The GitHub Actions job "Flink Connector JDBC CI Backwards Compatibility" on 
flink-connector-jdbc.git/FLINK-40392 has failed.
Run started by GitHub user patrashu (triggered by patrashu).

Head commit for run:
5e87ff3001a431f7648122f991532e6f32e410db / hmc7493513 <[email protected]>
[FLINK-40392][jdbc] Add BoundaryQuerySplitterEnumerator for 
scan.partition.boundary-query

Introduces AbstractDynamicSplitterEnumerator, a base class for
SplitterEnumerator implementations that must open a database
connection at job runtime to discover their splits, and
BoundaryQuerySplitterEnumerator, its first implementation - it derives
split boundaries by running a user-provided SQL query instead of a
statically-known numeric range, per FLIP-602.

Addresses review feedback: rethrow on discovery failure so the job
fails instead of retrying forever, drop the AS before the boundary
subquery alias (Oracle rejects it), read temporal boundary columns via
getObject(int, Class) instead of a driver-specific type, delay the
finished flag by one call to avoid a checkpoint losing discovered
splits, and document that baseSqlTemplate must not already contain a
WHERE/LIMIT clause. Also fixes a NotSerializableException the
discovery lock introduced, found while re-verifying these fixes
against Flink 2.1.3's coordinator-thread model.

Wiring these into scan.partition.boundary-query as a public table/
DataStream API option is left to a follow-up PR/JIRA (ST-5).

Generated-by: Claude Code (Claude Sonnet 5)

Report URL: 
https://github.com/apache/flink-connector-jdbc/actions/runs/36640051353

With regards,
GitHub Actions via GitBox

Reply via email to