<Swing Dev> [11] Review Request: JDK-6481195 ListSelectionListener indicates events on model.addElement after model.clear()
Semyon Sadetsky
semyon.sadetsky at oracle.com
Fri Feb 2 16:22:28 UTC 2018
Hi Pankaj,
I think it would be wrong to fix an issue by adding yet another case
that contradicts the specs.
In the most usages the clearSelection() call is followed by
setAnchorSelectionIndex(-1)/setLeadSelectionIndex(-1). I think that
would be more safe way to fix the issue by adding those calls into the
clearSelection(). And you will need to cleanup the usages.
Also, please file the issue about removal of anchor/selection with index
!=0.
--Semyon
On 02/01/2018 11:25 PM, Pankaj Bansal wrote:
>
> Hi Semyon,
>
> <<I'm not sure that the spec should point to specific numbers.
> <<Ok, I give you an example for clarification
> <<list.setSelectedIndices(new int[] {1, 0}); // selection is [0 ,1]
> lead/anchor = 0/0
> <<model.removeElementAt(0); // selection is [0] lead/anchor = 0/0
> <<after you fix: lead/anchor = -1/-1.
> <<the specs say
> << * Return the first/second index argument from the most recent
> call to
> << * setSelectionInterval(), addSelectionInterval() or
> removeSelectionInterval().
>
> Ok, I got your point. In the example you gave, as addSelectionInterval
> was called from setSelectedIndices with 0 at last, the lead/anchor
> should be 0/0 not -1/-1.
>
> But what I am saying is, this is true for all the indices not just 0.
> For all other indices, the anchor/lead is incorrect. It is made
> correct for just 0 index by handling it specially in code.
>
> Consider the same example you gave after adding 1 in every index.
>
> list.setSelectedIndices(new int[] {2, 1}); // selection is [1 ,2]
> lead/anchor = 1/1
> model.removeElementAt(1); // selection is [1] lead/anchor = 0/0
> (before the fix. I have verified it)
>
> If we go by the API description, even this is wrong as anchor/lead
> should be 1/1 and it will be wrong for all the indices except 0 as
> they have specially handled 0 to be correct. I have just made it
> incorrect for index 0 also. I think this is bug which needs to be
> addressed separately independent of the index.
>
> Again, please let me know if I am missing something here.
>
> Regards,
>
> Pankaj Bansal
>
> *From:*Semyon Sadetsky
> *Sent:* Thursday, February 1, 2018 9:51 PM
> *To:* Pankaj Bansal; swing-dev at openjdk.java.net
> *Subject:* Re: <Swing Dev> [11] Review Request: JDK-6481195
> ListSelectionListener indicates events on model.addElement after
> model.clear()
>
> On 01/31/2018 11:04 AM, Pankaj Bansal wrote:
>
> Hi Semyon,
>
> Thanks for the review.
>
> << This is directly specified in getLeadSelectionIndex()/
> getAnchorSelectionIndex().
>
> Sorry but I am not able to find anything specific to handling of
> case of 0th item being deleted when first row is selected or
> anything which states that anchor or lead can’t be -1. I guess I
> may be missing something here or not able to infer what you are
> saying. Can you please point out which specific statement
> describes the behavior you suggested.
>
> I'm not sure that the spec should point to specific numbers.
>
> Ok, I give you an example for clarification
>
> list.setSelectedIndices(new int[] {1, 0}); // selection is [0 ,1]
> lead/anchor = 0/0
> model.removeElementAt(0); // selection is [0] lead/anchor = 0/0
>
> after you fix: lead/anchor = -1/-1.
>
> the specs say
>
> * Return the first/second index argument from the most recent call to
> * setSelectionInterval(), addSelectionInterval() or
> removeSelectionInterval().
>
>
> <<Anchor and lead indexes are used in lot of places. Which tests
> did you run after the fix?
>
> I have run following tests. I have run most of them(closed/jck) on
> Windows, but I think if this change is to affect something, it
> would not be platform specific. All the following runs have passed
> successfully (no new failures/errors).
>
> All automated tests in: Open/test/jdk/javax/swing/JList,
> /JTable, /JTableHeader, /JTree
>
> All automated tests in: Closed/test/jdk/javax/swing/JList,
> /Jtable, /JTableeHeader, /JTree, /DefaultListSelectionModel,
> /DefaultListModel, /DefaultListCellRenderer
>
> All tests in: jck/swing_javax/JList, /JTable, /Tree,
> /DefaultListSelectionModel, /DefaultSingleSelectionModel ,
> /DefaultListModel, /DefaultListCellRenderer
>
> Few demos like SwingSet2, Java2D
>
> Please let me know if you think I should run some more tests.
>
> Regards,
>
> Pankaj Bansal
>
> *From:*Semyon Sadetsky
> *Sent:* Wednesday, January 31, 2018 9:49 PM
> *To:* Pankaj Bansal; swing-dev at openjdk.java.net
> <mailto:swing-dev at openjdk.java.net>
> *Subject:* Re: <Swing Dev> [11] Review Request: JDK-6481195
> ListSelectionListener indicates events on model.addElement after
> model.clear()
>
> Hi Pankaj,
>
> This is directly specified in getLeadSelectionIndex()/
> getAnchorSelectionIndex().
>
> Anchor and lead indexes are used in lot of places. Which tests did
> you run after the fix?
>
> --Semyon
>
> On 01/31/2018 12:38 AM, Pankaj Bansal wrote:
>
> Hi Semyon,
>
> << When only the first list element deleted the anchor and
> lead selection indexes become -1 after your fix while should be 0.
>
> Is it mentioned somewhere that it should be 0 ? I mean if lead
> index was 2 and only 3^rd element is deleted, lead becomes 1.
> Same should be true when lead is 0 and first element is
> deleted. Is it rule that lead or anchor can’t go beyond 0
> considering their default value is -1? I think there is a
> similar bug filled as why should not lead or anchor become -1.
>
> https://bugs.openjdk.java.net/browse/JDK-4334792
>
> Regards,
>
> Pankaj Bansal
>
> *From:*Semyon Sadetsky
> *Sent:* Wednesday, January 31, 2018 12:17 PM
> *To:* Pankaj Bansal; Jayathirth D V;
> swing-dev at openjdk.java.net
> <mailto:swing-dev at openjdk.java.net>; Sergey Bylokhov; Prasanta
> Sadhukhan
> *Subject:* Re: <Swing Dev> [11] Review Request: JDK-6481195
> ListSelectionListener indicates events on model.addElement
> after model.clear()
>
> Hi Pankaj,
>
> When only the first list element deleted the anchor and lead
> selection indexes become -1 after your fix while should be 0.
>
> --Semyon
>
> On 01/30/2018 01:13 AM, Pankaj Bansal wrote:
>
> Hi Jay,
>
> I have made the small correction you suggested.
>
> webrev:
> http://cr.openjdk.java.net/~pbansal/6481195/webrev.02/
> <http://cr.openjdk.java.net/%7Epbansal/6481195/webrev.02/>
>
> Regards,
>
> Pankaj Bansal
>
> *From:* Jayathirth D V
> *Sent:* Tuesday, January 30, 2018 12:13 PM
> *To:* Pankaj Bansal; swing-dev at openjdk.java.net
> <mailto:swing-dev at openjdk.java.net>; Sergey Bylokhov;
> Semyon Sadetsky; Prasanta Sadhukhan
> *Subject:* RE: <Swing Dev> [11] Review Request:
> JDK-6481195 ListSelectionListener indicates events on
> model.addElement after model.clear()
>
> Hi Pankaj,
>
> Changes are fine.
>
> As mentioned in previous suggestion Java code convention
> for different statements like if () and for ()
> loops(http://www.oracle.com/technetwork/java/javase/documentation/codeconventions-142311.html)
> we have to maintain single space between if/for and
> braces. In test case this needs to be updated. There is no
> need for another webrev please update it before you push
> the change.
>
> Thanks,
>
> Jay
>
> *From:* Pankaj Bansal
> *Sent:* Monday, January 29, 2018 3:39 PM
> *To:* Jayathirth D V; swing-dev at openjdk.java.net
> <mailto:swing-dev at openjdk.java.net>; Sergey Bylokhov;
> Semyon Sadetsky; Prasanta Sadhukhan
> *Subject:* RE: <Swing Dev> [11] Review Request:
> JDK-6481195 ListSelectionListener indicates events on
> model.addElement after model.clear()
>
> Hi Jay,
>
> Thanks for the review.
>
> I have made the suggested changes.
>
> webrev:
> http://cr.openjdk.java.net/~pbansal/6481195/webrev.01/
> <http://cr.openjdk.java.net/%7Epbansal/6481195/webrev.01/>
>
> Regards,
>
> Pankaj Bansal
>
> *From:* Jayathirth D V
> *Sent:* Thursday, January 25, 2018 6:28 PM
> *To:* Pankaj Bansal; swing-dev at openjdk.java.net
> <mailto:swing-dev at openjdk.java.net>; Sergey Bylokhov;
> Semyon Sadetsky; Prasanta Sadhukhan
> *Subject:* RE: <Swing Dev> [11] Review Request:
> JDK-6481195 ListSelectionListener indicates events on
> model.addElement after model.clear()
>
> Hi Pankaj,
>
> Please find my inputs:
>
> Test case is working properly before and after changes.
>
> But I think we should make changes in test to actually
> detect all failure cases properly instead of throwing
> RuntimeException when we hit first failure:
>
> if (numberOfEvents > 2) {
>
> throw new RuntimeException("Wrong number of Events. " +
>
> "Expected: 2, Actual: " + numberOfEvents);
>
> }
>
> if (list.getLeadSelectionIndex() != -1) {
>
> throw new RuntimeException("Wrong Lead Index. " +
>
> "Expected: -1, Actual: " + list
>
> .getLeadSelectionIndex());
>
> }
>
> if (list.getAnchorSelectionIndex() != -1) {
>
> throw new RuntimeException("Wrong Anchor Index. " +
>
> "Expected: -1, Actual: " + list
>
> .getAnchorSelectionIndex());
>
> }
>
> Before your code change we just hit first check throw
> exception which will not cover all the required
> conditions(proper number of events, proper lead selection
> index & proper anchor selection index).
>
> We can update a boolean variable in each of 3 cases and
> then check them and throw relevant RuntimeException.
>
> Also in test case there are some indentation problems in
> for loop, if condition and expressions please update them
> also.
>
> Thanks,
>
> Jay
>
> *From:* Pankaj Bansal
> *Sent:* Monday, January 22, 2018 6:43 PM
> *To:* swing-dev at openjdk.java.net
> <mailto:swing-dev at openjdk.java.net>; Sergey Bylokhov;
> Semyon Sadetsky; Prasanta Sadhukhan
> *Subject:* <Swing Dev> [11] Review Request: JDK-6481195
> ListSelectionListener indicates events on model.addElement
> after model.clear()
>
> Hi All,
>
> Please review the fix for JDK 11.
>
> Bug:
>
> https://bugs.openjdk.java.net/browse/JDK-6481195
>
> Webrev:
>
> http://cr.openjdk.java.net/~pbansal/6481195/webrev.00/
> <http://cr.openjdk.java.net/%7Epbansal/6481195/webrev.00/>
>
> Issue:
>
> Invalid ListSelectionEvents are being fired, when the data
> is added after the clear() function has been called on
> DefaultListModel. This only happens when the index 0 was
> selected before calling clear.
>
> When the index 0 is selected on a JList and clear()
> function is called, anchor and lead are not being updated
> properly in DefaultSelectionModel as they are not being
> reset to -1 inside the removeIndexInterval() function.
> Because of this, when new elements are added in model, a
> ListSelectionEvent is fired from insertIndexInterval()
> function.
>
> Fix:
>
> The code inside removeIndexInterval function handles the
> case when 0 index is selected in special way. Removed that
> piece of code.
>
> Note:
>
> This also fixes
> https://bugs.openjdk.java.net/browse/JDK-4334792partially
> when removeAllElements or clear() is called on
> DefaultListModel.
>
> Regards
>
> Pankaj Bansal
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/swing-dev/attachments/20180202/7bc1ac26/attachment.html>
More information about the swing-dev
mailing list