RFR: 8277817: java/awt/dnd/BadSerializationTest/BadSerializationTest.java failed: ClassNotFoundException: com.apple.laf.AquaImageFactory$SystemColorProxy [v9]

[Alexey Ivanov]

On Fri, 10 Dec 2021 17:03:49 GMT, Liam Miller-Cushon <cushon at openjdk.org> wrote:

>> This change updates the serialized objects used by `java/awt/dnd/BadSerializationTest/BadSerializationTest.java` using a similar approach to the previous fix in [JDK-8039082](https://bugs.openjdk.java.net/browse/JDK-8039082).
>
> Liam Miller-Cushon has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Add a few more missing calls to LookAndFeel.uninstallColors

I've been contemplating about what is being done in this PR… It feels wrong.

> > As of the usefulness of the serializations tests in Swing - it is a good coverage of the cleaning code in the L&fs. Each L&F should clean all its data when the app switch one LF to another, but if old L&F forget to cleanup then the app will get a memory leak.

Swing explicitly warns, “Serialized objects of this class will not be compatible with future Swing releases.”

Don't break this rule in this test? A component is serialized using one version of the JDK and is deserialized in another version of the JDK and even on another platform.

> This test doesn't switch L&Fs .. if it wasn't for mac defaulting to Aqua they'd all be using Metal.
> 
> So _this is an accidental by-product_, not the goal of the test.

It looks so.

If the L&Fs are expected to clean up when uninstalled, it should be tested thoroughly for all the components and Look-and-Feels. Only this way we can be sure the properties get reset.

While working on this code review, Liam has found and fixed quite a few places where clean-up wasn't done properly.

> And that goal be met by the test writing out the serialized objects and reading them back all at "test time", can't it ?
> 
> Seems to me right now we are stumbling over cases where what is serialized is inappropriate for persistent serialization.

Taking into account the above comments, this test could be simplified to write out the objects and then read them in. No persistent serialization which is _explicitly_ unsupported.


To be clear, I think the proposed change is good on its own: it adds convenience methods to clean up UI properties when a L&F is uninstalled, it adds the clean-up where it was missing. But it doesn't fit into the fix of this test failure. This work should be an enhancement performed under another bugid.

-------------

PR: https://git.openjdk.java.net/jdk/pull/6603