12.33. Release 0.203
- Fix spurious duplicate key errors from .
- Fix out of memory error caused by missing pushback checks in data exchanges.
- Fix execution failure for queries containing a cross join when using bucketed execution.
- Fix execution failure for queries containing an aggregation functionwith
DISTINCT
and a highly selective aggregation filter.For example: - Fix quoting in error message for
SHOW PARTITIONS
. - Eliminate redundant calls to check column access permissions.
- Improve query creation reliability by delaying query start until the clientacknowledges the query ID by fetching the first response link. This eliminatestimeouts during the initial request for queries that take a long time to analyze.
- Remove support for legacy
ORDER BY
semantics.
Security Changes
- Fix sending authentication challenge when at least two of the
KERBEROS
, , orJWT
authentication types are configured. - Allow using PEM encoded (PKCS #8) keystore and truststore with the HTTP serverand the HTTP client used for internal communication. This was already supportedfor the CLI and JDBC driver.
- Declare a dependency on
uuidgen
. Theuuidgen
program is required duringinstallation of the Presto server RPM package and lack of it resulted in an invalidconfig file being generated during installation.
Hive Connector Changes
- Fix complex type handling in the optimized Parquet reader. Previously, null values,optional fields, and Parquet backward compatibility rules were not handled correctly.
- Fix an issue that could cause the optimized ORC writer to fail with a error.
- Improve error message for max open writers.
- Fix retry of requests when the remote Thrift server indicates that theerror is retryable.