RFR: 8267425: Intermittent failure of HonorDeveloperSettingsTest unit test

[Kevin Rushforth]

On Thu, 20 May 2021 07:18:28 GMT, Ambarish Rapte <arapte at openjdk.org> wrote:

> The same failure was earlier fixed at PR #496, which seems insufficient.
> The changes in this PR were also tried ealier in the PR #496 but later removed, as they seemed not necessary.
> Fix is to clean up the test, by removing all children added to root and hiding the stage.
> The test only fails on the GHA run, if it continues to fail after this fix then we shall skip it till next test sprint.

I tried to instrument the code earlier this morning to see what is going on, but I couldn't make it fail, even running it in a loop via GitHub actions on Linux.

One thing to note is that the failure suggests some leftover state that has been partially, but not fully cleaned up. Given where the NPE is, there must be a node in the `cacheContainerMap` of `StyleManager` with a non-null scene where the root node of that scene is null. However, unless I'm missing something, this should not be possible since a scene with a null root node is an error. Scene::setRoot throws NPE if you attempt to construct a Scene with a null root or if you later set it to null. Very odd.

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

PR: https://git.openjdk.java.net/jfx/pull/509