[teamleaderleo]

On owned Mac minis, the cmux-tui macOS job (macOS lint and tests) now runs setup-go without the Actions cache. It also saves the cargo cache only from feat-cmux-next. Blacksmith VMs start empty, so they keep both caches as before.

Why (measured)

I sampled the 8 macOS lint and tests jobs on owned minis from 2026-10-29 14:30Z to 22:30Z, after setup-go landed in 5889445. The raw step timings and log excerpts are saved in the lane's scratch dir.

• setup-go restored 0 of 8 times. In 6 jobs it downloaded the ~590 MB entry (about 63 s mean, up to 128 s) and then failed to untar it. The mini keeps its own ~/go/pkg/mod, which Go makes read-only, so tar fails with Can't unlink already-existing object: Permission denied and then Failed to restore. Each job then tried to save the entry again: 48 s mean, ending in Unable to reserve cache ... another job may be creating this cache.
• rust-cache restored 0 of 8 times (No cache found.). Its key carries a hash of each mini's Rust environment, so minis produce different keys (ae76135b, b2523b1e, baf93c3c, ...). A branch dispatch can only read that branch's entries and the default branch's. Each branch save still took about 155 s (median 175 s).
• These saves fill the repository's GitHub cache store. At 2026-10-30 01:30Z it held 9.79 GB of its 10 GB: 6.80 GB under v0-rust-cmux-tui-macos-* and 2.87 GB under setup-go-macOS-*. 7.69 GB of that was on branch refs that no other run reads. The store evicts by least-recent use, so these entries push out the shared ones, such as ghostty-zig-packages. Only runners outside Blacksmith write this store; Blacksmith runners use Blacksmith's own store.

Changes

• .github/workflows/cmux-tui.yml (macos job):
  • setup-go sets cache: ${{ !contains(runner.name, '-glaeda') }}.
  • rust-cache sets save-if: ${{ !contains(runner.name, '-glaeda') || github.ref == 'refs/heads/feat-cmux-next' }}.
  • contains(runner.name, '-glaeda') is the repo's existing test for an owned mini; 12 other steps use it.
• tests/test_cmux_tui_owned_mini_caches.py pins both conditions. It is registered in tests/test-execution.toml (linux-guard) and run from ci-guards.yml.

Performance

Per owned-mini macOS lint and tests job. "Before" is the 8-job sample from 2026-10-29, at base 5889445 and later. "After" is a dispatch of this head on this branch.

metric	before	after	delta	harness or run link
setup-go step	62.5 s mean (0/8 restored)	AFTER_GO	AFTER_GO_DELTA	before: jobs 123671398605, 123684093131, 123664698315, 123684118038, 123732090143, 123772035002, 123660185485, 123653593592; after: AFTER_RUN
post setup-go (save attempt)	47.8 s mean	AFTER_POSTGO	AFTER_POSTGO_DELTA	same
post rust-cache on a branch (save)	155.5 s mean (0/8 restored)	AFTER_POSTRUST	AFTER_POSTRUST_DELTA	same
GitHub cache store written per branch dispatch	0.6-0.9 GB rust + 0.6 GB go	0 GB	-1.2 to -1.5 GB	/actions/caches listing, 2026-10-30 01:30Z

Restore hit rate on owned minis is 0% before and after. The change removes time and store space spent on restores that never succeed; it does not add hits. feat-cmux-next still saves the cargo cache that branches read.

Risk

This is CI only, and touches only the macOS job of cmux-tui.yml. If an owned mini lacks Go's module cache, go build downloads the modules, as the fleet step cmux-tui-rust-check.sh already does. Blacksmith macOS runs keep both caches unchanged.

Verification

• python3 tests/test_cmux_tui_owned_mini_caches.py: OK.
• python3 tests/test_ci_rust_cache_keeps_runner_bin.py: ok (25 steps).
• bash scripts/ci/tests/cmux-tui-rust-check.test.sh: ok.
• actionlint on both changed workflows: clean.
• tests/test_ci_test_execution_registry.py has 2 failures that fail the same way on base: test_verify_local_jobs_option_keeps_recipe_tests_live and test_cli_product_lane_selects_the_hook_spool_regression.

Part of cmuxterm-hq#5230 row "Blacksmith Actions cache ... cache inspection". The report half of that row is a separate PR into main, which adds an Actions cache section to the CI health report.

risk: normal (CI only)

🤖 Generated with Claude Code

Agent session: agent session

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Owned Mac minis in the cmux-tui macOS job now skip caches that never restore on them, saving job time and space in the repository's GitHub cache store.

• On an owned mini, setup-go's cache failed to restore in 8 of 8 sampled jobs: the mini keeps its own read-only ~/go/pkg/mod, which tar can't overwrite, so each job still downloaded the ~590 MB entry (~63 s) and re-saved it (~48 s).
• The cargo cache key hashes each mini's Rust environment, so branch dispatches never read those entries; each save took ~155 s and filled ~70% of the 10 GB store, evicting shared caches.
• setup-go now sets cache: ${{ !contains(runner.name, '-glaeda') }}; rust-cache saves only from feat-cmux-next. Blacksmith VMs start empty and keep both caches.
• Adds tests/test_cmux_tui_owned_mini_caches.py, registered in tests/test-execution.toml under the linux-guard lane, to pin both conditions.

^{Written for commit 9d9e4d6. 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 9d9e4d6: 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=pr27840 -f ref=ci/cmux-tui-owned-mini-caches -f platform=macos -f backend_mode=local

red: cmux-next checks (god files, concurrency, crash safety, l10n); NEW: 3 commit(s) change the sidebar's drawing code without a Sidebar-Look: trailer or an entry in plans/cmux-next/sidebar-owner.md is not red on the base 1d98fce; do not merge until green; see lane-rules

[blacksmith-sh]

Found 9 test failures on Blacksmith runners:

Failures

Test	View Logs
1 runner guard check(s) failed:/1 runner guard check(s) failed:	View Logs
1 runner guard check(s) failed:/1 runner guard check(s) failed:	View Logs
test_big_red (main.FleetRuleLints.test_big_red)/ test_big_red (main.FleetRuleLints.test_big_red)	View Logs
test_every_upload_is_short_lived_and_small_or_allowlisted (main.WorkflowTests.test_ every_upload_is_short_lived_and_small_or_allowlisted)/ test_every_upload_is_short_lived_and_small_or_allowlisted (main.WorkflowTests.test_ every_upload_is_short_lived_and_small_or_allowlisted)	View Logs
test_fork_pull_requests_into_manaflow_ai_never_reach_an_owned_runner_label (main.Fo rkRunnerRoutingTests.test_fork_pull_requests_into_manaflow_ai_never_reach_an_owned_runn er_label)/ test_fork_pull_requests_into_manaflow_ai_never_reach_an_owned_runner_label (main.Fo rkRunnerRoutingTests.test_fork_pull_requests_into_manaflow_ai_never_reach_an_owned_runn er_label)	View Logs
test_nightly_keeps_a_margin_under_the_limit (main.WorkflowFileSizeTest.test_nightly _keeps_a_margin_under_the_limit)/ test_nightly_keeps_a_margin_under_the_limit (main.WorkflowFileSizeTest.test_nightly _keeps_a_margin_under_the_limit)	View Logs
test_the_allowlist_has_reasons_and_no_stale_entries (main.WorkflowTests.test_the_al lowlist_has_reasons_and_no_stale_entries)/ test_the_allowlist_has_reasons_and_no_stale_entries (main.WorkflowTests.test_the_al lowlist_has_reasons_and_no_stale_entries)	View Logs
workflow references a self-hosted mac fleet label or bare self-hosted runner in a runne r-selection position./ workflow references a self-hosted mac fleet label or bare self-hosted runner in a runne r-selection position.	View Logs
workflow references a self-hosted mac fleet label or bare self-hosted runner in a runne r-selection position./ workflow references a self-hosted mac fleet label or bare self-hosted runner in a runne r-selection position.	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[teamleaderleo]

Review (codex gpt-6-astra high, next-review): SHIP at 6a03b09

No blocking findings.

[cmux-bot]

cmux-bot approves 6a03b09: an independent review passed this exact head (next-review receipt by teamleaderleo). A push needs a new review; a later BLOCK at this head dismisses this approval.

[teamleaderleo]

Owned by an agent session · session

A newer verdict blocks a386692 (no independent SHIP at the head).

[github-actions]

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4104504184001 (afe5e8ddc4)
• cmux NEXT nightly 1.0.0-nightly.4104787172001 (5ec8097e09)
• cmux NEXT nightly 1.0.0-nightly.4104925832401 (7f97b8485c)