On Thu, 21 Jan 2021 21:17:53 GMT, Dmitry Markov <dmar...@openjdk.org> wrote:
>>> > > Now we will commit the composition string if there is an active client. >>> > > That changes eliminates the issue described in JDK-8258805. Also the >>> > > behaviour of AWTTextTest1 is the same as it used to be on the previous >>> > > versions w/o the fix. >>> > >>> > >>> > Just to confirm: the updated behaviour is similar to what was happening >>> > in previous versions of Windows 10, that is if a compositing text isn't >>> > committed, it remains uncommitted in the text component when the focus >>> > changes instead of being committed as it was the case in the previous >>> > iteration. Is my understanding correct? >>> >>> I am sorry but you didn't get it right. The behaviour, which was introduced >>> in previous iteration, is still in place. It is required to get rid of the >>> problem described in JDK-8258805 >>> The purpose of of the second iteration is to eliminate negative side effect >>> (more details here [#2142 >>> (comment)](https://github.com/openjdk/jdk/pull/2142#issuecomment-763340186) >>> ) introduced by the first version of the fix. >> >> I admit I am even more confused now. To me, the description in the comment >> above is nearly the same as in [JBS >> comment](https://bugs.openjdk.java.net/browse/JDK-8258805?focusedCommentId=14391025&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14391025). >> Is the difference that the original test case disabled IME for the middle >> JTextField whereas in the test case above all JTextField support IME? >> >> In the updated version of the fix, we always commit the text on any focus >> change whether the newly focused component supports IME or not. > >> I admit I am even more confused now. To me, the description in the comment >> above is nearly the same as in [JBS >> comment](https://bugs.openjdk.java.net/browse/JDK-8258805?focusedCommentId=14391025&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14391025). >> Is the difference that the original test case disabled IME for the middle >> JTextField whereas in the test case above all JTextField support IME? > > Well.. I think the main difference between tests is that the [test attached > to the > bug](https://bugs.openjdk.java.net/browse/JDK-8258805?focusedCommentId=14391025&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-14391025) > uses `JTextField` (Swing) and the [test provided > above](https://github.com/openjdk/jdk/pull/2142#issuecomment-763491615) uses > `TextField` (AWT). The same input method events are processed differently for > Swing and AWT text components. Good example is the following test: > > import java.awt.*; > import java.awt.event.*; > > public class AWTTextTest1 extends Frame { > AWTTextTest1() { > setTitle("AWTTextTest1"); > setLayout(new GridLayout(0, 1)); > add(new TextField()); > add(new TextField()); > addWindowListener(new WindowAdapter() { > public void windowClosing(WindowEvent we) { > System.exit(0); > } > }); > setSize(400, 300); > setVisible(true); > } > public static void main(String[] args) { > new AWTTextTest1(); > } > } > > 1. Run test (originally it uses `TextField`) > 2. Click upper `TextField`, turn on IME, type some character (In case of > Japanese, type "aiu") > 3. Click lower `TextField`, the string is canceled. > 4. Replace `TextField` with `JTextField` in the test. Compile and run it > again. > 5. Click upper `JTextField`, turn on IME, type some character (In case of > Japanese, type "aiu") > 6. Click lower `JTextField`, the string is committed before focus transition. > >> >> In the updated version of the fix, we always commit the text on any focus >> change whether the newly focused component supports IME or not. > > That’s not quite right. Actually we commit the text if the current IM client > is “active client”. For example, `JTextField` is an “active client” but > `TextField` - NOT. The status “active client” depends on the implementation > of getInputMethodRequests() method. Hi, AWT's `TextComponent` is a `peered` input client, and Swing's `JTextComponent` is an `active` input client. Thus it is OK to behave differently. I would expect that AWT's one behaves as the same as native Windows apps, and Swing's one should commit text into the component that has lost the focus. ------------- PR: https://git.openjdk.java.net/jdk/pull/2142
