fix(coordinator): post-audit logic-review hotfixes

[appergb]

User description

Summary

Two real bugs surfaced by the 2026-05-10 end-to-end logic review of beta after the 8-PR audit-fix wave landed (full report at `docs/logic-review-2026-05-10.md`, kept local).

Bug 1: `coordinator.rs:2313` — QA path missing `.await`

PR #391 (`6171df61`) made `acquire_recording_mute` `async fn` and updated the dictation call site (`coordinator/dictation.rs:451`), but the QA call site was missed.

Effect: when `mute_during_recording = true` and the user triggers QA, the returned Future was dropped on the next line, `spawn_blocking` was never scheduled, `holders` never incremented, system audio was NOT muted (YouTube/etc. kept playing into the QA recording). Both matching `release_recording_mute(inner, "qa")` calls became no-ops (early return at `holders == 0`). The Rust compiler was emitting `unused_must_use` for this exact line.

Fix: `+ .await`. Compiler warning gone.

Bug 2: `coordinator/dictation.rs:843-849` — `focus_target` not cleared on cancel during Processing

PR #387 (`ce82fcd9`) was framed as "clear focus_target on cancel regardless of phase", but the unconditional clear only landed in `finish_cancel_session_state`. `cancel_session` deliberately skips that helper for the `Processing` branch (so `end_session` drives its own teardown), and `end_session`'s "ASR-finished, cancelled" exit set `phase = Idle` but never touched `focus_target`.

Effect: cancel-during-Processing leaves a stale `usize` focus slot in `state.focus_target` until the next `begin_session_state` overwrites it. Today bounded (no documented reader on that interval), but violates PR #387's stated contract and is a silent footgun for future readers.

Fix: in the cancelled-after-ASR exit branch, take `state.lock()` once and clear both `phase` and `focus_target` together.

Test plan

• `cargo test --lib` 185/185 pass.
• `cargo check` clean. `unused_must_use` warning at coordinator.rs:2313 disappears.
• CI Linux/Windows/macOS to verify.
• Manual for Bug 1: enable mute_during_recording, play YouTube, open QA panel, press Right Option — audio should mute and `[audio-mute] acquired by qa; holders=1` appear in `openless.log`.
• Manual for Bug 2: start dictation → release → during ~1s ASR await window press Esc; inspect logs / state. `focus_target` should be `None` immediately after cancel-during-Processing (was: stale until next session).

---

PR Type

Bug fix

---

Description

• Add missing .await on acquire_recording_mute in QA path

• Clear focus_target on Processing‑phase cancel in end_session

---

Diagram Walkthrough

flowchart LR
  QA["begin_qa_session"] -- ".await" --> Mute["acquire_recording_mute('qa') (mute engages)"]
  Cancel["end_session (cancel after ASR)"] -- "clear focus_target" --> Focus["state.focus_target = None (no stale handle)"]

Loading

File Walkthrough

Relevant files

Bug fix

coordinator.rs Await QA audio mute call                                                                  openless-all/app/src-tauri/src/coordinator.rs Add .await to acquire_recording_mute(inner, "qa") to prevent dropped future +1/-1      dictation.rs Clear focus_target on Processing cancel                                    openless-all/app/src-tauri/src/coordinator/dictation.rs Clear focus_target when cancelling during Processing phase to prevent stale handle Use single lock acquisition to set both phase and focus_target +9/-1

---

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis 🔶 387 - Partially compliant Compliant requirements: Clear focus_target on accepted cancels, including Processing. Keep the Processing -> Idle transition in end_session. Non-compliant requirements: Regression test updates are not visible in the provided diff. Requires further human verification: Run the updated coordinator state-machine tests. Manually verify cancel-during-Processing behavior in a real build. 391 - Partially compliant Compliant requirements: Await acquire_recording_mute() in the QA path. Non-compliant requirements: The non-blocking implementation changes for the broader audio-mute flow are not visible in the provided diff. Requires further human verification: Confirm the associated spawn_blocking changes in the audio-mute implementation. Run the macOS timing/manual verification described in the ticket. Verify the CI build.

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ No major issues detected