<i18n dev> <AWT Dev> [8] Review request for 6789984: JPasswordField can not receive keyboard input

[Naoto Sato]

Thank you for trying stopListening() out. Later I figured out that this 
method is only used when the input method instance switches the 
underlying input method engine from native OS's InputMethodAdapter 
instance to pure Java input method, so in this problem case it may not 
seem to apply. Let me know your further findings.

Naoto

On 11/7/12 8:11 AM, Anton Litvinov wrote:
> Hello Naoto,
>
> Yes, sure, I tried calling "stopListening()" method. But it does not
> lead to complete disabling of the input methods and appearance of the
> opportunity of entering a text into JPasswordField, when some
> composition started in JTextField before switch of focus from the last
> one to JPasswordField. Such situation happens, because "resetXIC()" is
> not called after "disableInputMethod()" inside "stopListening()" of
> "sun.awt.X11InputMethod" for the reason of unset flag "needResetXIC"
> after a call to "endComposition()". I think that I need to substitute a
> part of the fix calling to "endComposition()" and "disableInputMethod()"
> inside "disableNativeIM()" of "sun.awt.im.InputContext" for one call to
> "stopListening()". But before providing a next version of the fix for a
> review I will have to run regression tests connected with IM on
> platforms other than Linux and it will require some time.
>
> Thank you,
> Anton
>
> On 11/6/2012 11:21 PM, Naoto Sato wrote:
>> Hi Anton,
>>
>> In order for native input methods to not grab the native events, I
>> would guess that calling InputMethodAdapter.stopListening() would be
>> the right solution to me. Have you tried considering calling this?
>>
>> Naoto
>>
>> On 11/5/12 3:14 AM, Anton Litvinov wrote:
>>> Hello Naoto,
>>>
>>> Thank you for the review. A call to "endComposition()" was added to the
>>> shared code for resolution of another problem which was additionally
>>> observed. The problem consists in the fact that some not completely
>>> composed string inside JTextField can stay in composition state after
>>> the focus is moved from JTextField to JPasswordField. This not completed
>>> string marked with underscore disappears from JTextField at the moment
>>> when JTextField receives focus again. So I decided to make this string
>>> disappear when a component with disabled input methods, in current case
>>> JPasswordField, receives focus by a call to "endComposition()" method.
>>> Perhaps, it would be better to remove this call from the fix at all,
>>> because absence of end of text composition during switch of focus
>>> between component with enabled and disabled input methods can be another
>>> separate issue, would not it? But, on the other hand, since this fix
>>> brings the case with not composed string to a view, it should contain
>>> some solution which is specific to Unix platforms.
>>>
>>> Also a new comment concerning a root cause of the bug from my view point
>>> and a proposed solution was added to a page of the bug. Since it is not
>>> visible on "http://bugs.sun.com" yet, it can be seen on Oracle's
>>> internal bug tracking resource.
>>>
>>> Thank you,
>>> Anton
>>>
>>> On 11/2/2012 9:08 PM, Naoto Sato wrote:
>>>> Hi Anton,
>>>>
>>>> What's the reason for adding endComposition() in the shared code?
>>>> endComposition() is supposed to end the unfinished composed string
>>>> that is currently in composition, and the bug does not mention
>>>> anything regarding the composition. Also, it looks like the bug is
>>>> localized only to SCIM on Unix platforms, and calling resetXIC() looks
>>>> just enough in XIM handling code to me.
>>>>
>>>> Naoto
>>>>
>>>> On 11/2/12 5:43 AM, Anthony Petrov wrote:
>>>>> Hi Anton,
>>>>>
>>>>> I'm not an expert in IM code, but generally the changes look good
>>>>> to me.
>>>>> I'm also CC'ing i18n-dev@ to take a look at this fix.
>>>>>
>>>>> My only concern is that you're changing shared code here. Did you run
>>>>> any IM tests on other platforms to make sure no regressions are
>>>>> introduced?
>>>>>
>>>>> Also, could you please add a comment to the bug report explaining the
>>>>> root cause of the issue and how the proposed solution helps resolve
>>>>> it?
>>>>>
>>>>> --
>>>>> best regards,
>>>>> Anthony
>>>>>
>>>>> On 11/1/2012 8:40 PM, Anton Litvinov wrote:
>>>>>> Hello,
>>>>>>
>>>>>> Please review the following fix for a bug.
>>>>>>
>>>>>> Bug: http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6789984
>>>>>> Webrev: http://cr.openjdk.java.net/~alitvinov/6789984/webrev.00
>>>>>>
>>>>>> This bug consists in inability of JPasswordField to react on any
>>>>>> keyboard events under certain conditions on Linux OS. The webrev does
>>>>>> not contain any new "jtreg" test, because a particular Linux OS
>>>>>> distribution with additionally installed packages are required for
>>>>>> reproduction of the bug.
>>>>>>
>>>>>> Requirements to the environment:
>>>>>> 1. Oracle Enterprise Linux (OEL) 5.6 x86 64/Oracle Enterprise Linux
>>>>>> (OEL) 5.5 x86 64
>>>>>> 2. Package "scim-anthy-1.2.0-6.el5.x86_64" and its all dependent
>>>>>> packages installed on the OS.
>>>>>> 3. JDK 8.
>>>>>> 4. A simple Java Swing application, where a dialog with both
>>>>>> JTextField and JPasswordField is displayed.
>>>>>>
>>>>>> Thank you,
>>>>>> Anton
>>>>
>>