Hi Alexey,
Could you please file a bug?
I'll try to fix it.
Thanks,
SAM
On 31.10.2013 0:58, Alexey Makhmutov wrote:
Hello,
We've noticed that Introspector behavior has changed in 7u40 compared
to 7u25. Here is the example to show the difference. Let's assume that
we have two classes:
public class A {}
public class B extends A {}
And two beans:
public class BeanA
{
public void setData(A value) {}
public A getData() {return null;}
}
public class BeanB extends BeanA
{
public void setData(B value) {}
public B getData() {return null;}
}
Here BeanB overrides getData method by using covariant return type. If
we run Introspector for BeanB, then 7u25 returns PropertyType of
'data' equal to 'B', while 7u40 reports it as 'A' instead. It seems,
that in 7u40 (and later versions) if return type of getter method can
be casted to the return type of getter with same name defined in
super-class, then Introspector just uses return type of getter in
parent class.
If I understand correctly, this change is result of following fix,
which was introduced in jdk8 and then ported to 7u40:
https://bugs.openjdk.java.net/browse/JDK-7189112.
So, here is my question: is the currently observed behavior in 7u40 a
desired one, which will be kept for future java updates? Should we
adapt our code to such behavior or just wait for some changes in next
updates? Well, overriding properties in children classes (with
covariant return type change) doesn't look like a particularly good
idea - and such situation was just impossible when JavaBeans spec was
originally created (it wasn't possible to override getter with
different return type until covariant returns were introduced in Java
5). However, current behavior looks a little counter-intuitive - we
have class with pair of matching getter and setter methods, while type
of property is resolved to the getter defined in base class. And, of
course, such change may break some existing code while upgrading to
the 7u40 (as it was in our case).
Thanks,
Alexey
