CodeWithPravinMaske commented on PR #12608: URL: https://github.com/apache/seatunnel/pull/12608#issuecomment-6010307115
Thanks @SEZ9, all three are in: 1. **Restore E2E**: `AbstractMysqlCDCITBase#testMysqlCdcNullInNullableUniqueKeyAfterRestore` (Zeta only, like the other restore tests). It snapshots the tables, runs `ALTER TABLE uk_added_null ADD UNIQUE KEY` and NULL inserts in the binlog phase, takes a savepoint, restores the job, inserts another NULL row into each table and asserts the sink holds NULL (and no `0`). After the restore, `uk_added_null`'s schema comes from the table history saved in the savepoint, since it changed in the binlog phase. The binlog insert step is shared with `testMysqlCdcNullInNullableUniqueKeyWithoutPrimaryKey`. It passes locally. 2. **Comments on the override**: the class Javadoc says it is a copy of Debezium 1.9.8, that `parsePrimaryIndexColumnNames` is the only modified method, and that the change must be re-applied (or the copy dropped) when upgrading Debezium. The method Javadoc explains why: Debezium also uses it to promote a unique key, which can hold NULLs, so only a real primary key marks its columns NOT NULL. 3. **F4 assertion**: `MySqlSchemaTest#testNullableColumnConfiguredAsPrimaryKeyEmitsNull` builds the catalog table through `CatalogTableUtils.mergeCatalogTableConfig(catalogTable, config)` with `primaryKeys = ["code"]` (the real `table-names-config` path, which marks the catalog column NOT NULL). It then checks that the Debezium column stays optional, and that converting a row with NULL in `code` through the Debezium table schema yields NULL, not `0`. With Debezium's original behaviour restored, this test fails, as do the 3 other nullability tests. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
