HibernateTransactionManager (unintentionally) bound to Hibernate 5.2 SharedSessionContractImplementor [SPR-17557]

[spring-projects-issues]

Taras Tielkes opened SPR-17557 and commented

The compiled version of HibernateTransactionManager seems to contain a reference to SharedSessionContractImplementor (perhaps from the lambda call?).

When running with Hibernate 5.1.x in the classpath, resolution of SharedSessionContractImplementor will fail, triggering a NoClassDefFoundError.

Please take a look at the output of "javap -v" for the compiled version of HibernateTransactionManager in the latest release jar.

---

Affects: 5.1.3

Issue Links:

• Propagate read-only status to Hibernate Session through setDefaultReadOnly [SPR-16956] #21494 Propagate read-only status to Hibernate Session through setDefaultReadOnly
• HibernateTransactionManager should lazily acquire JDBC Connection (like HibernateJpaDialect) [SPR-17216] #21749 HibernateTransactionManager should lazily acquire JDBC Connection (like HibernateJpaDialect)

Referenced from: commits 0841dc5

[spring-projects-issues]

Zhang Jie commented

spring-orm-5.1.3 uses hibernate-core-5.3.7.Final as dependency, you shouldn't use Hibernate 5.1.x in the classpath.

[spring-projects-issues]

Juergen Hoeller commented

We aim to support any Hibernate ORM 5.x version at runtime, so this is nevertheless a bug... even if we generally recommend Hibernate ORM 5.3.7.

[spring-projects-issues]

Juergen Hoeller commented

This seems to be caused by the sessionImpl::connection method reference, where sessionImpl is a SessionImplementor type but the connection() method is inherited from SharedSessionContractImplementor since Hibernate ORM 5.2 there.

Taras Tielkes, have you seen this as a recent regression in Spring Framework 5.1.x? Does it work with Spring Framework 5.0.x still?

As far as I can see, a simple change to () -> sessionImpl.connection() solves the problem at bytecode level since it compiles to a SessionImplementor reference again. Assuming it still works with Spring Framework 5.0.x (where a similar direct call in in place), that's the change I intend to apply for the 5.1.4 release.

[spring-projects-issues]

Taras Tielkes commented

Juergen Hoeller, for this project we're just now migrating from 4.3.x to 5.1.x, so I don't have any feedback for 5.0.x versions.
However, I'd be happy to check later this evening.

[spring-projects-issues]

Juergen Hoeller commented

No worries, I'll commit the change to master assuming that it solves the problem, so please rather try 5.1.4.BUILD-SNAPSHOT tonight.

[spring-projects-issues]

Taras Tielkes commented

Juergen Hoeller I have tried the SNAPSHOT from last night, and I can confirm that it fixes the issue for us.

This regression was a subtle one - not apparent from reading the source. I'm wondering if it would be possible to catch such cases in an automated way...

As always, thanks for the quick service.