BUG: Properties of inherited models can be used on different child models

[rPraml]

Hello @rbygrave

we currently are updating to 13.20 and noticed, that https://github.com/ebean-orm/ebean/pull/3057 has triggered a bug.

What we do is, that we fetch a property of an inherited root, but the property itself lives in one child node:

Property prop = DB.getDefault().pluginApi().beanType(Animal.class).property("name");

As this property seems to be "animal" compatible, we use prop.value(animal) to read the property value of (any) animal.

But if the animal is a dog, and the property comes from cat, you are just reading the same property index from an other bean. (Note DB.getDefault().pluginApi().beanType(Animal.class).property("registrationNumber") has the same property index as name)

before #3057 this works by luck in our use case, as the property was unloaded and so, it just returns null. Now as a lazy load occurs, we were wondering, why we get a value, where we do not expect one

The PR provides a test and a suggested fix

[rPraml]

@rbygrave I've updated this PR. The check is now perfomed in the setIntercept/getIntercept code paths, but is more strict:

non inherited beans

• The property must always be fetched from the descriptor of this bean class.
• There are slight functional changes (edge cases), like fetching Customer.creTime which is owned by the superclass BasicDomain
  • before: it was possible to read/write all instances of BasicDomain
  • now: it will only work on customer:
  • if you fetch a property with db.beanType(Customer.class).expressionPath("cretime"), use it for Customer only (and not for ContactGroup....)
  • it is (currently) not possible to get a "universal" property with db.beanType(BasicDomain.class).expressionPath("cretime") as there is no descriptor.
  • I think this behaviour is acceptable.

inherited beans

• If there is a inherit hierarchy, you might get properties from the inheritance-root, which are not part of all child models.
• Reading such a (non existent) property returns null
• Writing to a property will fail

[rPraml]

@rbygrave I review all my open PRs. This PR is a small one and IMHO ready to review and merge.