All reported problems originated from clippy; flutter analyze was already
clean. The dominant cluster (44 errors) was in chronogrid_core::sync, whose
test module targeted a CRDT API that no longer existed - meaning that test
target had not compiled since Task 2.6.
Lint fixes:
- core/sync: realign tests to real TaskCrdt/NoteCrdt API (constructor arity,
set_*(value, hlc) arg order, from_bytes vs load, Result<Option<String>>
getters, and &mut receivers for Automerge save())
- core/crdt/doc: add impl Default for CrdtDoc; avoid approx_constant in test
- daemon: derive Default for Config; wire [logging].level into the tracing
subscriber so the previously dead fields are functionally used
- crypto/keychain: use div_ceil
- net/swarm_key: drop needless borrow
- net/swarm_manager: collapse if-let into match arm
- ai: replace assert!(true) with a real VERSION assertion
Bug fixes exposed once the crate compiled:
- core/sync: SyncEngine::run moves a SwarmHandle (holding a broadcast::Sender
clone) into the task, so RecvError::Closed is unreachable from an external
event_tx drop. The old test asserted an impossible contract and hung; it is
now torn down via JoinHandle::abort() with a new regression guard.
- ffi: serialize the 21 tests mutating the global ENGINE singleton behind a
poison-tolerant engine_test_guard(), fixing an intermittent parallel-test
race that was failing verify_agent_work.ps1.
Verification: clippy --workspace --all-targets --all-features = 0 warnings and
0 errors; cargo test --workspace --all-features = 221 passed / 0 failed;
flutter analyze clean; verify_agent_work.ps1 ALL CHECKS PASSED.
Mike Kell