Spring Cloud discovered a recent regression with the binder where the ValueObjectBinder is used instead of the JavaBeanBinder. They use a pattern like this:
public class Foo {
Foo() {
}
Foo(String name) {
..
}
// getName / setName
}
This allows for properties to be defined in two ways: test[0]=hello or test[0].name=hello.
In the first instance the ObjectToObject converter is used to create the instance. In the second case, bean binding should be used.
The regression only happens then the bound object doesn't have a supplied value (e.g. it's in a list).