Trevon
|
090a5ca69c
|
fix(testing): stabilize TSAN coverage, harden daemon shutdown, and speed up KG/storage paths (#53)
* perf(search, metadata): add KG write buffer, score cache, adaptive budgets
* perf(indexing): wire KGWriteBuffer into indexing service
Converts per-document addEdgesUnique calls into batched flush via
KGWriteBuffer. 500-doc ingest: 500 SQL round-trips → 1 batch.
Verified: 14.2× edge insert gap (14.8K→195K edges/s) in kg_edge bench.
Tests: 10/10 Catch2 pass (105 assertions).
* perf(storage): add fsync durability, parallel batch, read integrity
* test(storage): add stress, crash-sim, cold-cache, batch benchmarks
* fix: CI optimizations and work coordinator shutdown fix
* fix(testing): stabilize TSAN coverage gate
* fix: harden KG flush, score cache, and benches
- keep KG buffers intact when batch commit fails
- count flushed edges per batch instead of lifetime totals
- guard the Simeon score cache and clear it on rebuild
- fall back to fsync when F_FULLFSYNC fails on macOS
- remove interactive sudo from cold-cache benchmarks
|
2026-06-26 17:57:02 -04:00 |
|
Trevon
|
b603738f44
|
Stabilize sanitizer lanes and daemon concurrency (#40)
* fix(repair): keep RPC progress alive
* fix(daemon): propagate DB open failure to FSM and add shutdown cancellation
* fix(db): classify FTS5 inverted-index errors as repairable, not transient locks
* fix(daemon): clean up stale corrupt files after salvage and suppress optional readiness keys
* test(db_recovery): update FTS5 classification test for new transient-lock exclusion
* test(gc): add garbage collector unit tests for collection, dry-run, scheduling
* test(repair): expand repair manager tests with hash-verified recovery and source ordering
* fix(gc): dry-run mode no longer increments blocksDeleted/bytesReclaimed
* fix(mock,mcp): add atomic bool for lock consistency, move field assignments
* style: fix unnecessary-copy, dead-store, and const-refable findings from Infer
* fix(daemon): increase default shutdown timeout from 15s to 120s
* fix(daemon): demote non-critical warnings to info (FTS5, VACUUM, WorkCoordinator timeout)
* fix(db): re-enable WAL auto-checkpoint to prevent unbounded WAL growth
* test(pool): add WAL auto-checkpoint and concurrency coverage for connection pool
* test(pool): add exhaustion, stats, lifecycle, and read-only connection pool tests
* test(pool): add tagging, WAL journal, reuse, and busy timeout tests
* test(metrics): relax session snippet assertion to tolerate stub behavior
* fix(test): stabilize sanitizer triage failures
* fix(build): wire version and zyp deps
* fix(daemon): close TSAN race clusters
* fix(test): guard optional zyp dep
* fix(test): address PR review findings
|
2026-06-01 12:14:43 -04:00 |
|