Add ObserverMethod#getDeclaringBean method.

[manovotn]

Fixes #563

I am deliberately keeping the return value for synth observers undefined because I understand some implementations can choose to return null but in Weld this is apparently never the case and instead synth. OMs belong to the extension declaring them.

[Ladicek]

The Build Compatible Extension API demands that for synthetic observers, ObserverInfo.bean() returns null. Can that be implemented on top of this?

[manovotn]

The Build Compatible Extension API demands that for synthetic observers, ObserverInfo.bean() returns null. Can that be implemented on top of this?

@Ladicek that sounds like a mistake we made. I think the declaring class and the declaring bean are both (in case of Portable Extensions) derived from https://github.com/eclipse-ee4j/cdi/blob/master/api/src/main/java/jakarta/enterprise/inject/spi/configurator/ObserverMethodConfigurator.java#L76-L82
According to spec, it'll always result in a non-null value. On top of that, for your StillDI impl, I expect this to default to the class of the Portable Extension via which we implement build compatible extensions.

So the way I see it, we have two options:

• State in Lite it is undefined return value for synth observers
• State in Lite that the return value is the extension defining it
  • I don't think this is possible because in Lite, compatible extensions are not beans

[Ladicek]

Actually I'm looking at it again and it seems even my ObserverInfo.isSynthetic() method implementation is wrong (because ProcessSyntheticObserverMethod.getAnnotatedMethod() is non-portable). Wow.

[Ladicek]

At the same time, one can distinguish synthetic observers from non-synthetic by the event type, and so it is not necessary to invoke any of these non-portable methods. So it's all good.

[Ladicek]

With this change weld/core#2605, all should be fine and this can be merged.

[manovotn]

With this change weld/core#2605, all should be fine and this can be merged.

Thanks a lot for that!