RFR: 8293862: javax/swing/JFileChooser/8046391/bug8046391.java failed with 'Cannot invoke "java.awt.Image.getWidth(java.awt.image.ImageObserver)" because "retVal" is null' [v4]

[Alexey Ivanov]

On Sun, 8 Jan 2023 09:09:56 GMT, Sergey Bylokhov <serb at openjdk.org> wrote:

>>> Did we check how the native code works w/o using JFileChooser, buy FileSystemView, or by the native example? Is it possible that it is intermittent because of timing related to "com" thread?
>> 
>> 
>> 
>>> 
>> 
>> I have tested in headful system multiple time, never reproduced this issue...... Can timing affect in headless and not in headful......?
>
>> I have tested in headful system multiple time, never reproduced this issue...... Can timing affect in headless and not in headful......?
> 
> It should not. Does it actually have any dependency on "headful" envirement? The native code we used does not create any UI element or something like that. It is better to debug it on the same/similar config/system where the bug is reproduced, since as far as I understand it is not a rare case.

@mrserb, @TejeshR13, There could be a dependency on the host being headful even if we don't create UI. You may remember there were a couple of test failures where colours in `VolatileImage` were different from the colours in `BufferedImage` because `VolatileImage` used `IndexedColorModel` rather than the expected 32-bit colour.

Recently I saw another problem where a seemingly headless test didn't work as expected on the headless system. Windows L&F wasn't set correctly and `JTextField` didn't have the expected border.

Windows Server systems may not have full-featured desktop environment, in this case we may see problems with *headless* tests which in fact need *headful (full-featured)* desktop environment even though such tests don't show the UI.

I agree it is worth exploring further why fetching the icon of the _Downloads_ folder returns `null`. Could either of you submit a bug, please?

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

PR: https://git.openjdk.org/jdk/pull/11104