[hibernate-issues] [Hibernate-JIRA] Commented: (HHH-6744) TransientObjectException thrown on dirty check during flush with CHECK_NULLABILITY set to false

TransientObjectException thrown on dirty check during flush with CHECK_NULLABILITY set to false ----------------------------------------------------------------------------------------------- Key: HHH-6744 URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-6744 Project: Hibernate Core Issue Type: Bug Components: core Affects Versions: 4.0.0.CR4 Reporter: Gail Badner Fix For: 4.0.0.next TransientObjectException thrown on dirty check of a many-to-one association during flush with CHECK_NULLABILITY set to false when the entity being dirty-checked: - has a "delayed-insert" ID generator (e.g., sequence, increment), - has been saved, but not flushed This may also happen on an update. I believe JDBCException should be thrown instead. This is reproduced by testMergeEntityWithNonNullableTransientEntity() in: - MultiPathCircleCascadeTest for Oracle10gDialect and PostgreSQLDialect (because these dialects use sequences for the native ID generator and the default for CHECK_NULLABILITY is false) - MultiPathCircleCascadeCheckNullibilityFalseTest for Oracle10gDialect and PostgreSQLDialect (because these dialects use sequences for the native ID generator and CHECK_NULLABILITY is explicitly set to false) - MultiPathCircleCascadeDelayedInsertTest for all dialects (because the test uses an "increment" generator and the default for CHECK_NULLABILITY is false) - MultiPathCircleCascadeCheckNullFalseDelayedInsertTest (because the test uses an "increment" generator and CHECK_NULLABILITY is explicitly set to false) The failure is masked in MultiPathCircleCascadeDelayedInsertTest and MultiPathCircleCascadeCheckNullFalseDelayedInsertTest because they incorrectly allow TransientObjectException as valid behavior.