6.2.0 broke with "Could not register object [@someHash] under bean name 'blabla': there is already object [@sameHash] bound"

[fmarot]

Hello, until 6.1.7 my program had been working fine for many years, but 6.2.0 introduced a change that made it break with:

"Could not register object [presetClientService@5404b7d9] under bean name 'blabla': there is already object [presetClientService@5404b7d9] bound"

The hash Id is the same so I'm under the impression that some bean is found in the bean registry and spring tries to re-add it in the registry. Why ?!
I was not able to create a simple reproducer yet.
To give you more details the code looks like this with 3 @configuration classes referenced:

AnnotationConfigApplicationContext sphereContext = new AnnotationConfigApplicationContext(MainSphereSpringContext.class);
AnnotationConfigApplicationContext moduleContext = new AnnotationConfigApplicationContext();
moduleContext.register(ModuleSpringConfig.class, OleaDockingSpringConfig.class);
moduleContext.setParent(sphereContext);
moduleContext.refresh();

And the failing @configuration class contains:

@Bean
public PresetClientService presetClientService() {
    return new PresetClientService(someService());
}

@Bean
public ModuleConfigPresenterImpl configPresenter() {
    return new ModuleConfigPresenterImpl(segmentActionDirectory(), presetClientService());
}

The problem comes from the call to presetClientService(). If I replace this call with null, the loading of the context is OK.
Was working fine since years and broke with 6.2.0.
I could re-try to create a real-failing sample if really needed, but for now I failed at it.
Regards

[bclozel]

We are going to need that sample unfortunately. I'm not sure we have enough information to understand where the problem is.

[jhoeller]

Also, please try to reproduce this against 6.2.3 rather than 6.2.0. Several regressions have been fixed in the meantime which makes it easier to isolate where a remaining problem could be.

[fmarot]

Sorry, I should have mentioned I also tested against the latest 6.2.3 and 7.0.0-M2 but they also have the same problem.

[jhoeller]

@fmarot for a start, it would help to see the stacktrace for that exception. That's essentially an assertion, for some reason addSingleton seems to be called twice for the same bean instance in your scenario.

[fmarot]

Here is the stacktrace. I'll try to write a SSCCE (https://sscce.org/) but no guarantee.

java.lang.IllegalStateException: Could not register object [com.oleamedical.module.viewing.PresetClientServiceImpl@285489cc] under bean name 'presetClientService': there is already object [com.oleamedical.module.viewing.PresetClientServiceImpl@285489cc] bound
	at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.addSingleton(DefaultSingletonBeanRegistry.java:155)
	at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.getSingleton(DefaultSingletonBeanRegistry.java:372)
	at org.springframework.beans.factory.support.AbstractBeanFactory.doGetBean(AbstractBeanFactory.java:337)
	at org.springframework.beans.factory.support.AbstractBeanFactory.getBean(AbstractBeanFactory.java:202)
	at org.springframework.beans.factory.support.DefaultListableBeanFactory.instantiateSingleton(DefaultListableBeanFactory.java:1155)
	at org.springframework.beans.factory.support.DefaultListableBeanFactory.preInstantiateSingleton(DefaultListableBeanFactory.java:1121)
	at org.springframework.beans.factory.support.DefaultListableBeanFactory.preInstantiateSingletons(DefaultListableBeanFactory.java:1056)
	at org.springframework.context.support.AbstractApplicationContext.finishBeanFactoryInitialization(AbstractApplicationContext.java:987)
	at org.springframework.context.support.AbstractApplicationContext.refresh(AbstractApplicationContext.java:627)
	at com.oleamedical.module.application.AbstractSeriesViewerModuleApp.createSpringContext(AbstractSeriesViewerModuleApp.java:480)
	at com.oleamedical.module.application.AbstractSeriesViewerModuleApp.newDocumentRunnable(AbstractSeriesViewerModuleApp.java:704)

[jhoeller]

Interesting, that's for a top-level bean even - no nesting, no circle. Somehow we seem to end up in a pre-registered state in that regular code path, that's clearly a bug.

That helps a lot for investigating this, thanks! I'll try to sort it out for the 6.2.4 release this Thursday which we'll do just-in-time for the Spring Boot 3.4.3 release.

[fmarot]

Thanks @jhoeller , I got the latest spring-beans-6.2.4-20250218.142954-8.jar and I can confirm that you fixed the problem. Big thank you !

[jhoeller]

Thanks for the immediate testing of the snapshot, much appreciated!