On Tue, 24 Jun 2025 16:14:28 GMT, Prasanta Sadhukhan <psadhuk...@openjdk.org>
wrote:
> I dont think the test is invalid..It was in closed repo and tested for so
> many years before it got opensourced L&F is anyway applied to entire
> component tree in the fix via `SwingUtilities.updateComponentTreeUI(frame);`
I believe what fixed the test is that you removed the line
`UIManager.setLookAndFeel(save)` which restored the L&F to whatever L&F that
was set before at the time when the test started, in this particular case
Nimbus L&F according to the bug description.
This ensures the current L&F remains Windows L&F so that the missing resources
are now found. If there were another L&F to test, or if Windows L&F weren't the
last in the list, the test will continue to fail because the current L&F will
again be different from Windows L&F.
Indeed, it does. If I comment out `UIManager.setLookAndFeel(save)` from the
original test code, the menu items under Windows do have check marks and
bullets.
At the same time, `SwingUtilities.updateComponentTreeUI(frame)` has no effect…
however, it does change the L&F of the menu bar and the menu items on the menu
bar.
The menu items that are displayed in popups aren't visible and aren't part of
the component hierarchy, therefore they remain at the assigned L&F which was
active when they were created.
> Similar method was used in opensourced test
> javax/swing/JTabbedPane/bug4499556.java by @prrace ..Its the similar strategy
> used in many closed/open manual tests for testing different L&Fs in applet era
Yes, I've seen other tests which employ this strategy… where several components
in a container are created with different L&Fs. And often everything seem to
work… except for the cases where it doesn't.
This is the case where it doesn't…
Continuing with my example above where I commented out the line
`UIManager.setLookAndFeel(save)`, the **Windows** menu looks good but clicking
the **Metal** menu now throws NPE:
java.lang.NullPointerException: Cannot invoke "java.awt.Font.hashCode()"
because "font" is null
at
java.desktop/sun.font.FontDesignMetrics$MetricsKey.init(FontDesignMetrics.java:213)
at
java.desktop/sun.font.FontDesignMetrics.getMetrics(FontDesignMetrics.java:282)
at
java.desktop/sun.swing.SwingUtilities2.getFontMetrics(SwingUtilities2.java:1235)
at
java.desktop/javax.swing.JComponent.getFontMetrics(JComponent.java:1700)
at
java.desktop/sun.swing.MenuItemLayoutHelper.reset(MenuItemLayoutHelper.java:126)
at
java.desktop/javax.swing.plaf.synth.SynthMenuItemLayoutHelper.<init>(SynthMenuItemLayoutHelper.java:118)
at
java.desktop/javax.swing.plaf.synth.SynthGraphicsUtils.paint(SynthGraphicsUtils.java:578)
at
java.desktop/javax.swing.plaf.synth.SynthMenuUI.paint(SynthMenuUI.java:278)
at
java.desktop/javax.swing.plaf.synth.SynthMenuUI.update(SynthMenuUI.java:245)
at
java.desktop/javax.swing.JComponent.paintComponent(JComponent.java:852)
at java.desktop/javax.swing.JComponent.paint(JComponent.java:1128)
at
java.desktop/javax.swing.JComponent.paintChildren(JComponent.java:961)
at java.desktop/javax.swing.JComponent.paint(JComponent.java:1137)
at
java.desktop/javax.swing.JComponent.paintToOffscreen(JComponent.java:5324)
This makes sense, the **Metal** menu itself was created while Nimbus was the
active L&F. (Yes, you moved the menu creation after the new L&F was set.) This
just proves my point that multiple L&Fs aren't really supported.
> @prrace Any comments on this test regarding its validity and fix?
I'm looking forward to seeing Phil's comments.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/25907#issuecomment-3001288429
