[teamleaderleo]

Base-red-job: cmux-actions / Tests (expressions, matrix, every workflow fixture)

#32457 (bbb2b89) moved the iOS kit picker's fallback in .github/workflows/cmux-next-ios-kit.yml to fromJSON(vars.RUNNER_POOLS).linux_8. That pool is declared in .github/runners.json and set live, so the workflow is right. The cmux-actions fixture, though, had RUNNER_POOLS = {"linux":"fleet-linux"} only, so the picker's runs-on came out null and 4 "iOS kit picker fallback" cases failed on feat-cmux-next.

Fix, test-only (the workflow is unchanged and nothing is pinned):

• The fixture RUNNER_POOLS now holds every pool declared in .github/runners.json at its first member, which is what scripts/ci/runner_pools.py keeps it at. linux stays fleet-linux, so trigger planning still sees a fleet candidate.
• The iOS kit cases now test the pool reads: the eight vCPU pool, a moved pool being followed, the bounded LINUX_RUNNER 4/8/16 sizes, a fleet or invalid LINUX_RUNNER being rejected for the pool, and saved overflow reading linux_16. I dropped the unset and empty variable cases: RUNNER_POOLS is required, and an unset one fails the run at fromJSON() (tests/runs_on_labels.py).

Checks: vitest run test/ passes 1032/1032 (4 failed before), tsc --noEmit passes, and tests/test_ci_no_single_size_pins.py is OK.

Agent session: agent session

🤖 Generated with Claude Code

---

^{Open in [code]smith · Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Fixes the cmux-actions fixture so the iOS kit picker fallback tests pass on feat-cmux-next.

#32457 moved the picker's fallback to fromJSON(vars.RUNNER_POOLS).linux_8, but the fixture variable held only linux, so the picker resolved to a null runs-on and four cases failed. The fixture now holds every pool declared in .github/runners.json at its first member — what scripts/ci/runner_pools.py keeps them at — and the iOS kit cases test the pool reads: the eight vCPU pool, a moved pool, the bounded LINUX_RUNNER 4/8/16 sizes, and saved overflow reading linux_16. The unset and empty variable cases are gone; RUNNER_POOLS is required and an unset one fails the run at fromJSON().

Test-only: the workflow is unchanged and nothing is pinned.

^{Written for commit fec6edd. Summary will update on new commits.}

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review fec6edd: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr32653 -f ref=lane/fleet-ios-kit-runs-on -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4154676661101 (e467c2433a)
• cmux NEXT nightly 1.0.0-nightly.4154787783901 (9c772fa667)
• cmux NEXT nightly 1.0.0-nightly.4155084601701 (1923ed1867)

[teamleaderleo]

Review (subagent claude-opus-5-5 medium): SHIP at fec6edd

I checked that the fixture pools derive from .github/runners.json, that the cases match the ios-kit runner runs-on order, and that dropping the unset/empty cases is justified by runs_on_labels.pool_members. vitest passes 1032/1032 and the no-single-size-pins guard is OK.

Follow-ups: none.

[teamleaderleo]

Review (codex gpt-6-astra high, sweep, via the review router; full diff from merge base db16812 to this head): SHIP at fec6edd

No P1/P2 findings.