Bump to Hibernate 5.6.3.Final

[gastaldi]

It finally undeprecates the TypeDef annotations

[Sanne]

thanks for preparing the PR!

[gastaldi]

Any chance to backport this PR to 2.6.1.Final? (#22114 (comment))

[Sanne]

Any chance to backport this PR to 2.6.1.Final? (#22114 (comment))

I have no strong objections but there's no strong reason to do it either... just a little risk. IMO it's best to calm down a little with backports, unless there's good reasons, no? People will get it soon, it's not like 2.7 will take ages :)

[gastaldi]

@Sanne for me the removal of the deprecation warnings is the main driver to get it asap but I think we can wait for 2.7 next year :)

[famod]

Looking at #22925 (comment) and given this PR here is marked for imminent 2.6.3, isn't this going to break projects that are using currently latest blaze-persistence 1.6.4?

/cc @beikov @gsmet

[gsmet]

Not sure to understand? Current version of BP in Quarkus is not compatible with 5.6.3?

[famod]

@gsmet I got that impression from Blazebit/blaze-persistence#1413

[beikov]

There is a breaking change in the org.hibernate.hql.spi.id.MultiTableBulkIdStrategy SPI in 5.6.3.Final but AFAIU the change is necessary. Not sure if we can somehow make the change in Hibernate in a backwards compatible way, but I'll look into this tomorrow.

[famod]

Just checked my project (which is using blaze-persistence-integration-hibernate-5.4) with current Quarkus main (which has Hibernate 5.6.3) and I'm no seeing any issues.
So I suppose this error only pops up if using some specific BP feature.

[beikov]

Are you overriding the hibernate.hql.bulk_id_strategy config property?

[famod]

No, we don't.

[beikov]

Then the contributor is maybe not applied correctly when using Quarkus for some reason. I'll dig into that.

[famod]

Hum, we are generating the ids of our entities on our own (UUIDs), so maybe that's the reason?

[gsmet]

@beikov @Sanne we need this sorted out before Wednesday evening. Either by avoiding the SPI breakage or by integrating a new BP version in the Platform for 2.6.3.Final.

[beikov]

Ok, so the Quarkus Hibernate bootstrap somehow strips off configuration values contributed in MetadataContributor implementations, which is why Blazebit/blaze-persistence#1413 is not an issue for Quarkus users.

Working on a fix for that now.

Apart from that, yes, I think that we should try to get a Hibernate 5.6.4.Final out that avoids breaking SPIs. I'll work on that first thing on Monday. From my experiments in Blaze-Persistence I can see that constructing the new additional parameter is possible through the other parameters, so we could add a default method:

interface MultiTableBulkIdStrategy {
    default void prepare(JdbcServices jdbcServices, JdbcConnectionAccess connectionAccess, MetadataImplementor metadata, SessionFactoryOptions sessionFactoryOptions) {
        final SqlStringGenerationContext sqlStringGenerationContext = SqlStringGenerationContextImpl.fromExplicit(
                jdbcServices.getJdbcEnvironment(), metadata.getDatabase(),
                sessionFactoryOptions.getDefaultCatalog(), sessionFactoryOptions.getDefaultSchema());
        prepare(jdbcServices, connectionAccess, metadata, sessionFactoryOptions, sqlStringGenerationContext);
    }
}

[beikov]

FYI I created https://hibernate.atlassian.net/browse/HHH-15032 for the issue.

[Sanne]

I was about to tag 5.6.4.Final but I'll hold off for now, and changed HHH-15032 to a blocker.

Please remember I won't be around in the next few days.