RFR: 8277922: Unable to click JCheckBox in JTable through Java Access Bridge [v3]

[Alexey Ivanov]

On Thu, 10 Mar 2022 15:10:42 GMT, Anton Litvinov <alitvinov at openjdk.org> wrote:

>> src/java.desktop/share/classes/javax/swing/JTable.java line 8417:
>> 
>>> 8415:                 if (ac != null) {
>>> 8416:                     return ac.getAccessibleAction();
>>> 8417:                 }
>> 
>> This is the fix for possible NPE, right?
>
> Yes, correct, this "if" condition is a fix for possible NPE, the method "javax.swing.JTable.AccessibleJTable.AccessibleJTableCell.getCurrentAccessibleContext()" may return "null" according to its documentation and to its code. I already took into account this possible NPE in the 1st and 2nd versions of the fix for this bug, and saw it amoral not to take it into account in the 3rd version of the fix, therefore I added this "if (ac != null) {" block in the 3rd fix version.

Got it! Thanks.

>> test/jdk/javax/accessibility/JTable/BooleanRendererHasAccessibleActionTest.java line 48:
>> 
>>> 46: import javax.swing.table.TableCellRenderer;
>>> 47: 
>>> 48: public class BooleanRendererHasAccessibleActionTest {
>> 
>> Would `BooleanRendererHasNoAccessibleActionTest` be a better name? After all, you test for *no* accessible action.
>
> For almost 10 years of fixing bugs in JDK, I have always given names to the regression tests to describe exactly the failing test scenario, rather then the expected and not failing behavior. So for me the test name "BooleanRendererHasNoAccessibleActionTest" would mean that the test should fail when BooleanRenderer does not have "AccessibleAction", and this is the opposite from what the bug and what I am fixing by the 3rd fix version. I am not going to change the test name.

Makes sense.

-------------

PR: https://git.openjdk.java.net/jdk/pull/7416