Abe White wrote:
Is that a hint to set openjpa.jdbc.SchemaFactory=native
(ForeignKeys=true)?
No. I'm saying that if these secret tests rely on OpenJPA to detect
foreign keys based on the above setting,
Secret tests?
then it's likely that the
problem is in OpenJPA's ability to detect the existing foreign keys
rather than in OpenJPA's constraint analyses and SQL ordering. For
example, I know in some versions of MySQL the standard
DatabaseMetaData.getExportedKeys call didn't work unless the keys
were created with ALTER TABLE statements rather than in the CREATE
TABLE statement.
Understood. I was however under the impression that correctly annotated
relations would make it unnecessary to resort to database metadata to
build a correct dependency graph --it appears that's not the case.
FWIW, setting SchemaFactory=native(ForeignKeys=true) does fix that
particular problem with ConstraintUpdateManager.
