v3.4.3

Added

• Added EnableThriftNativeMetadata to request and consume supported Thrift-native SEA metadata results.
• Real-Time (Reyden) SQL warehouses now work over the default Thrift connection with no configuration change: when the gateway rejects a Thrift OpenSession with SQLSTATE KP001, the driver transparently re-opens the session on SEA and remembers the warehouse (per host, ~6h) so later connections skip Thrift. An explicit UseThriftClient=1 is always honored and never auto-switched.
• Added session-version tracking from synchronous SQL Exec API responses for Lakehouse Real-Time. With asynchronous execution, subsequent statements may not observe session changes.

Updated

• UseBoundedSeaApi and EnableThriftNativeMetadata now default to 1; when unset, activation is controlled by the server-side enableSqlExecForJdbc rollout flag.
• DatabaseMetaData.getColumns(...) with a null catalog now issues a single SHOW COLUMNS IN ALL CATALOGS statement (consistent with getSchemas/getTables) instead of enumerating every catalog and issuing a per-catalog SHOW COLUMNS. Older DBR versions that do not support the syntax transparently fall back to the previous enumerate-and-fan-out behavior.
• Updated bundled Jackson, lz4-java, Netty, and Apache HttpComponents Client and Core dependencies to patched versions to address security findings.
• Updated bundled Apache Thrift to 0.24.0 to address CVE-2026-43871.

Fixed

• Fixed DatabaseMetaData.getTypeInfo() returning the INTERVAL row out of DATA_TYPE order.
• Fixed later logging-enabled connections being unable to produce logs when an earlier connection used LogLevel=OFF. The first enabled connection now establishes the shared JUL handler, while a later OFF connection does not disable it.
• Invalid or incomplete Databricks JDBC URLs now fail with a descriptive DatabricksSQLException instead of leaking a NullPointerException when required connection parameters are missing.
• Fixed connections failing when the same parameter is provided in both the JDBC URL and the connection properties, with the JDBC URL taking precedence.
• Fixed IdleConnectionEvictor thread leak in long-running applications. Driver-side resources (HTTP client, background threads) are now always released when Connection.close() is called, even if statement cleanup or server-side session termination fails.
• Throw DatabricksSQLException instead of an unchecked ClassCastException when a complex-type getter (getArray, getStruct, getMap) is called on a column of a different complex type.
• Fixed NullPointerException when reading collated string columns (e.g. STRING COLLATE UTF8_LCASE) over Arrow. Such columns report a type_name that does not map to a ColumnInfoTypeName, leaving it null; the value read now recovers STRING from the type text and the result set metadata reports VARCHAR instead of OTHER, while getColumnTypeName() still preserves the collated type text.
• Fixed ResultSet.getObject(int) on the Arrow result path leaking a raw java.lang.IndexOutOfBoundsException (with a null SQLState) for an out-of-range column index. It now throws a DatabricksSQLException (SQLState INVALID_STATE, "Column index out of bounds: <n>"), matching the JDBC contract and the Thrift/inline result implementations. Affects the Arrow/CloudFetch path used by SEA and by Thrift CloudFetch results.
• Fixed connecting with an unsupported AuthMech (e.g. AuthMech=99) intermittently failing with an internal IllegalStateException: Recursive update or StackOverflowError on both the SEA and Thrift paths. The value is now validated at connect time and rejected deterministically with a SQLException (SQLState=INPUT_VALIDATION_ERROR).
• Improved SEA connection-failure error messages.
• Fixed NullPointerException being thrown when materializing an array containing nested object types (other arrays, structs or maps) as DatabricksArray when some or all elements are literal null.