[lawrencecchen]

Two acpmux lint base reds on feat-cmux-next.

1. macOS clippy: function start_logged is never used

cargo clippy -D warnings on macOS (job fb8745cdbcf3231af7557c1d, aws-m4pro-4, at #28041's merged head) fails at cmux-tui/crates/acpmux/tests/daemon_memory.rs:82. #27991 (941b7a4) moved Daemon::start onto start_inner, so start_logged has one caller left, retained_session_list_allocation_benchmark, which is #[cfg(target_os = "linux")]. Fix: start_logged is #[cfg(target_os = "linux")] too.

2. Panic ratchet: acpmux arithmetic_side_effects over its baseline 1430

The tracked count grew from the baseline that #28169 (66c3d03) measured (warning now). From the ratchet's clippy JSON at 66c3d03 (cmux-tui-spec run 41076340459) and at a current feat head (run 41086018852), per file:

merge	file	new hits	now
4543835 (#28161)	hub/session_pages.rs	3 (len() + 1, used + bytes twice)	saturating_add
55df640 (#28353)	daemon/unused_exit.rs	2 (clients += 1, arrivals += 1)	saturating_add

7a5538d (#27868) removed one in agent_host/host.rs. No value here can reach usize::MAX or u64::MAX, so behavior is the same. The baseline is not lowered here: a lower count needs the Linux measurement of this PR's own ratchet run.

Evidence

	run	result
macOS clippy before (feat tip 4ab956c)	fleet 5937adaa407e3121bf249dd6 (aws-m4pro-2)	fails: associated function start_logged is never used
macOS clippy after (c73f1cc, the dead-code fix alone)	fleet b62faf44852ef09f28767510 (aws-m4pro-2)	passes
Linux clippy after (c73f1cc)	fleet bf122b19e08a2d0c4e790703	passes (Linux never had the warning)
ratchet after (head)	this PR's cmux-tui-spec run	pending

R9: no new cap, limit or timeout.
R10: no hot path changes.

Observed / Not run

• Observed: the runs above; prepush-gate.py passes.
• Not run: the ratchet at the head (it runs in this PR's checks).

Changelog

• none

No capture: no UI change.

[github-actions]

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

Review 1f73eb4: 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=pr28418 -f ref=base-red/daemon-memory-macos-dead-code -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4115130212201 (0e12c602f8)
• cmux NEXT nightly 1.0.0-nightly.4115488208801 (dd66817742)
• cmux NEXT nightly 1.0.0-nightly.4115755828101 (7c0adaf184)

[blacksmith-sh]

Found 2 test failures on Blacksmith runners:

Failures

Test	View Logs
test_every_listed_hook_is_one_glaeda_released_with_the_check (main.SigningHooksInSt epWithGlaedaTests.test_every_listed_hook_is_one_glaeda_released_with_the_check)/ test_every_listed_hook_is_one_glaeda_released_with_the_check (main.SigningHooksInSt epWithGlaedaTests.test_every_listed_hook_is_one_glaeda_released_with_the_check)	View Logs
test_the_repository_highlights_validate (main.DigestTests.test_the_repository_highl ights_validate)/ test_the_repository_highlights_validate (main.DigestTests.test_the_repository_highl ights_validate)	View Logs

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

[lawrencecchen]

Review (independent review subagent, Claude Opus 5.5): SHIP at 6825a75
reviewer-session: rbasered-7b4fbe43
author-session: a1c85faf51fad876b
R9: pass. No new cap, queue, limit or deadline.
perf claim: none

Scope: PR #28418, git diff origin/feat-cmux-next...6825a752c11 (daemon/unused_exit.rs, hub/session_pages.rs, tests/daemon_memory.rs). I read the code, the culprit commits and the fleet logs. I did not compile or run tests.

Findings

1. Required before landing (merge, not a code defect). GitHub says this head is CONFLICTING. acpmux: a session list too large for one message goes in pages #28182 (49c648c) changed the same lines in hub/session_pages.rs byte_page: the base now has let bytes = size(&value) + 1; and still has used + bytes two times. Merge the base with scripts/merge-main.sh and keep the saturating form with the new size: let bytes = size(&value).saturating_add(1); let total = used.saturating_add(bytes);. After the push, the merge is a new head, so re-add ready-to-land.

2. Pass (verified). start_logged gets #[cfg(target_os = "linux")]. Its only caller is at daemon_memory.rs:566, inside retained_session_list_allocation_benchmark, which is #[cfg(target_os = "linux")] (line 525). The macOS clippy red is associated function start_logged is never used in 5937adaa407e3121bf249dd6 (4ab956c, aws-m4pro-2). The same clippy passes at c73f1cc in b62faf44852ef09f28767510. This is the root-cause fix: the function and its caller now have the same cfg. No lint is allowed or skipped.

3. Pass (verified). Each arithmetic change keeps behavior:

   • unused_exit.rs arrive: clients (usize) counts open connections. It cannot reach usize::MAX (each needs an fd and memory). arrivals (u64) increases once per connection; it cannot reach u64::MAX. saturating_add is correct where overflow cannot happen. Drop already uses saturating_sub. The arrivals != seen checks (lines 114, 118, 126) do not change.
   • session_pages.rs byte_page: bytes is a JSON string length plus 1, and used is the sum of rows that fit budget. Both are bounded by memory. total is computed once, and the same value goes into the compare and into used, so the page cut does not change.
     The 5 hits came from 4543835 (acpmux: ACP session/list goes in pages when the whole list is too large to send #28161: len() + 1, used + bytes 2 times) and 55df640 (acpmux: a daemon nobody uses stops after an hour #28353: clients += 1, arrivals += 1). I confirmed both in the culprit diffs.

4. Pass (ratchet). This PR's clippy (changed cmux-tui packages) run (41087613416) passes the panic ratchet and says "counts went down in acpmux". The PR does not raise the baseline.

5. Low (non-blocking). The body's evidence table still says the ratchet run is "pending". It passed; update the row.

checks: (gh pr checks 28418) 70 pass, 85 skipping, 4 fail.

• cmux-next daemon tests (same-tree) (run 41090080784): 2 Swift tests in ClassicUpgradeSimulationTests.swift (:129, :184) fail on workspace order. This PR changes only acpmux Rust, so the red is not from this PR (a base finding or a flake).
• tooling-tests, nightly-guards / workflow-guard-tests / release-notary and nightly-guards / Guard status: base findings, the same on base red: acpmux: a cmux agent message is not an inbox item (agents_chat_tools, reply in place) #28416, base red: acpmux: the chat index cache and the unattended folder are not written under a lost home #28417 and base red: acpmux: the carried-turn cap test queues its agent message with the inbox off #28427.
• After the merge with the base, expect the acpmux panic ratchet to be red for indexing_slicing (691 > 686 at 49c648c, seen on base red: acpmux: the carried-turn cap test queues its agent message with the inbox off #28427). That is a base finding from the base tip, not from this PR.

[lawrencecchen]

Review (independent review subagent, Claude Opus 5.5): SHIP at 1f73eb4
reviewer-session: rbr3-407362a5
author-session: a1c85faf51fad876b
R9: pass. No new cap, queue, limit or deadline (saturating adds on existing counters and an existing byte budget).
perf claim: none

Scope: PR #28418. The last receipt was SHIP at 6825a75. The new head 1f73eb4 is a merge of feat-cmux-next 45d6f70 into 6825a75, with one resolved conflict in cmux-tui/crates/acpmux/src/hub/session_pages.rs. I checked the merge and the PR's patch after it.

Findings

None.

Checked and found fine

• The PR's own commits are unchanged: git range-diff of the PR range before and after the merge shows c73f1cce9c9 = c73f1cce9c9 and 6825a752c11 = 6825a752c11.
• The PR's patch against the new base (git diff 45d6f7093d1 1f73eb4d379) is the same 3 files and 9+/6- as before: daemon/unused_exit.rs (saturating clients/arrivals), hub/session_pages.rs, tests/daemon_memory.rs (#[cfg(target_os = "linux")] on start_logged). Nothing else changed.
• The conflict resolution (git show --cc) keeps both intents:
  • acpmux: a session list too large for one message goes in pages #28182 (49c648c) made byte_page generic (byte_page<R>, row: impl Fn(usize) -> R, size: impl Fn(&R) -> usize) and counted bytes with size(&value) + 1. The merge keeps size(&value), so acpmux: a session list too large for one message goes in pages #28182's behavior (the caller's own size function, not value.to_string().len()) is kept.
  • This PR's change (no overflow, panic ratchet) is kept: size(&value).saturating_add(1), total = used.saturating_add(bytes), the break test on total > budget, and (used, last) = (total, Some(hash)). used stays 0_usize.
  • Behavior is the same as base for every value that does not overflow (the page still takes at least one row, never splits a hash and stops at the first row past the budget). The other base change to this file in the range (Base red: acpmux panic ratchet again (indexing_slicing 703 > 693, string_slice 97 > 94): rewrite the new hits without panics #28386) is not in conflict and is unchanged.
• Fleet: macOS clippy a4a8b0a15237a472ecf7a898 fetched 1f73eb4 and ran cargo clippy --workspace --exclude cmux-holder-v3 --all-targets --locked -- -D warnings: clippy pass, cmux-tui-rust-check: green, exit 0. This is the check the start_logged part of the PR fixes on macOS.

checks

gh pr checks 28418: 75 pass, 82 skipping, 2 fail. clippy (changed cmux-tui packages), rustfmt, cmux-next crash ratchet, cmux-next checks (god files, concurrency, crash safety, l10n), cmux-next fast tier (cargo), tooling-tests, risk evidence, merge freeze, ci-status pass. The 2 fails are a base finding: nightly-guards / workflow-guard-tests / release-notary and its wrapper nightly-guards / Guard status, red on feat-cmux-next until #28441 lands (same on #28416, #28417, #28467).

[lawrencecchen]

Ready to land at 1f73eb4:

• Independent review: SHIP at this exact head (base red: acpmux: macOS clippy dead start_logged; arithmetic ratchet back under its baseline #28418 (comment)).
• Risk tier: risk:normal.
• Red before: macOS clippy 5937adaa407e3121bf249dd6 (start_logged is never used).
• Green after: macOS clippy a4a8b0a15237a472ecf7a898 at this merged head.
• Base merged; session_pages.rs conflict resolved as size(&value).saturating_add(1). Takes back nothing.
• No push after the review.