The test is java/awt/event/KeyEvent/KeyChar/KeyCharTest.java.

First, in this PIT build there is a new quite unexpected exception running Windows (7 and 10) (plus a dump of some EncodeD):

java.lang.IllegalArgumentException: invalid keyChar
        at java.awt.event.KeyEvent.<init>(java.desktop@9-ea/KeyEvent.java:1172)
        at sun.awt.windows.WToolkit.eventLoop(java.desktop@9-ea/Native Method)
        at sun.awt.windows.WToolkit.run(java.desktop@9-ea/WToolkit.java:314)
        at java.lang.Thread.run(java.base@9-ea/Thread.java:843)

It is new in this PIT: b124 doesn't throw this.

Second, a test bug:
This test depends on a char returned by getKeyChar() from KEY_PRESSED and KEY_RELEASED. However that char is not necessary defined. According to the spec, "KEY_PRESSED and KEY_RELEASED events are not intended for reporting of character input." -- it's only KEY_TYPED event must report a valid keychar.

As a result, the test fails on Windows with "Key char is not delete" and indeed, it is 0xFFFF (CHAR_UNDEFINED) which is expected.