[haywolf]

Fixes #77

Root cause: release built --mac --x64 --arm64 from a single npm ci on arm64 macos-latest, so the x64 app shipped arm64 sharp/libvips, canvas and Prisma engines. Linux x64 also crashed on launch, because libvips-cpp.so.8.17.3 stayed inside the asar.

Changes

• release.yml: each architecture builds on its own runner (macos-15-intel, macos-latest, windows-latest, ubuntu-latest, ubuntu-24.04-arm).
  • Install uses npm ci --ignore-scripts, and builds run with --publish never.
  • The per-arch latest-mac.yml files are merged; the job fails if an architecture is missing.
• Windows arm64 installer dropped: Prisma has no win-arm64 engine. Jorben approved dropping it.
• asarUnpack now includes node_modules/@img/**.
• build:* scripts build only the host architecture.
• New scripts/verify-package-arch.mjs, scripts/package-load-check.cjs and the verify:package script check that packaged binaries match the target architecture, that required modules are present, and that they load when the host can run them.
  • The check runs before upload in release, and on a Linux x64 package in CI.
  • 27 unit tests.

Tests:

• test:unit 973 passed, typecheck and build pass.
• The Linux x64 package passes verify:package and launches under xvfb (it crashed before the fix).
• Simulated mac x64, mac arm64, linux arm64 and win x64 packages pass the architecture check.

Not verified: a real launch on an Intel Mac or an arm64 Mac, Windows packaging, and actual runs of release.yml and ci.yml. A release candidate run is needed to confirm these.

Merge order: #72 → #75 → #77 (none of the three PRs touch the same files), then run a release candidate to validate the release pipeline.

[github-actions]

AI Code Review Summary

PR: #80 (fix(release): build each arch natively and verify package architecture (#77))
Preferred language: English

Overall Assessment

No blocking issue was detected in the reviewed diff; keep focused regression testing before merge.

Major Findings by Severity

No major issues identified from the reviewed diff.

Actionable Suggestions

• Address the highest severity findings first and add targeted tests for changed logic.

Potential Risks

• Potential hidden risks remain in edge cases not covered by the current diff context.

Test Suggestions

• Add happy-path + boundary + failure-path tests for touched modules.

File-Level Coverage Notes

• package.json: File was in scope but not fully reviewed before budget limits. (coverage_incomplete)
• vitest.config.ts: File was in scope but not fully reviewed before budget limits. (coverage_incomplete)

Inline Downgraded Items (processed but not inline)

• None

Coverage Status

• Target files: 2
• Covered files: 0
• Uncovered files: 2
• No-patch/binary covered as file-level: 0
• Findings with unknown confidence (N/A): 0

Uncovered list:

• package.json: coverage_incomplete
• vitest.config.ts: coverage_incomplete

No-patch covered list:

• None

Runtime/Budget

• Rounds used: 1/4
• Planned batches: 1
• Executed batches: 1
• Sub-agent runs: 1
• Planner calls: 2
• Reviewer calls: 2
• Model calls: 4/64
• Structured-output summary-only degradation: YES

Reasons:

• planner_structured_output_failed_round_1: Structured output failed after repair: You have no permission to access this resource (api_key_source: payg) (request_id: 01627672289e4225855f1b88b1c8bf67)
• reviewer_structured_output_failed_round_1_general: Structured output failed after repair: You have no permission to access this resource (api_key_source: payg) (request_id: 80afd0940f4b4f93b218d422f3752be2)