[Rev 01] RFR: 8236259: MemoryLeak in ProgressIndicator

[Kevin Rushforth]

On Mon, 2 Mar 2020 12:53:21 GMT, Florian Kirmaier <fkirmaier at openjdk.org> wrote:

>>> It's now part of the systemtests, because the memory-semantics for the tests in controls is changed due to the TestToolkit.
>> 
>> just curious: does that imply that you think the tests in controls ... rather useless?
>
> @kleopatra 
> It's not useless, but having a special Toolkit which is only used inside of JavaFX itself is very unreasonable. It would make much more sense, to use the same basis for unit-test as the rest of the JavaFX-Community which is monocle. The systemTests also seem to work quite well.

The unit tests in the `javafx.controls` and `javafx.graphics modules` have used the StubToolkit since the early days. These tests are actually quite useful, but are not able to test anything that depends on the glass toolkit, prism rendering, etc.

[JDK-8090538](https://bugs.openjdk.java.net/browse/JDK-8090538) was filed several years ago to explore replacing the StubToolkit with a headless, Monocle-based glass toolkit, but it didn't get very far. There would be some advantages in doing this, but some drawbacks as well (see the discussion thread referred by the above JBS issue). More importantly, it would be a lot of work to do it right, and it isn't clear at this point whether it would be justified by the benefit.

So as far as _this_ PR goes, a test in the systemTests project seems fine.

I'll put this back on my review queue.

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

PR: https://git.openjdk.java.net/jfx/pull/71