FWD » Core Development » User Interface

ChUI Swing: when I press Escape, the client crashes with NPE here:

Thread [AWT-EventQueue-1] (Suspended (exception NullPointerException))    
LinkedBlockingQueue<E>.offer(E) line: 411    
LinkedBlockingQueue<E>(AbstractQueue<E>).add(E) line: 95    
LinuxKeyboardReader.keyPressed(KeyEvent) line: 214    
ChuiClient(Component).processKeyEvent(KeyEvent) line: 6497    

The origin in this LinuxLeyboardReader, l. 214:

         queue.add(null);

The queue field is defined as

   /** Incoming keystroke buffer. */
   protected final LinkedBlockingQueue<KeyCode> queue = new LinkedBlockingQueue<>();

And the LinkedBlockingQueue explicitly does not support adding nulls:

    public boolean offer(E e) {
        if (e == null) throw new NullPointerException();

Are we sure we need to add a null keyboard event to the queue?

Tomasz: Have you encountered this issue in <large_chui_app> regression testing?

Greg Shah wrote:

Are we sure we need to add a null keyboard event to the queue?

Yes, it is used to signal a modifier key in multiple places. I am just checking in a fix.

Hynek Cihlar wrote:

Greg Shah wrote:

Are we sure we need to add a null keyboard event to the queue?

Yes, it is used to signal a modifier key in multiple places. I am just checking in a fix.

Have you created a branch for this issue?

The issue is fixed in 7742a revision 14707. The problem only showed in Swing ChUI client. Please review.

Vladimir, please also retest.

Greg Shah wrote:

Are we sure we need to add a null keyboard event to the queue?

Tomasz: Have you encountered this issue in <large_chui_app> regression testing?

No, it was a problem with limited taBuf Deque size causing IllegalStateException

I added missing code comments in 7742a revision 14708.

Vladimir Tsichevski wrote:

Hynek Cihlar wrote:

Vladimir, please also retest.

The NPE disappeared now, but it seems the code works incorrectly.

Pressing Escape has no visible effect, and the boolean escPressed just flips on and off.

Please compare the behavior with the terminal client. If you find differences post your test case.

Hynek Cihlar wrote:

Vladimir Tsichevski wrote:

Hynek Cihlar wrote:

Vladimir, please also retest.

The NPE disappeared now, but it seems the code works incorrectly.

Pressing Escape has no visible effect, and the boolean escPressed just flips on and off.

Please compare the behavior with the terminal client. If you find differences post your test case.

Do you mean I should compare two FWD ChUI native mode?

Vladimir Tsichevski wrote:

Hynek Cihlar wrote:

Vladimir Tsichevski wrote:

Hynek Cihlar wrote:

Vladimir, please also retest.

The NPE disappeared now, but it seems the code works incorrectly.

Pressing Escape has no visible effect, and the boolean escPressed just flips on and off.

Please compare the behavior with the terminal client. If you find differences post your test case.

Do you mean I should compare two FWD ChUI native mode?

I see no differences.

Also: after the Escape was pressed, the new key press is ignored. This holds both for Swing and native modes.

Vladimir Tsichevski wrote:

I see no differences.

Also: after the Escape was pressed, the new key press is ignored. This holds both for Swing and native modes.

How does native OE behave?

Hynek Cihlar wrote:

Vladimir Tsichevski wrote:

I see no differences.

Also: after the Escape was pressed, the new key press is ignored. This holds both for Swing and native modes.

How does native OE behave?

Procedure complete. Press space bar to continue.

Vladimir Tsichevski wrote:

Hynek Cihlar wrote:

Vladimir Tsichevski wrote:

I see no differences.

Also: after the Escape was pressed, the new key press is ignored. This holds both for Swing and native modes.

How does native OE behave?

Procedure complete. Press space bar to continue.

This is different from the original issue. I will check how 6667e behaves and how does it resolve this regression.