The gossip listener used synchronous newSocket + recvFrom,
which blocks the async event loop and prevents other async
operations from running while waiting for UDP packets.
Switch to newAsyncSocket + async recvFrom so the loop can
continue processing other tasks between gossip rounds.
The previous implementation allowed multiple concurrent async calls
(query, execute, ping) to interleave writes on the same socket,
which corrupted the binary protocol framing when NodeBB fired
parallel database operations.
Add an internal _requestQueue and _requestLock so that all TCP
requests are serialized: each async operation enqueues a task,
and tasks are drained one at a time via setImmediate().
The JavaScript client reads floats via readFloatBE/readDoubleBE,
which expect IEEE-754 values in big-endian byte order.
The Nim server was writing them with copyMem in native byte order,
so on little-endian machines (x86_64) the JS side deserialized
FLOAT64 values as garbage (e.g. 1.0 became 3.03865e-319).
Fix serialization by casting to int32/int64 and using bigEndian32/
bigEndian64, mirroring the existing big-endian handling for ints.
Apply the same fix to deserialization and to fkVector elements.
- Added missing irNeg case in evalExpr (query/executor.nim)
- Unary minus previously fell through to else: return 'false'
- Added 2 tests: SELECT expression and WHERE condition
- disttxn: rollback now correctly updates participant state (was modifying a copy)
- disttxn: bare except replaced with CatchableError
- sharding: migrateData condition was always false, now correctly migrates data
- sharding: unsafe cast[string] replaced with manual byte-to-char conversion
- sharding: bare except replaced with CatchableError
- server: parseUInt replaced with parseBiggestUint for uint64 portability
Commit 6aaabb5 added std/locks usage in vector/engine.nim (HNSW thread-safety).
std/locks requires --threads:on. Local builds worked because /etc/nim/nim.cfg
had threads:on, but GitHub Actions runners (setup-nim-action) compile without
threads by default, causing the test job to fail silently at runtime.
Also updated baradadb.nimble tasks for consistency.
Fixes GitHub Actions CI failure on test / clients-ci workflows.
- backup.tla: models backup/restore/verify/cleanup with properties:
BackupSnapshotsValid, RestoreIntegrity, VerifyIntegrity,
RetentionInvariant, HistoryConsistency, BackupIdValid
- recovery.tla: models WAL replay REDO/UNDO with properties:
RedoCommitted, RecoveryCompleteness, WalIntegrity
- Added models/backup.cfg and models/recovery.cfg with bounded constants
- Updated run_all.sh to include both new specs
- All 9 TLA+ specs pass TLC model checking
Поправени проблеми:
- LSM-Tree: WAL write вече е извън db.lock (отделен walLock) → по-добра concurrency
- MVCC: добавен compactVersions() който се изпълнява на всеки 100 commits
Премахва стари overwritten версии, които не са видими за active транзакции
- DistTxn: commit() вече НЕ rollback-ва commit-нали participants
Ако някой participant е commit-нал, транзакцията се маркира като committed
- Raft: добавен disk persistence за currentTerm, votedFor, log
saveState() при всяка промяна; loadState() при стартиране
292 теста, 0 failure-а.
Поправени проблеми:
- Deadlock detector: cycle reconstruction вече проверява пълен цикъл
- Graph cypher: executeCypher acquire-ва lock при четене
- Graph engine: loadFromFile acquire-ва lock при зареждане
- Graph engine: lock поле е exported за външен достъп
- Query: parseRowData поддържа escape-ване на , и = в стойности
- Query: execUpdateRow escape-ва стойности при сериализация
- Lexer: readNumber спира при втора точка (1.2.3 вече не е валиден float)
- Lexer: skipBlockComment хвърля ValueError при незатворен коментар
292 теста, 0 failure-а.
dimgigov