Repository navigation
Tags
Tags
test: don't assert temp-dir cleanup on Windows in the registry buildp… …ack test (buildpacks#2722) `TestBuild/.../buildpack_is_from_a_registry/all_buildpacks_are_added_to_ ephemeral_builder` fails intermittently on the Windows runner: build_test.go:1798: Expected nil: unlinkat C:\Users\RUNNER~1\AppData\Local\Temp\registry<N>\bp-<N>.tar: The process cannot access the file because it is being used by another process. It failed three consecutive runs of main today, and the first attempt on buildpacks#2721, a PR that changed only .grype.yaml. Since release -> build-additional-archs -> test, a red Windows leg blocks cutting a release. This block creates a buildpack layer tar under tmpDir (ifakes.CreateBuildpackTar) and hands it to the build through a fake package (fakePackage.AddLayer). pkg/buildpack exposes that layer as an openerBlob whose reader is opened lazily via pkg.GetLayer -- a plain os.Open for imgutil/fakes.Image. A reader that outlives the build keeps a handle on the file, and os.RemoveAll then fails with a sharing violation. POSIX allows unlinking an open file, which is why Linux and macOS never see it. Uses the guard this repo already applies to exactly this error, introduced in 686ecb5 ("Fixing flaky error on windows", 2023) and now in 20 places including pkg/buildpack/buildpack_test.go, pkg/archive/archive_test.go and pkg/client/inspect_buildpack_test.go: still assert the cleanup everywhere except Windows, so a real cleanup regression is caught on the other platforms. `runtime` is already imported in this file. Verified: `go test -count=2 -run TestBuild ./pkg/client/` ok, `go vet` clean, `make verify` 0 issues. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
build(deps): bump go-git, otel and klauspost/compress to fix CVEs (bu… …ildpacks#2666) Addresses the vulnerabilities grype reports against the released v0.40.8 image: - github.com/go-git/go-git/v5 v5.19.1 -> v5.19.2 GHSA-hc8v-wwc9-vgxm (High), GHSA-qgq7-7hm3-q39j (Medium) - go.opentelemetry.io/otel v1.43.0 -> v1.44.0 GO-2026-5158 / CVE-2026-41178 (Medium), baggage header not length-capped - github.com/klauspost/compress v1.18.6 -> v1.18.7 GO-2026-5841 / GHSA-259r-337f-4rfw, OOB read in s2 The fifth finding, GO-2026-5932 on golang.org/x/crypto, has no fixed version: it is the advisory marking golang.org/x/crypto/openpgp as unmaintained, introduced at v0 with no fix, so it matches every release of the module and cannot be remediated by bumping. pack does not link that package -- its only importer is go-github/v30 through acceptance/config, which sits behind the `acceptance` build tag and is not part of the shipped binary. Suppressed in .grype.yaml with that rationale; govulncheck agrees the code does not call it. Deliberately scoped to these three modules. The moby/ggcr/docker-cli group bump in buildpacks#2663 also carries the go-git and compress fixes, but it is being held back until imgutil supports the newer moby client. Signed-off-by: Juan Bustamante <bustamantejj@gmail.com>
build(deps): bump Go to 1.25.12 and safe x/* deps to fix CVEs (buildp… …acks#2653) Fixes the CVEs reported in buildpacks#2652 and takes the safe subset of the grouped dependabot bump in buildpacks#2650. The two CVEs failing the released-image grype scan are both stdlib, fixed in Go 1.25.12, so the actual remediation is the go directive: - GO-2026-4970 (High) root escape via symlink plus trailing slash in os - GO-2026-5856 (Medium) Encrypted Client Hello privacy leak in crypto/tls CI builds with go-version-file: go.mod, so bumping the directive is sufficient to rebuild the release image on 1.25.12. Also takes the safe x/* bumps from buildpacks#2650, which resolve without disturbing the docker/moby image stack: - golang.org/x/crypto v0.53.0 -> v0.54.0 - golang.org/x/mod v0.37.0 -> v0.38.0 - golang.org/x/sync v0.21.0 -> v0.22.0 - golang.org/x/sys v0.46.0 -> v0.47.0 - golang.org/x/term v0.44.0 -> v0.45.0 - golang.org/x/text v0.38.0 -> v0.40.0 The docker/cli, go-containerregistry and moby/moby/{api,client} bumps from buildpacks#2650 are held: imgutil still pins the older moby split-client, so forcing pack ahead breaks daemon rebase and manifest annotate acceptance. Tracked upstream in buildpacks/imgutil#307. Signed-off-by: Juan Bustamante <bustamantejj@gmail.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
chore: suppress unreachable docker daemon CVEs in grype config (build… …packs#2644) The grype scan against the pack binary still flagged six docker findings that our .grype.yaml intended to ignore. The existing ignore rules listed the GHSA aliases, but grype matches the printed primary ID and the go-module scan prints the GO-2026-* IDs, so the suppressions never fired. All six are non-impactful for pack: - docker/docker GO-2026-4887, GO-2026-4883, GHSA-x86f-5xw2-fm2r, GHSA-rg2x-37c3-w2rh, GHSA-vp62-88p7-qqf5 are daemon-side (AuthZ bypass, plugin-privilege off-by-one, docker cp races, decompression RCE). pack only uses docker as a client (api/types, client, volume/mounts, pkg/* helpers) and never runs the daemon paths. None are fixed in the github.com/docker/docker module; the fix exists only in the rewritten github.com/moby/moby/v2 module, which the ecosystem has not adopted. - docker/cli GO-2026-4610 is already remediated: we ship v29.4.3, newer than the fixed v29.2.0; grype mis-orders the +incompatible version. Listing every ID form (GO + GHSA) makes the suppressions apply across both the go-module binary scan and the released-image scan. With this, `grype out/pack` reports no vulnerabilities. Signed-off-by: Juan Bustamante <bustamantejj@gmail.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
build: add `syntax` parser directive to Dockerfile (buildpacks#2420) Signed-off-by: JamBalaya56562 <jambalaya.pyoncafe@outlook.jp> Co-authored-by: Juan Bustamante <bustamantejj@gmail.com>
chore: add .grype.yaml to ignore unreachable docker daemon CVEs (buil… …dpacks#2582) Mirror the ignore list used by buildpacks/lifecycle for the same reasons: pack uses docker/docker as a client only, so daemon-side vulnerabilities (plugin install privilege validation, AuthZ plugin bypass) are not reachable from pack's code paths. Grype also flags two long-standing protobuf false positives that are already filtered by lifecycle. This silences the scheduled check-latest-release.yml grype scan, which currently fails on: - GHSA-pxq6-2prw-chj9 (Medium) - GHSA-x744-4wpc-v9h2 (High) Both are fixed upstream in github.com/moby/moby/v2 but not backported to the github.com/docker/docker module, so grype cannot auto-resolve them. Signed-off-by: Juan Bustamante <bustamantejj@gmail.com> Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
chore: upgrade Go to 1.25.8 to fix stdlib CVEs (buildpacks#2548) Fixes CVE-2026-25679 (High), CVE-2026-27142 (High), and CVE-2026-27139 (Low) in the Go standard library by bumping the minimum Go version to 1.25.8. Signed-off-by: Juan Bustamante <bustamantejj@gmail.com> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
chore: fix security vulnerabilities in dependencies (buildpacks#2539) - Bump go stdlib 1.25.6 → 1.25.7 (CVE-2025-68121 critical, CVE-2025-61732 high) - Bump github.com/chainguard-dev/kaniko v1.25.7 → v1.25.10 (GHSA-6rxq-q92g-4rmf high) Signed-off-by: Juan Bustamante <bustamantejj@gmail.com> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
PreviousNext