Backport fix for multiple invocations of ImportBeanDefinitionRegistrars [SPR-9939]

[spring-projects-issues]

Oliver Drotbohm opened SPR-9939 and commented

I think this is a reasonable backport candidate as with Spring Data we don't want to force users on to 3.2 abruptly and the logs for the overridden bean definitions might lead users into thinking they're setting up their application context invalidly. Furthermore, even if the duplicate invocation doesn't cause any issues with our ImportBeanDefinitionRegistrar it will for implementations which use auto-generated bean names as they'll accidentally register two bean instances which might break autowiring by type as no unique bean then can be found.

---

This issue is a backport sub-task of #14558

Issue Links:

• @Enable registrars invoked with subclass for annotation placed on superclass (3.2.x) [SPR-11251] #15876 @Enable registrars invoked with subclass for annotation placed on superclass (3.2.x)

Referenced from: commits spring-projects/spring-hateoas@48b039a, spring-projects/spring-hateoas@3976ae1

[spring-projects-issues]

Oliver Drotbohm commented

Maybe worth mentioning that I am perfectly fine for this one not making it into 3.1.3. I just think it's important to have in the 3.1.x timeline before we officially declare it as done.

[spring-projects-issues]

Chris Beams commented

Acknowledged, thanks Ollie.

[spring-projects-issues]

Chris Beams commented

commit d547570bcd606abc6ac34a82930d36fb83c24002
Merge: a420e84 a9290d8
Author: Chris Beams <cbeams@vmware.com>
Commit: Chris Beams <cbeams@vmware.com>

    Merge branch 'backport-SPR-9925' into 3.1.x
    
    * backport-SPR-9925:
      Backport "Polish @Imports search code"
      Backport "Ensure @Imports are processed in correct order"
      Backport "Prevent duplicate @Import processing"
      Backport "Polish Javadoc for @Import"