You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Every push to main gets a CI result from the macOS check #397
Since 2026-11-04 at 14:56Z, every CI run on main ends as failed, so main shows red even though nothing in hub is broken. The check (ubuntu-latest) job passes. The check (macos-latest) job is never given a runner, runs no steps, and is cancelled about 15 minutes after it was queued. GitHub's annotation reads "The job was not acquired by Runner of type hosted even after multiple attempts", alongside "Due to capacity constraints, jobs targeting macOS arm64 runners may experience longer queue times." The last run where the macOS job passed was for 7b2cb2d, and nothing under .github/ has changed since.
Ideal state
Every push to main gets a CI result from the macOS check
A CI failure on main means hub failed to build, lint or test, not that GitHub had no runner
Checked 2026-11-05: the check (macos-latest) job has passed on every CI run on main since 2026-11-04 19:32Z (0a5fb30 through 5c7dec4, 7 runs). The last cancelled one was 68cba87 at 19:07Z. #401 has since changed .github/workflows/ci.yml, pinning actions and tools, but the job still runs on macos-latest. Nothing in hub stops the failure returning the next time GitHub runs short of macOS runners.
One workaround under consideration is running CI on Linux only, cited as cutting cost about 10x. Trade-offs to think through before choosing it:
Cost: GitHub says standard hosted runners are free in public repositories, and hub is public, so dropping macOS saves hub nothing today. The 10x figure matches the private-repo rates ($0.062 a minute for macOS against $0.006 for Linux 2-core). It applies to hub only if hub goes private, or to a private repo with the same matrix.
Coverage: hub runs only on macOS. The code behind cfg(target_os = "macos") (disclaimed_spawn/, config/src/op_process.rs and config/src/screen_look.rs) would no longer be built, linted or tested in CI. The pre-commit just lint on a Mac still covers it for local commits, but not for commits made on GitHub, such as Dependabot's.
The decision:031 says CI checks every push on both platforms, so Linux-only CI means amending it first.
Signal: with no macOS job, a runner shortage can't turn main red. That is what this issue asks for, but it gets there by losing the macOS result rather than by telling the two kinds of failure apart.
Options that keep some macOS coverage:
run macOS only on pushes to main, or on a schedule
make only the Linux job required, so a macOS job that never gets a runner doesn't fail the run
set a short queue timeout on the macOS job, and report it separately from build failures
Activity
ooloth commentedon Nov 5, 2026
Checked 2026-11-05: the
check (macos-latest)job has passed on every CI run onmainsince 2026-11-04 19:32Z (0a5fb30through5c7dec4, 7 runs). The last cancelled one was68cba87at 19:07Z. #401 has since changed.github/workflows/ci.yml, pinning actions and tools, but the job still runs onmacos-latest. Nothing in hub stops the failure returning the next time GitHub runs short of macOS runners.One workaround under consideration is running CI on Linux only, cited as cutting cost about 10x. Trade-offs to think through before choosing it:
cfg(target_os = "macos")(disclaimed_spawn/,config/src/op_process.rsandconfig/src/screen_look.rs) would no longer be built, linted or tested in CI. The pre-commitjust linton a Mac still covers it for local commits, but not for commits made on GitHub, such as Dependabot's.mainred. That is what this issue asks for, but it gets there by losing the macOS result rather than by telling the two kinds of failure apart.main, or on a schedule