the branch's existing eve-alpine-base, rather than a populated eve-alpine cache: build-cache.sh retains cached APKs, so the previously selected image would retain vulnerable OpenSSH 10.0_p1-r10.

Update the cache rebuild marker to produce a new image tag and regenerate dependent pins. Preserve stable package sources, Go 1.24.6 and the compactor base. Keep cross-compilers and all their consumer references unchanged; add cross-compilers to the hash-check exceptions as requested in review. Existing ACRN, acrn-kernel, kvm-tools and bpftrace-compiler exceptions remain unchanged.

The bootstrap commit adapts master commit fb4c22e409e14a824eae7b995560091ee3fea21d; the dependent-image update derives from fc203c28772424f4d469cf9541edd827ee63e67e. Public-key-only hardening is already present on this branch. The manual APK overlay and unused openssh-test stage are absent.

How to test and validate this PR

  • Passed Docker hash consistency (make -o pkg-deps.mk check-docker-hashes-consistency), SPDX and whitespace checks. The make option suppresses dependency-file generation, not hash validation.
  • Independently computed with linuxkit pkg show-tag: Alpine cda433a7600c2486a9011c2de8b7e676483e25ac; debug 8e18f3bedf5f8ff681713ebe0773cf20c2a907af. The installer debug pin matches.
  • Confirmed no changes to the cross-compilers Dockerfile or cross-compiler image references relative to current 16.0-stable.
  • CI must build the new cache and dependent packages on supported architectures. Verify the resulting debug image contains Alpine OpenSSH 10.0_p1-r11 or a newer fixed revision; new image contents have not been verified locally.
  • Verify sshd -T disables password and keyboard-interactive authentication, and test authorized SSH keys and SFTP. Run Eden/evetest after approval.
  • Full image builds, runtime tests and device validation have not been run locally for this revision.

Changelog notes

Refresh Alpine packages to include the OpenSSH password-authentication security fix. SSH remains public-key only.

PR Backports

Checklist

  • I've provided a proper description
  • I've added the proper documentation (scope and validation instructions)
  • I've tested my PR on amd64 device
  • I've tested my PR on arm64 device
  • I've written the test verification instructions
  • I've added a reference link to the original PR
  • PR's title follows the template
  • I've provided validation limitations for unchecked items

@shjala shjala added stable Should be backported to stable release(s) security Provides a security fix labels Nov 5, 2026
@github-actions

github-actions Bot commented Nov 5, 2026 •

Copy link
Copy Markdown
Contributor

Code Review — [16.0-stable] OpenSSH 10.6p2 backport

  • Read PR diff and full pkg/debug/Dockerfile context
  • Check that test accounts/keys don't ship to the final image
  • Verify debug-image pin consistency across the repo
  • Review register-sbom-pkg.sh and tests/openssh.sh
  • Post findings

This is a clean, well-documented backport that mirrors the already-reviewed master #7046. Verified the important safety properties:

  • Test accounts are confined. tests/openssh.sh appends ssh-test/ssh-salt to /out/etc/{passwd,shadow}, but it only runs in the test stage; the shipped FROM scratch image copies --from=build /out/, so no test account or password hash ships. ✅
  • Publickey-only mitigation is enforced and verified. The Include is prepended (first-match wins), the drop-in sets PasswordAuthentication no / KbdInteractiveAuthentication no, and the test stage asserts it via sshd -T. ✅
  • Supply-chain pinning is sound. OpenSSH tarball is SHA‑256 pinned via ADD --checksum; SBOM registration records 10.6p2. ✅
  • Debug-image pin. pkg/installer/Dockerfile is the only consumer of the build pin; the other eve-debug: references in the repo (bpftrace test fixture, docs) are unrelated pins and are not affected. ✅

One minor finding (posted inline):

  • The openssh-test stage (pkg/debug/Dockerfile:48‑51) is dead/redundant — it isn't in the dependency graph of the shipped image or the test stage, so it never runs during a normal build, and it re-runs the same openssh.sh the test stage already executes. Wire it in as an early-fail gate or drop it.

No correctness, security, or performance blockers found.

View job run

@shjala
shjala force-pushed the backport/16.0-openssh-10.6p2 branch from 3f978b1 to 9da95e5 Compare November 5, 2026 17:01
Comment thread pkg/debug/Dockerfile Resolved
@github-actions

github-actions Bot commented Nov 5, 2026 •

Copy link
Copy Markdown
Contributor

Claude finished @shjala's task in 4m 49s —— View job


PR Review: Backport OpenSSH 10.6p2 to 16.0-stable

  • Fetch PR diff
  • Analyze Dockerfile changes
  • Analyze register-sbom-pkg.sh
  • Analyze tests/openssh.sh
  • Analyze installer debug-image pin
  • Post inline review

Summary

The backport is carefully constructed and I found no blocking correctness or security issues. Key things I verified:

  • Source build & runtime wiring — openssh/openssh-client/openssh-server are dropped from PKGS and replaced by a checksum-pinned source build. Binaries land at the same paths Alpine used (/usr/sbin/sshd, /usr/bin/ssh, /usr/lib/ssh/sftp-server), the sshd account is recreated, /var/empty is created by install-nokeys, and runtime libcrypto3/libz come from the retained openssl/zlib PKGS. The sshd account and the modified sshd_config propagate into /out and then into the final image.
  • Publickey-only hardening — the Include sshd_config.d/*.conf directive is prepended to line 1 so the 10-eve-pubkey-only.conf drop-in (PasswordAuthentication no / KbdInteractiveAuthentication no) wins by first-match. The test stage validates this with a real sshd -T dump, which correctly reflects production config (the test's own PasswordAuthentication yes lives only in /ssh-test/sshd_config).
  • Test isolation — test accounts/keys are appended to /out only in the test stage (FROM build), which the final image does not COPY from, so they are not shipped. Confirmed.
  • register-sbom-pkg.sh — self-contained, only consumer is pkg/debug; the synthetic C: checksum length check (28 + "Q1" = 30) is correct; xxd is available via busybox. Registered after the last apk -p /out call, so the record survives.

Finding (1, minor)

  • Orphaned openssh-test stage (pkg/debug/Dockerfile L48–51) — a sibling of build that isn't reachable from the final stage, so it only runs with an explicit --target openssh-test. The test stage already runs the same tests/openssh.sh. Posted inline to confirm intent or drop. (inline)

Could not verify (out of scope / no tooling here)

  • The 9ae448…29d5b SHA-256 for openssh-10.6p2.tar.gz and the 1a7658cf… debug-image pin (make check-docker-hashes-consistency / linuxkit pkg show-tag) — relying on the author's stated local checks. A wrong download checksum fails the build safely (no silent bypass).
  • arm64/RISC-V build-time behavior of the chroot-based test (mknod/chroot in BuildKit) — only amd64 was validated per the description.
    · branch backport/16.0-openssh-10.6p2

Comment thread pkg/debug/Dockerfile Resolved
@codecov

codecov Bot commented Nov 5, 2026 •

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 33.55%. Comparing base (c46b596) to head (18f1927).

Additional details and impacted files
@@               Coverage Diff                @@
##           16.0-stable    #7048       +/-   ##
================================================
+ Coverage             0   33.55%   +33.55%     
================================================
  Files                0       15       +15     
  Lines                0     2962     +2962     
================================================
+ Hits                 0      994      +994     
- Misses               0     1968     +1968     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@shjala
shjala force-pushed the backport/16.0-openssh-10.6p2 branch from 9da95e5 to 5cb7b16 Compare November 6, 2026 09:22
@shjala shjala changed the title [16.0-stable] debug: upgrade OpenSSH to 10.6p2 [16.0-stable] debug: use Alpine OpenSSH password-authentication fix Nov 6, 2026
@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Claude finished @shjala's task in 2m 34s —— View job


PR Review: [16.0-stable] OpenSSH backport

  • Read PR diff and full Dockerfile context
  • Verify version/approach consistency
  • Review tests/openssh.sh and openssh.sha256
  • Verify debug-image pin + test-account isolation
  • Post inline findings

Summary

Heads-up: the current branch no longer matches its own PR description. The latest commit (5cb7b160 debug: use Alpine OpenSSH password-authentication fix) replaced the from-source 10.6p2 build with Alpine APK openssh 10.0_p1-r11 pinned by openssh.sha256. The title, body, changelog ("Upgrade OpenSSH to 10.6p2") and test instructions ("confirm ssh -V reports 10.6p2") are now stale — on this branch ssh -V reports OpenSSH_10.0p1. The two earlier bot reviews in this thread describe the old source-build diff and no longer apply.

The safety properties I re-verified against the current code hold up:

  • Test accounts don't ship — tests/openssh.sh appends ssh-test/ssh-salt to /out/etc/{passwd,shadow} only in the test stage (FROM build); the shipped scratch image copies --from=build, so no test account/hash ships. ✅
  • Publickey-only hardening — 10-eve-pubkey-only.conf (PasswordAuthentication no / KbdInteractiveAuthentication no) is added in build, included first-match via the base config, and the test stage asserts it with a real sshd -T. ✅
  • APK install is integrity-pinned — all 6 APKs per arch are SHA-256 verified before apk add --no-network, and openssh-server-10.0_p1-r11 is asserted installed. ✅

Findings (all inline)

  1. Version/doc mismatch (Dockerfile:27) — code installs 10.0_p1-r11; PR title/body/changelog/test-steps say 10.6p2. Reconcile the docs, and state which CVE/upstream commit the Alpine r11 revision backports. (inline)
  2. Build-durability risk (Dockerfile:37) — dl-cdn.alpinelinux.org/.../v3.22/main/ serves only the current revision; when Alpine ships -r12, these -r11 URLs 404 and this LTS build breaks. The SHA-256 pin guards integrity, not availability. Consider mirroring the APKs or returning to the retained OpenBSD source tarball. (inline)
  3. Dead openssh-test stage (Dockerfile:46‑49) — still orphaned (flagged in prior reviews). Not in the build graph; pkg/debug/Makefile only runs --target test, which already runs the same tests/openssh.sh. Wire it in as an early-fail gate or drop it. (inline)

Could not verify (no tooling here)

  • That Alpine openssh 10.0_p1-r11 actually carries the musl password-auth bypass fix, and the 18 APK SHA-256 values — relying on the author's checks. A wrong checksum fails the build safely.
  • make check-docker-hashes-consistency / the 838ab936… debug-image pin, and arm64/riscv64 build-time behavior of the chroot/mknod test — amd64-only per the description.

No correctness or security blocker found in the build logic itself — the main asks are reconciling the stale PR docs and the two Dockerfile cleanups above.

Comment thread pkg/debug/Dockerfile Resolved
Comment thread pkg/debug/Dockerfile Resolved
Comment thread pkg/debug/Dockerfile Resolved
@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Claude finished @shjala's task in 3m 24s —— View job


PR Review — [16.0-stable] OpenSSH password-authentication fix

  • Fetch PR diff and read full pkg/debug/Dockerfile
  • Review tests/openssh.sh and installer pin
  • Verify test-account isolation and publickey-only hardening
  • Post inline findings

Summary

This revision has changed substantially from what the earlier bot reviews in this thread describe: it no longer builds OpenSSH 10.6p2 from source, but overlays Alpine's signed 10.0_p1-r11 APKs via per-arch ADD --checksum scratch stages. The PR title/body/changelog now match that approach. The core safety properties hold up:

  • Test accounts don't ship — tests/openssh.sh appends ssh-test/ssh-salt to /out/etc/{passwd,shadow} only in the test stage (FROM build); the shipped scratch image copies --from=build, so no test account/hash/key is shipped. ✅
  • APKs are integrity-pinned — all 18 APKs (6 × 3 arches) are SHA-256 verified by ADD --checksum before apk add --no-network, and openssh-server-10.0_p1-r11 is asserted installed. ✅
  • Publickey-only hardening is build-time verified — the test stage runs a real sshd -T and asserts PasswordAuthentication no / KbdInteractiveAuthentication no. ✅

Findings (all inline)

  1. Mitigation now depends on Alpine's default Include (Dockerfile:133‑137) — the explicit Include /etc/ssh/sshd_config.d/*.conf sed from the source-build revision was dropped. The 10-eve-pubkey-only.conf drop-in only applies if Alpine's packaged sshd_config carries an active Include; the test stage's sshd -T is the only guard. (inline)
  2. Build-durability / 404 risk (Dockerfile:14, all 18 ADD lines) — dl-cdn…/v3.22/main/ serves only the current revision; when Alpine ships -r12, these -r11 URLs 404 and this LTS build breaks. SHA-256 guards integrity, not availability. (inline)
  3. Dead openssh-test stage (Dockerfile:67‑70) — still orphaned after multiple prior reviews; not in the build graph, never runs, and tests a weaker config than the real test stage. Wire it in or drop it. (inline)

Could not verify (no tooling here)

  • That Alpine 10.0_p1-r11 carries the musl password-auth bypass fix, and the 18 APK SHA-256 values — relying on the author's checks; a wrong checksum fails the build safely.
  • make check-docker-hashes-consistency / the aca4a2b2… debug-image pin, and arm64/riscv64 build-time behavior of the chroot/mknod test — amd64-only per the description.

No correctness or security blocker found in the build logic itself; the main asks are findings 1–3 above.
· branch backport/16.0-openssh-10.6p2

Comment thread pkg/debug/Dockerfile Resolved
Comment thread pkg/debug/Dockerfile Resolved
Comment thread pkg/debug/Dockerfile Resolved
@shjala
shjala force-pushed the backport/16.0-openssh-10.6p2 branch 2 times, most recently from e4ea828 to 00d1598 Compare November 9, 2026 17:07
@shjala shjala changed the title [16.0-stable] debug: use Alpine OpenSSH password-authentication fix [16.0-stable] Bump alpine for openssh update Nov 9, 2026
@shjala
shjala marked this pull request as ready for review November 9, 2026 17:07
@github-actions

github-actions Bot commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

Claude finished @shjala's task in 3m 26s —— View job


PR Review — [16.0-stable] Bump Alpine for openssh update

  • Read the current PR diff (git diff origin/16.0-stable...HEAD)
  • Verify this is an image-reference-only cache refresh
  • Check pkg/debug / installer pin consistency
  • Verify legacy pins preserved and no stale references left behind
  • Post findings

Summary

Heads-up for anyone reading the thread: this revision bears no resemblance to what the earlier bot reviews describe. The branch no longer touches pkg/debug/Dockerfile's SSH logic, tests/openssh.sh, or openssh.sha256 at all — the from-source-build and APK-overlay approaches are both gone. The current diff is a pure image-reference cache refresh: eve-alpine is bumped a5885b38… → a5f43dc9… across 46 Dockerfiles, and the dependent intermediate images are re-pinned. The OpenSSH musl password-auth fix now rides in through the rebuilt shared Alpine package cache rather than any per-package change. All the prior inline findings (dead openssh-test stage, ADD --checksum 404 risk, publickey-only Include) no longer apply to this code.

What I verified against the current diff:

  • eve-alpine bump is consistent — every FROM lfedge/eve-alpine:… and EVE_ALPINE_IMAGE= now points at a5f43dc9…; no package was left on the old a5885b38… hash except the intentionally-excluded eve-tools/bpftrace-compiler (called out in the PR body). ✅
  • Dependent intermediate pins are internally consistent — eve-cross-compilers ee54e131… (bsp-imx, optee-os, pillar), eve-uefi 4bae43fe… (pillar, xen-tools), eve-dom0-ztools 24ee2780… (pillar, vtpm) each match across every file that references them; recovertpm/bpftrace/fscrypt/dnsmasq/gpt-tools/optee-os/debug single-consumer bumps are coherent. ✅
  • pkg/alpine/Dockerfile eve-alpine-base → eve-alpine swap is the standard cache-refresh mechanism (rebase onto the prior published eve-alpine, apk update && apk upgrade -a, rebuild the mirror) — expected for this kind of PR, not a defect. ✅
  • eve-debug has one build-graph consumer (pkg/installer/Dockerfile:31); the other eve-debug: hits are docs/test fixtures with unrelated pins. ✅

Finding (1, minor — inline)

  • Stale verified debug pin (pkg/installer/Dockerfile:31) — the PR description says the debug pin was verified as 0a1efca4…, but the committed pin is 08a4ef58…. The pkg/debug content changed in this PR, so the earlier verification is stale. Re-run linuxkit pkg show-tag pkg/debug and reconcile the description. (inline)

Could not verify here (no build tooling / registry access)

  • That the rebuilt a5f43dc9… Alpine cache (and the dependent digests) are the correct content hashes — this is exactly what make check-docker-hashes-consistency and the CI package builds enforce; a wrong/placeholder hash fails that check safely rather than shipping.
  • That the new cache actually carries OpenSSH 10.0_p1-r11 (or a newer fixed Alpine revision) with the password-auth bypass fix, and that sshd -T on the built debug image still reports PasswordAuthentication no / KbdInteractiveAuthentication no — the PR's own test steps cover this; please confirm on the built image before merge.

No correctness, security, or performance blocker found in the diff itself — it's a clean, mechanical, internally-consistent hash bump. The one ask is reconciling the stale debug pin in the description.
· branch backport/16.0-openssh-10.6p2

Comment thread pkg/installer/Dockerfile Resolved
Comment thread pkg/alpine/Dockerfile Resolved
@rene

rene commented Nov 13, 2026

Copy link
Copy Markdown
Contributor

@shjala , please, do not bump cross-compilers in this PR, we don't need to bump it (it doesn't use any affected package from alpine), and if we bump it here the CI/CD will most likely timeout and not publish the images. We can bump cross-compilers later, btw that's why it's an exception for the docker hash checker....

Rebuild the 16.0-stable package cache from its existing alpine-base. Do not seed it from eve-alpine, whose cached OpenSSH r10 archives are retained by build-cache.sh. Update the rebuild marker to give the fresh cache a new content tag.

Adapted from master commit fb4c22e: retain the stable bootstrap base instead of copying the populated master cache.

Signed-off-by: Shahriyar Jalayeri <shahriyar@posteo.de>
@shjala
shjala force-pushed the backport/16.0-openssh-10.6p2 branch from 300a8a2 to 6dfd9ed Compare November 16, 2026 09:27
@github-actions

Copy link
Copy Markdown
Contributor

Claude Code is working…

I'll analyze this and get back to you.

View job run

@github-actions
github-actions Bot requested a review from rene November 16, 2026 09:27
This commits bumps Alpine package to its latest version so it picks up a
critical fix for openssh.

Signed-off-by: Renê de Souza Pinto <rene@renesp.com.br>
(cherry picked from commit fc203c2)

Backport adjustments: retain 16.0-stable package sources, compactor
base and Go 1.24.6 setting; omit packages absent from this branch.
Regenerate stable Alpine and dependent image tags while preserving
the excluded legacy ACRN, kvm-tools and bpftrace-compiler pins.
Keep cross-compilers and their consumers on the existing published
images and exclude cross-compilers from the hash consistency check.

Signed-off-by: Shahriyar Jalayeri <shahriyar@posteo.de>
@shjala
shjala force-pushed the backport/16.0-openssh-10.6p2 branch from 6dfd9ed to 18f1927 Compare November 16, 2026 09:29
Comment thread pkg/installer/Dockerfile
RUN cp "/usr/local/my-installer/target/$CARGO_BUILD_TARGET/release/installer" /usr/local/my-installer/target/installer

FROM lfedge/eve-debug:0f92dcfac96c6115e8aa1c600d0e369b5415fc11 AS debug
FROM lfedge/eve-debug:8e18f3bedf5f8ff681713ebe0773cf20c2a907af AS debug

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

The committed debug pin here is 8e18f3bedf5f8ff681713ebe0773cf20c2a907af, but the PR description (both the test section and the 2026-11-12 rebase status) still says the installer debug pin was verified as 08a4ef585ddbdacc36cca84351254b642fac26a4. Since pkg/debug's FROM pins (recovertpm/bpftrace/eve-alpine) were bumped again in this revision, the computed pkg/debug tag changed and the documented value is stale. Please re-run linuxkit pkg show-tag pkg/debug, confirm it yields 8e18f3be…, and update the description so reviewers following the test steps don't hit a mismatch. (make check-docker-hashes-consistency remains the real gate.)

@github-actions

github-actions Bot commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

Claude finished @shjala's task in 2m 34s —— View job


PR Review — [16.0-stable] Bump Alpine for openssh update

  • Read the current diff (git diff origin/16.0-stable...HEAD)
  • Confirm @rene's two review requests were addressed
  • Verify eve-alpine bump + dependent-pin consistency
  • Check cross-compilers handling and the hash-check exception
  • Post findings

@rene's feedback — both addressed ✅

  1. Rebase from alpine-base, not a populated cache — pkg/alpine/Dockerfile keeps FROM lfedge/eve-alpine-base:17217d43… (only the comment changed), so the rebuilt cache no longer retains the vulnerable cached OpenSSH. ✅
  2. Don't bump cross-compilers — pkg/cross-compilers/Dockerfile is untouched and all three consumers (bsp-imx, optee-os, pillar) still pin eve-cross-compilers:65d07c4c…. It's added to IGNORE_DOCKERFILE_HASHES_PKGS so the consistency checker tolerates it staying on the old eve-alpine hash. ✅

Consistency — clean mechanical bump ✅

  • eve-alpine a5885b38… → cda433a7… across all 44 consuming references; the only remaining a5885b38… references are the two lines in pkg/cross-compilers/Dockerfile, which is intentionally excluded.
  • Dependent intermediate pins are internally consistent — each of eve-uefi f83932c7… (×2), eve-dom0-ztools 0d0bd5c8… (×2), eve-recovertpm 11dd0074…, eve-bpftrace f8fb9cd1…, eve-optee-os 8a7a0e47…, eve-fscrypt, eve-dnsmasq, eve-gpt-tools resolves to a single hash.
  • Debug pin now matches the PR description — pkg/installer/Dockerfile pins eve-debug:8e18f3be…, matching the body's stated verified value (the earlier 08a4ef58… mismatch is resolved).

Finding (1, minor — inline)

  • Undocumented hash-check exception (Makefile:551) — cross-compilers is added to the ignore list without the explanatory comment the other entries have; it reads as permanent when it's meant to be temporary. Suggested a one-line note inline. (inline)

Could not verify here (no registry/build tooling)

  • That the rebuilt cda433a7… cache and the dependent digests are the correct content hashes — make check-docker-hashes-consistency and the CI package builds are the real gate; a wrong hash fails safely rather than shipping.
  • That the new cache actually carries the fixed Alpine OpenSSH (10.0_p1-r11 or newer) and that sshd -T on the built debug image still reports PasswordAuthentication no / KbdInteractiveAuthentication no — the PR's own test steps cover this; please confirm on the built image before merge.

No correctness, security, or performance blocker found — the diff is an internally-consistent hash bump that correctly incorporates @rene's feedback.
· branch backport/16.0-openssh-10.6p2

Comment thread Makefile
# based eve-alpine on this branch: their upstream sources (ACRN 2.x,
# qemu 5.2) do not build with the alpine 3.22 toolchain.
IGNORE_DOCKERFILE_HASHES_PKGS=alpine acrn acrn-kernel kvm-tools
IGNORE_DOCKERFILE_HASHES_PKGS=alpine acrn acrn-kernel kvm-tools cross-compilers

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Adding cross-compilers here is the right call per @rene's request (avoids the CI timeout and keeps the consumers pinned to the old eve-cross-compilers:65d07c4c…). However, every other entry in this list is explained by the comment block just above (installer's rationale, and why acrn/acrn-kernel/kvm-tools stay on alpine 3.16), while cross-compilers is added with no explanation. A future maintainer won't know this is a temporary exclusion meant to be lifted once cross-compilers is rebuilt on the new eve-alpine.

Consider a one-line note, e.g.:

# cross-compilers is temporarily excluded: it does not consume any affected
# alpine package, so it is intentionally left on the old eve-alpine hash to
# avoid rebuilding it here (CI timeout); bump it in a follow-up.

Also worth confirming before merge: pkg/cross-compilers/Dockerfile still FROMs eve-alpine:a5885b38…, so that old image must remain pullable whenever cross-compilers is next rebuilt.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

security Provides a security fix stable Should be backported to stable release(s)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants