DnD fails with JTextArea and JTextField (original) (raw)

Sean Chou zhouyx at linux.vnet.ibm.com
Mon Oct 10 07:35:22 UTC 2011

Hi Pavel, Artem,

I run jtreg on another machine and it passes. The attachment is the testcase and patch.

The following modification to testcase are done as suggested last time:

  1. Copyright in agreed form (the ibm info)
  2. Date in copyright
  3. All swing code in EDT
  4. Use realSync()

Please have a look, thanks.

On Thu, Sep 29, 2011 at 11:55 PM, Artem Ananiev <artem.ananiev at oracle.com>wrote:

On 9/28/2011 9:48 PM, Sean Chou wrote:

Hi Artem,

I can not validate the testcase after I changed to use realSynch, it always fails with throwing an exception. Testcase test\javax\swing\JSlider**6848475\bug6848475.java fails on my machine with timeoutexception as well. I run the test in its directory with command like this: jtreg -verbose:all -testjdk:...j2sdk/ bug6848475.java It may be a bug with realSync() implementation, possibly related to 6421984: realSync throws InfiniteLoop some times. Thanks, Artem 2011/9/28 Artem Ananiev <artem.ananiev at oracle.com_ _<mailto:artem.ananiev at oracle.**com <artem.ananiev at oracle.com>>>

On 9/27/2011 5:19 AM, Sean Chou wrote: Hi Pavel, Is toolkit.realSync() doing the job of "sleep 10s and throw an timeout exception" ? Definitely, not. SunToolkit.realSync() is what Robot.waitForIdle() was initially supposed to be: this call should wait until all the pending events (native, Java) to be processed. Sometimes, however, it fails to do so, e.g. when native events are turned into Java events, and Java event handlers call more native methods - then the TimeoutException is thrown. Thanks, Artem I found the main thread exit after realSynch is invoked. Should I catch that exception? bug6848475.java seems not catching it. 2011/9/26 Pavel Porvatov <pavel.porvatov at oracle.com_ _<mailto:pavel.porvatov at oracle.**com <pavel.porvatov at oracle.com>> <mailto:pavel.porvatov at oracle.**_com_ _<mailto:pavel.porvatov at oracle.**com <pavel.porvatov at oracle.com>>>>

Hi Sean, Hi Pavel, Thanks for new comments. I was looking at test\javax\swing\JTextArea_**6940863 and copied the copyright from there, and just put the initialization part of the swing code into EDT. I didn't find realSynch in Toolkit class so used synch. I'll modify that. The realSync is a SunToolkit method. About the first comment, do you mean to put the testcase and the fix together in a webrev ? Yep! It's hard to reconstruct fix from different mails. Don't forget to select an appropriate folder for the test... Thanks, Pavel

2011/9/26 Pavel Porvatov <pavel.porvatov at oracle.com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> > <mailto:pavel.porvatov at oracle.**_com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> >>> Hi Sean, Hi Pavel, Thanks for the comments. I modified the testcase according to the comments, please have a look at it again. I tried to run with jtreg and it runs well. Looks much better. Still have several comments: 1. Could you please sent complete fix as a webrev, but not parts of the fix as a single file? 2. Date of copyright... 3. You are still using Swing components on non-EDT thread. As I wrote before take a look at the test\javax\swing\JSlider_**6848475\bug6848475.java test... 4. Use toolkit.realSync() instead of toolkit.sync(). BTW: as described in javadoc realSync cannot be invoked on the EDT thread... Regards, Pavel P.S. Sorry for my stubborn =) But on some machines not accurate tests actually fail (e.g. on Solaris). Therefore later we must fix tests and that's a really boring task...

2011/9/20 Pavel Porvatov <pavel.porvatov at oracle.com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> > <mailto:pavel.porvatov at oracle.**_com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> >>> Hi Sean, The have several comments : 1. Could you please read http://openjdk.java.net/jtreg/ is it possible run your test via jtreg? 2. There is no copyright in the begin of test 3. There are no jtreg tags 4. All Swing code must be initialized on the EDT thread 5. Keep test minimal as possible, please. It helps other people to understand your code.... E.g. there is no need to create JButton with listener. 6. Note that the "frame.setVisible(true)" doesn't guarantee that after that line Frame is visible, you should use toolkit.realSync() here 7. No TODO, please 8. Are you sure your test should pass if exceptions occurs (see your catch blocks) Please take a look at other tests and use them as a good examples.... Regards, Pavel Hi Pavel, I wrote a test case for the behavior of DefaultCaret. Please take a look, it is attached. 2011/9/15 Pavel Porvatov <pavel.porvatov at oracle.com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> > <mailto:pavel.porvatov at oracle.**_com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> >>> Hi Sean, Hi Pavel, I'm comfortable with moving the checking into DefaultCaret#updateSystemSelection method._ About regression test, I'm not sure how to write, because it contains user operation. Can you give me a similar test so I can write one for this bug? Yes, you can find a lot examples in the test/javax/swing directory by word Robot, e.g. _test\javax\swing\JSlider__ 6848475\bug6848475.java. One hint: use reflection ONLY if there are no another ways to write a test... Regards, Pavel 2011/9/13 Pavel Porvatov <pavel.porvatov at oracle.com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> > <mailto:pavel.porvatov at oracle.**_com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> >>> Hi Sean, Hi Pavel , I'm sorry I didn't make update for this bug for a long time, and here is some recent investigation. The scenario is as follows: Suppose we are dragging "abcde" over TextField tf, which have "hello dragging" as its content. When we are dragging from start to end, there is a cursor moving from "h" to "g", which means the place to insert "abcde" if we drop it. When we dragging "abcde" exit tf, there will be a dragExit event, the tf needs to restore its original status after we drag out. Eg. if its cursor is between "h" and "e" in "hello", which appears like "h|ello", when we are dragging over it, it may like "hello dr|agging", and when drag exit, it needs to be "h|ello" again. So in dragExit handler, it calls javax.swing.TransferHandler.**cleanup(false),_ which means only to restore the original state. cleanup calls javax.swing.text.JTextComponent.setDropLocation_ to set the cursor to original position. And setDropLocation calls DefaultCaret.setDot and DefaultCaret.moveDot to set the state. The problem is moveDot doesn't know this is just to restore the original state, it treats the invocation as an action to select something. And it calls updateSystemSelection which will call java.awt.datatransfer._ Clipboard.setContent. And the selected content is changed from "abcde" to the original selected part of "hello dragging", then the drop operation finds it is not the string dragged and nothing is dropped. So I made a simple patch(attached) . It just check if the textField owns focus before updateSystemSelection, if it is not focused, it does not treat the moveDot as a selection action and does not call Clipboard.setContent. This works on Linux, however, DefaultCaret is shared by Linux and Windows while windows doesn't have this problem. So I don't think this is a correct patch, but it brings my question. I think it is strange for DefaultCaret to use setDot and moveDot to restore original state, especially moveDot will cause an updateSystemSelection operation, which makes moveDot much like an action from user instead of just restoring state. I'm not sure why it works well on windows, but I don't think it is right to call updateSystemSelection or it is not right to use setDot and moveDot to restore the original state. Is there any reason for that ? Thanks for the patch! I believe you are right and we shouldn't update system selection clipboard when the component doesn't have focus. I'd like to modify your fix and move checking into the DefaultCaret#*updateSystemSelection method:_ if (this.dot != this.mark && component != null && component.hasFocus()) { We also must write regression tests for fixes if possible, so an automatic test is needed as well. Could you please write a test for the fix? > I'm not sure why it works well on windows, That's because Windows doesn't have system selection clipboard... > Is there any reason for that ? No, that's a just bug... Regards, Pavel 2011/6/6 Pavel Porvatov <pavel.porvatov at oracle.com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> > <mailto:pavel.porvatov at oracle.**_com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> >>> Hi Sean, Hi, I reported, but the system doesn't reply me a bug number. It says "will give me email", but I haven't got one yet. Is this the right process, or I might make a problem when reporting? I don't know why the system didn't report bug ID, but your bug was filed successfully. You can find it here: http://bugs.sun.com/_** bugdatabase/viewbug.do?bug*id=7049024<http://bugs.sun.com/_bugdatabase/viewbug.do?bug_id=7049024>_ <http://bugs.sun.com/**_ _bugdatabase/viewbug.do?bug**id=7049024<http://bugs.sun.com/bugdatabase/viewbug.do?bugid=7049024> > Regards, Pavel 2011/5/27 Pavel Porvatov <pavel.porvatov at oracle.com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> > <mailto:pavel.porvatov at oracle.**_com_ _<mailto:pavel.porvatov at oracle.**com<pavel.porvatov at oracle.com> >>> Hi Sean, Hi all, I have a testcase related to DnD failure with JTextArea and JTextField on linux. The testcase is as follows: _/_ * DnDTest.java _/_ import java.awt.Color; import java.awt.Component; import java.awt.Dimension; import java.awt.FlowLayout; import java.awt.Frame; import java.awt.event.WindowAdapter; import java.awt.event.WindowEvent; import javax.swing.JTextArea; import javax.swing.JTextField; public class DnDTest extends Frame { Component c; public DnDTest() { super("Single Frame --- AWT Frame"); super.setBackground(Color.**gray);_ // set layout here. setLayout(new FlowLayout()); c = new JTextArea("JTextArea component"); c.setPreferredSize(new Dimension(400, 100)); add(c); c = new JTextField("JTextField component(No IM)"); c.setPreferredSize(new Dimension(400, 20)); add(c); addWindowListener(new WindowAdapter() { public void windowClosing(WindowEvent event) { System.exit(0); } }); setSize(850, 360); setVisible(true); } public static void main(String[] args) { new DnDTest(); } } Reproduce steps: 1. Run the testcase with b143 2. Open a new file with gedit and input some words like "abcde" 3. Drag "abcde" into JTextField and drop it there. 4. Once more, drag "abcde" into JTextField and then move out of the Frame (keep draging) and drag into JTextField again and drop it. Expectation: The second DnD inputs another "abcde" into JTextField. Result: The second DnD inputs nothing into JTextField. Yes, looks like a bug. The test case works on Windows as expected. Investigation: The JTextArea as well has this problem, and in step 4, if we drag "abcde" over JTextField and then drop into JTextArea, nothing is input into JTextArea either. However, if "abcde" is drag into JTextField or JTextArea directly or when JTextArea/Field are empty as in step 2, it works. Are there any comments? And can anyone file a bug for it please ? Anybody can file a bug, http://bugreport.sun.com/_**bugreport/<http://bugreport.sun.com/_bugreport/> <http://bugreport.sun.com/**bugreport/<http://bugreport.sun.com/bugreport/> > Regards, Pavel

-- Best Regards, Sean Chou

-- Best Regards, Sean Chou -- Best Regards, Sean Chou -- Best Regards, Sean Chou -- Best Regards, Sean Chou -- Best Regards, Sean Chou -- Best Regards, Sean Chou -- Best Regards, Sean Chou

-- Best Regards, Sean Chou -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.openjdk.java.net/pipermail/swing-dev/attachments/20111010/796b3858/attachment.html> -------------- next part -------------- A non-text attachment was scrubbed... Name: webrev.7049024.zip Type: application/zip Size: 151351 bytes Desc: not available URL: <http://mail.openjdk.java.net/pipermail/swing-dev/attachments/20111010/796b3858/webrev.7049024.zip>

More information about the swing-dev mailing list