On Wed, 1 Dec 2021 04:27:03 GMT, Sergey Bylokhov <s...@openjdk.org> wrote:
> 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. 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. 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. ------------- PR: https://git.openjdk.java.net/jdk/pull/6603