Repository navigation
[APPX-14752] Remediate Arnica code risk findings for Agent SDK - #468
Open
Matjaz Pirnovar (Mat001) wants to merge 7 commits into
Open
Matjaz Pirnovar (Mat001) wants to merge 7 commits into
Matjaz Pirnovar (Mat001) wants to merge 7 commits into
Conversation
Clears the Arnica SCA/SAST/IaC/license/reputation findings on optimizely/agent@master. Scoped against govulncheck ground truth (28 reachable vulnerabilities -> 0), not just the Arnica version list (several of whose versions/line numbers were stale). Go dependencies (go.mod / go.sum): - Bump Go toolchain 1.24.7 -> 1.26.11 (clears 14 Go-stdlib findings: net/http, crypto/tls, html/template, crypto/x509). Updates the 8 go-version pins and GIMME_GO_VERSION (feeds the Docker build) in CI. - otel stack 1.21.0 -> 1.45.0, grpc 1.59.0 -> 1.83.1, x/net -> 0.60.0, go-chi/chi/v5 -> 5.5.0, plus transitive bumps (protobuf, x/crypto, x/text, x/sys). testify is pulled to 1.11.1 by otel and cannot be pinned back without regressing otel/grpc to vulnerable versions. - Remove the stale replace block; its forced "safe" versions were now below the new baselines (and were the source of the x/crypto v0.45.0 flag). Code & IaC: - SAST XSS (pkg/handlers/notification.go): write the already JSON-encoded SSE payload with w.Write instead of fmt.Fprintf, removing the flagged format-string-to-ResponseWriter pattern; drop the now-unused fmt import. - Add a wget-based HEALTHCHECK to Dockerfile.alpine (admin /health on :8088). Bump its base images golang/alpine 3.21 -> 3.23 since Go 1.26.11 is not published for alpine3.21. Dockerfile.static is FROM scratch with no shell/HTTP client, so its omitted HEALTHCHECK is documented inline. Python (example/test deps only, not the Agent binary): - examples/requirements.txt: bump certifi, requests, urllib3, idna. - tests/acceptance/requirements.txt: bump requests; pytest 7.0.0 -> 8.4.2. Documentation: - docs/dependency-security-notes.md records the accepted (non-remediable) reputation and MPL-2.0 license findings, all confirmed indirect deps. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
❌ ErrorsYour PR has failed checks. Please review the issues below and take necessary action before merging. 🚦 5 Pipeline jobs failed
Useful? React with 👍 / 👎 This comment will be updated automatically if new data arrives.🔗 Commit SHA: 698c0bf | Docs | View more details | Give us feedback! |
Arnica flags direct advisories regardless of reachability, so it reported grpc@v1.83.1 (1 vuln) and x/net@v0.60.0 (2 vulns) even though govulncheck found them unreachable. Bump grpc 1.83.1 -> 1.83.2 and x/net 0.60.0 -> 0.62.0 (pulling transitive x/crypto/x/sys/x/text/x/sync bumps) to clear the findings. govulncheck remains at 0 affected vulnerabilities. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The tests_acceptance_test job failed at setup-python with "Version 3.9 with arch x64 not found" -- the upgraded ubuntu-latest runner no longer provides Python 3.9 (runner drift, not caused by the dependency bumps; the failure is before pip install). Bump actions/setup-python v3 -> v5 and Python 3.9 -> 3.12 (quoted to avoid float parsing). 3.12 is compatible with the pinned acceptance dependencies. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Go 1.26 toolchain bump broke the lint job: golangci-lint v1.64.2
(pinned in the Makefile) is built with Go 1.24 and refuses to analyze a
module targeting a newer Go ("the Go language version (go1.24) used to
build golangci-lint is lower than the targeted Go version (1.26.11)").
The v1 line never shipped a Go 1.26 build. The code uses no post-1.24
language features, so pin run.go to 1.24 as a bridge until golangci-lint
is upgraded to a v2 release built with Go 1.26+.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ately Reverts the setup-python v3->v5 / 3.9->3.12 change. The acceptance job is broken by pre-existing runner drift (ubuntu-latest dropped Python 3.9) and the old pinned openapi_core 0.17.1 does not work on modern Python, which is independent of these Arnica security changes and would fail on master too. Keeping this PR scoped to the security findings; the acceptance-test modernization is tracked under a separate infra ticket. The tests/acceptance/requirements.txt pytest/requests bumps are retained since they clear the Arnica Python findings (static scan). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The lint job timed out at golangci-lint's default 1m ("Timeout exceeded").
The dependency graph grew (otel v1.45, grpc v1.83) and the Go 1.26
toolchain makes the first uncached analysis slower. Raise run.timeout to
5m. (No lint findings -- it was purely a timeout.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Go 1.26 toolchain bump broke the lint job: golangci-lint v1.64.2 is built with Go 1.24 and cannot analyze a module targeting Go 1.26 (x/net v0.62.0 requires go 1.26, forcing the module's go directive to >=1.26, so lowering it is not an option). The v1 line never shipped a Go 1.26 build. - Makefile: bump golangci-lint v1.64.2 -> v2.15.0 and update the lint invocation for v2 flags (--out-format removed; use --output.tab.path). - .golangci.yml: migrate to the v2 schema (golangci-lint migrate). gosimple and stylecheck are merged into staticcheck in v2; typecheck is dropped; run.timeout is no longer needed (disabled by default in v2); the stale revive.min-confidence setting is removed. Fix the findings v2's stricter staticcheck surfaces in non-test code: - pkg/metrics/metrics.go: strings.Replace(-1) -> strings.ReplaceAll; preallocate convertedArray. - pkg/server/server.go: drop tls.Config.PreferServerCipherSuites (ignored since Go 1.18, SA1019). - pkg/syncer/pubsub.go: convert if/else-if chain to a tagged switch. - pkg/optimizely/optimizelytest/config.go, plugins/userprofileservice/ services/rest_ups.go: preallocate slices. golangci-lint v2.15.0 run (matching the Makefile flags) reports 0 issues; build and affected package tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Remediates the Arnica code risk findings for the Agent SDK (
optimizely/agent@master). Scoped against govulncheck ground truth — 28 reachable vulnerabilities → 0 — rather than Arnica's list alone (several of whose versions/line numbers were stale, e.g.x/crypto v0.45.0/x/net v0.47.0were already recent).Go dependencies (
go.mod/go.sum)1.24.7 → 1.26.11— the only way to clear the 14 Go-stdlib findings (net/http,crypto/tls,html/template,crypto/x509). Also updates the 8go-versionpins andGIMME_GO_VERSION(which feeds the Docker build) in.github/workflows/agent.yml.1.21.0 → 1.45.0, grpc1.59.0 → 1.83.1, x/net→ 0.60.0, go-chi/chi/v5→ 5.5.0, plus transitive bumps (protobuf, x/crypto, x/text, x/sys).replaceblock — its forced "safe" versions were now below the new baselines (and were the source of Arnica'sx/crypto v0.45.0flag).testifyis pulled to1.11.1by otel and cannot be pinned back without regressing otel/grpc into vulnerable versions.Code & IaC
pkg/handlers/notification.go): write the already-json.Marshal'd SSE payload viaw.Writeinstead offmt.Fprintf, removing the flagged format-string-to-ResponseWriter pattern; dropped the now-unusedfmtimport.wget-based probe (/healthon admin:8088) toDockerfile.alpine, and bumped its base imagesgolang/alpine3.21 → 3.23(Go 1.26.11 is not published foralpine3.21).Dockerfile.staticisFROM scratch(no shell/HTTP client, no health subcommand), so its omitted HEALTHCHECK is documented inline with the orchestrator-probe alternative.Python (example/test deps only — not the Agent binary)
examples/requirements.txt: bumpcertifi,requests,urllib3,idna.tests/acceptance/requirements.txt: bumprequests;pytest 7.0.0 → 8.4.2.Documentation
docs/dependency-security-notes.mdrecords the accepted / non-remediable findings — the 5 reputation and 4 MPL-2.0 license findings, all confirmed indirect deps (jwx→lestrrat, prometheus→matttproud, go-kit→VividCortex, viper/go-sdk→hashicorp). MPL-2.0 is compatible for unmodified deps in this Apache-2.0 project.Test Plan
go build ./...✅go vet ./...— no new warnings (thecopylocks/lostcancelwarnings in*_test.goare pre-existing onmaster) ✅gofmtclean on changed files ✅go mod tidystable (no diff) ✅govulncheck ./...→ 0 affected vulnerabilities ✅go test ./...passes exceptpkg/syncer/pubsub(requires a live Redis — environmental, unrelated to this change; CI runs these with a Redis service) ✅golang:1.26.11,golang:1.26.11-alpine3.23,alpine:3.23. (Docker daemon unavailable locally for a full image build — CI will exercise it.)Jira
https://optimizely-ext.atlassian.net/browse/APPX-14752
🤖 Generated with Claude Code