drasi-project / drasi-core / 26309368977

71%

Build Type

push

github

Committed by web-flow

Commit Message

feat(reactions): reaction recovery persistence (#401-#404) (#459)

* feat: implement reaction recovery and bootstrap (#397, #398, #399, #400)

Add reaction recovery infrastructure with checkpoint-based state tracking,
outbox gap detection, and configurable recovery policies (Strict, AutoReset,
AutoSkipGap).

Key changes:
- BootstrapContext with trait-based BootstrapBackend for in-process and FFI use
- Snapshot and outbox fetch APIs on Query trait (fetch_snapshot, fetch_outbox)
- QueryOutputState with im::HashMap for O(1) snapshot cloning and ring-buffer outbox
- ReactionCheckpoint persistence via StateStoreProvider
- Startup validation rules (durable store, policy compatibility)
- Bootstrap gate pattern: buffer live events while bootstrap runs
- Per-query recovery: config hash mismatch detection, outbox catchup
- Lifecycle supervisor: monitors forwarder tasks, transitions to Error on exit
- FFI bootstrap bridge: FfiBootstrapContext, vtable bootstrap_fn, host callbacks
- Reaction trait extensions: bootstrap(), needs_snapshot_on_fresh_start(),
  default_recovery_policy(), is_durable()
- End-to-end bootstrap test with snapshot consumption verification

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix: address 14 review issues in reaction recovery system

Bootstrap gate & ordering (Critical):
- Replace Notify gate with watch<bool> channel to prevent missed wakeups
- Reorder subscribe_and_bootstrap: wire subscriptions BEFORE bootstrap
  so events buffer in broadcast channel during bootstrap
- Add sequence filtering in forwarders to skip stale events

Forwarder lifecycle (Critical):
- Store AbortHandles for both forwarder and supervisor tasks so
  stop_reaction properly cancels everything
- Supervisor checks current status before transitioning to Error,
  preventing override of intentional stop_reaction

FFI safety (High):
- Fix Arc leak on bootstrap thread panic by reclaiming before error
  propagation
- Parse earliest_available ... (continued)

Coverage Stats

541 of 779 new or added lines in 13 files covered. (69.45%)

1 existing line in 1 file now uncovered.

78648 of 110444 relevant lines covered (71.21%)

41.71 hits per line