Skip to content

INTEL-MACOS-QUALIFICATION.md

Latest commit

 

History

History
49 lines (38 loc) · 2.91 KB

INTEL-MACOS-QUALIFICATION.md

File metadata and controls

49 lines (38 loc) · 2.91 KB
Raw

Intel macOS (x86_64) Qualification — Status

Status: Runner-availability qualified, not release-matrix qualified. Owner lane: tauri-intel-qualification.yml (workflow_dispatch only).

History

  • macos-13 (the last GitHub-hosted Intel runner using that label) was removed from tauri-build.yml's production matrix on 2026-07-28 after 3 consecutive tagged-release runs each had the job sit in GitHub's queue indefinitely, never reaching in_progress — see docs/TAURI-CI.md for the full incident detail.
  • On 2026-08-26, GitHub's replacement label macos-15-intel (introduced 2025-09-18, available through August 2027) was empirically probed via a throwaway workflow_dispatch workflow: it scheduled and completed successfully on this organization's plan.

Current qualification lane

tauri-intel-qualification.yml runs a real pnpm exec tauri build on macos-15-intel, manually triggered only (workflow_dispatch). It is deliberately not:

  • part of the production release matrix (tauri-build.yml's bundle job),
  • wired into latest.json generation or any darwin-x86_64 updater key,
  • able to mutate a GitHub Release (permissions: contents: read only, no signing secrets, no release-publication steps of any kind — verified structurally sound by scripts/workflow-policy-check.mjs's publishing-boundary check, since this job is not on the PUBLISHING_ALLOWLIST).

A single successful qualification build proves the runner label works — it does not by itself prove the build is ready for regular release inclusion.

What's still open before promoting Intel to the release matrix

  1. Soak/repeat testing. One successful run isn't enough evidence that macos-15-intel provisions reliably at release time (the exact failure mode that removed macos-13 was queue unavailability, not build failure — that requires repeated real-world observation, not a single manual dispatch, to rule out).
  2. Artifact parity check against the existing macos-latest (Apple Silicon) bundle — bundle size, .dmg/.app.tar.gz structure, and updater-signature behavior should be verified equivalent before shipping both architectures under one release.
  3. Ongoing maintenance-cost decision. Adding a second macOS architecture to the release matrix roughly doubles macOS CI minutes per release; whether that's worth it for the current Intel-Mac user base is a product decision, not a technical one — out of scope for this qualification lane.
  4. Explicit promotion PR. When (1)–(3) are resolved, promoting means: adding macos-15-intel to tauri-build.yml's bundle matrix, adding a darwin-x86_64 entry to the latest.json generator's expected-platform set, and updating docs/TAURI-CI.md's Build matrix table.

Tracking

Issue: #507.