Repository navigation
Migrate Swift Lambda build to Amazon Linux 2023 - #6
Merged
Merged
Conversation
provided.al2 is EOL: AWS blocks new-function creation Jul 31 2026 and updates Aug 31 2026, and recommends migrating to an AL2023-based runtime. Also fixes the CI swift-test job, which was failing outright: GitHub Actions now defaults to Node 24 for marketplace actions, and Node 24 needs a newer glibc than AL2 ships, so actions/checkout couldn't even run inside the old container. swift:6.3-amazonlinux2023 has a current glibc and ships git + openssl-devel already, so the CI workflow's manual install step is no longer needed either. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019Ex6D4aZJirG2b5yhjZsNR
swift test had never actually run on Linux before (previously blocked by the CI checkout failure). It surfaced real portability gaps: on Linux, Foundation splits URLSession/URLRequest/HTTPURLResponse into a separate FoundationNetworking module, and ImageIO doesn't exist at all. Guard the imports/usages with #if canImport(...), matching the pattern already used in NetworkRequestManager.swift/ValidatedRequestManager.swift. No behavior change on macOS; ImageResolutionChecker (a macOS-only local dev CLI, never run in CI) gets a Linux stub that returns nil so the package still compiles. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019Ex6D4aZJirG2b5yhjZsNR
Another gap swift test never caught before running on Linux for the first time: Glibc's stdout/stderr overlay (unlike Darwin's) doesn't mark these extern C globals Sendable, so Swift 6 strict concurrency rejects direct references in CollectionReviewer, ImageResolutionChecker, and CollectionLister. These are single-threaded interactive/batch CLI tools, so aliasing the globals as nonisolated(unsafe) locals (the SE-0412-documented escape hatch) is safe and requires no behavior change. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019Ex6D4aZJirG2b5yhjZsNR
nonisolated(unsafe) on my own wrapper globals alone wasn't enough — the diagnostic was on the *read* of Glibc's stdout/stderr in the initializer, which nonisolated(unsafe) on the LHS doesn't launder. Per swiftlang/swift#77866, the actual fix is @preconcurrency import Glibc, and critically it has to be the first import in the file — otherwise Foundation's own transitive (non-preconcurrency) Glibc import wins and the diagnostic persists. Keeping nonisolated(unsafe) on the wrapper lets too: that addresses a separate, independent diagnostic (a new global of non-Sendable type UnsafeMutablePointer<FILE>). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019Ex6D4aZJirG2b5yhjZsNR
bradwindy
added a commit
that referenced
this pull request
Jul 9, 2026
…nstant-time auth, input validation (#31) * fix: address code-review findings (races, SSRF, constant-time auth, input validation) Web app (collection-reviewer-web): - Route progress writes through the existing write mutex so /api/sync and /api/progress no longer race and clobber the progress index (#1). - Give atomicWrite a unique per-write temp name + failure cleanup, so concurrent writers can't collide on a shared *.tmp and leave a partial write / ENOENT crash (#4). - Validate and clamp the /api/images count param (parseImageCount): a non-numeric value no longer silently yields 0 images, and a huge value can't spawn unbounded concurrent lambda fetches (#2). - Clamp a stale/out-of-range progress index in mergeCollections (remapProgress) instead of silently resetting the reviewer to 0 (#5). - Guard await request.json() in the decision + progress routes so a malformed body returns 400 instead of a raw 500 (#7). Swift API Lambda: - constantTimeEquals now reduces each operand to a fixed 32-byte SHA-256 digest before the XOR compare, removing the secret-length timing leak from the previous early length-mismatch return (#6). Converter Lambda: - Enforce an http/https scheme allowlist, re-validate every redirect target against internal IP ranges (IMDS / loopback / RFC1918) while still permitting the legitimate public S3 redirect, and return generic error bodies instead of echoing raw exceptions (#3). Tests: add Vitest plus web unit tests (atomicWrite, remapProgress, parseImageCount, progress mutex, JSON-body guards); extend AuthorizedConsumerTests with unequal-length cases; extend the converter pytest suite with scheme/redirect SSRF and error-opacity coverage. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019YqwmybDadEYqqfDMrHvxs * fix: reject null/primitive JSON bodies in decision + progress routes CodeRabbit review: request.json() returns null (or a bare primitive) without throwing for a body like "null" or "42"; the subsequent destructuring then threw a TypeError outside the try/catch, yielding a 500 instead of the intended 400. Add an explicit non-null-object guard before destructuring in both write routes, and cover it with valid-JSON "null" body tests. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019YqwmybDadEYqqfDMrHvxs --------- 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
1 participant
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
provided.al2is EOL: AWS blocks new-function creation Jul 31 2026 and updates Aug 31 2026, and officially recommends an AL2023-based runtime.swift-testjob was failing outright: GitHub Actions now defaults marketplace actions to Node 24, which needs glibc ≥2.28 — newer thanswift:6.3-amazonlinux2ships, soactions/checkoutcouldn't run inside the container at all.scripts/build.sh, theMakefile,template.yaml'sRuntime, and the CI container toamazonlinux2023/provided.al2023. The AL2023 Swift image already bundlesgit+openssl-devel, so the CI workflow's manual install step is removed too.README.md's stale manual-deploy instructions to match (also fixed an already-staleswift:6.0-amazonlinux2reference along the way).Test plan
sam validate --lintpasses clean (previously warned aboutprovided.al2EOL).swift-testjob goes green on this PR (real AL2023 container on GitHub-hosted arm64 runner — this is the actual verification, since Docker wasn't available locally to dry-run).python-teststill passes (unaffected by this change, included as a sanity check).deployjob stays skipped on this PR (by design — only runs on push tomain).🤖 Generated with Claude Code
https://claude.ai/code/session_019Ex6D4aZJirG2b5yhjZsNR