Hibernate bumps to leads to Hibernate Reactive NPE and Hibernate Validation constraint validation failures where it shouldn't about quarkus

[fedinskiy]

@DavideD I prepared a minimalistic reproducer, which doesn't involve the TS.

1. git clone https://github.com/fedinskiy/reproducer -b db-panache-no-update
2. mvn quarkus:dev
3. curl -X PUT http://localhost:8080/library/book/2/Around_the_World_in_Eighty_Days # fails
4. mvn quarkus:dev -Dquarkus.platform.version=3.10.0

$ curl -X PUT http://localhost:8080/library/book/2/Around_the_World_in_Eighty_Days # succeeds
4
$ curl -XGET http://localhost:8080/library/book/4
Around_the_World_in_Eighty_Days

There is also a test, ResourceTest, which does the same. mvn clean test fails, mvn clean test -Dquarkus.platform.version=3.10.0 succeeds

UPD: It also looks like, that this error affects Postgres and Oracle, but not MariaDB, MySQL, or MSSQL

[DavideD]

I guess. I will check.

[DavideD]

I have some test failures, but I see a different error.
I'm working on it

[quarkus-bot]

/cc @DavideD (hibernate-reactive), @gavinking (hibernate-reactive)

[michalvavrik]

cc @yrodiere @Sanne

[yrodiere]

Hey @DavideD , could this be a regression in Hibernate Reactive?

[DavideD]

@michalvavrik Not sure if this is related, but an association in Hibernate is supposed to be mapped in a different way than the one in the tests.

It should be:

@Entity
class Book {
   ...
    @ManyToOne
    Author author;
   ...
}

Is there a specific reason for the test to use just the id?

[michalvavrik]

I have some test failures, but I see a different error.

Sorry, it was late night, if you want adjustments to the reproduce, just shout!

Is there a specific reason for the test to use just the id?

Hey @DavideD ,

In general, we just try to identify legit cases, I don't think it is required for this particular test. So we can change this. Before we do so, I'd like to understand this is not supported / encouraged use case:

class Book {
@NotNull
private Integer author;
}

I'd think that mapping only id when I don't need to work with entity is logical. I suppose if an entity instance is lazy loaded but still contains PK (that is loaded eagerly) then I can use the entity data type instead. So people shouldn't do this but it worked (with quotes :-)) in past and now the validation is improved?

[DavideD]

It's supported in the sense that it should work for simple cases. Not encouraged.

But:

• it's not mentioned in the official Hibernate documentation
• Hibernate is not aware that that's an association, meaning that they are not covered by these tests
• users might not understand that this approach is not encouraged and copy it

That said, I don't know the requirements for this project.
And I don't think it's related to this issue, so I will move on.

[DavideD]

Sorry, I couldn't figure out why it's not working in Quarkus.
I've created a test for Hibernate Reactive and it seems to work as expected.

I couldn't really launch the test suite, when I've tried manually the validation didn't seem to work.
I think an error is happening when Hibernate is trying to initialize Validator: https://github.com/hibernate/hibernate-orm/blob/9ba0dd7af03d63dc39ce0ac8acd6688803ea0af9/hibernate-core/src/main/java/org/hibernate/boot/beanvalidation/BeanValidationIntegrator.java#L139

In theory, it should call TypeSafeActivator, but something happens before reaching that point. But the debugger couldn't tell me anything.

It's worth mentioning, that if something goes wrong with the creation of the validator factory, Hibernate won't say anything about it.

I don't think I will have more time to work on this, if somebody wants to continue.

[michalvavrik]

Thank you for the time you put into the investigation @DavideD .

[DavideD]

Thanks a lot! I can finally see the nullpointer and I'm trying to figure out what's going on.

[DavideD]

Ooops...

[fedinskiy]

@DavideD "debugging is a murder mystery, there you are a detective, a culprit and a victim".

Can we expect a fix during this week, or we better disable the affected tests for a while?

[yrodiere]

Can we expect a fix during this week, or we better disable the affected tests for a while?

The latter. Even if @DavideD is going to work during his PTO (bad!), there's little chance the fix makes its way to a Hibernate Reactive release and a Quarkus release by the end of the week.

[DavideD]

Yes, sorry. Like Yoann said. At the moment, I can't make any promises.