We will need to make a slight change to execution of JPA native SQL queries to add an "auto flush".
The problem with this is that, unlike with HQL/JPQL, we do not know the tables affected by a native SQL query (and doing so would be extremely difficult and not worth the effort). Historically, Hibernate users would have relied on the addSynchronizedQuerySpace/addSynchronizedEntityName methods of org.hibernate.SQLQuery to indicate the tables needing flushed. JPA however has no such notion.
As I see it, I think the change here needs to be to implicitly perform a full flush when a JPA native SQL query is executed where the wrapped/underlying org.hibernate.SQLQuery does not define any "query spaces"; a partial flush just is not possible in that case. That allows us to pass the TCK, but still allows users to set "query spaces" if they wish to do the more efficient partial auto-flush.
For posterity, HEM users can still accomplish this by unwrapping to Hibernate's native org.hibernate.SQLQuery adding to the "synchronized query spaces", like:
There are 3 forms of adding "synchronized query spaces":
addSynchronizedQuerySpace - Add a table to synchronize on
addSynchronizedEntityClass - Add the tables to which the given entity (by class) is mapped
addSynchronizedEntityName - Add the tables to which the given entity (by name) is mapped
Bulk closing tickets resolved in released versions
I tested this with the 4.3.5 version an SQL queries don't force a flush before execution.
Check the unit test I attached on this issue: https://hibernate.atlassian.net/browse/HHH-9315