Re: [10] RFR JDK-8176072: READING attributes are not available on TSF

[semyon . sadetsky]

+1

--Semyon


On 11/17/17 4:35 AM, Sreeprakash Sreedharan wrote:

Hi Semyon,

Thanks for your inputs. You are right about the need for redirecting AwtInputTextInfor::GetAttributeInfor also to m_pResultTextInfor.

This scenario can happen when for the main AwtInputTextInfor (having the composition string in this case), the size of the current composition string is zero or if there is a mismatch in the size of the composition attribute and the composition string.

However, I have not yet been able to regenerate this scenario at all, as composition string being zero indirectly implies that WM_IME_COMPOSITION will not be send with GCS_COMPSTR.

This could probably explain why GetAttributeInfor was never redirected to m_pResultTextInfor and the results were only merged.

But there is still a theoretical possibility of this happening.

I have updated the webrev with this change.

Updated Webrev : http://cr.openjdk.java.net/~ssreedharan/8176072/jdk10/webrev.01/ <http://cr.openjdk.java.net/%7Essreedharan/8176072/jdk10/webrev.01/>

Thanks And Regards,

Sreeprakash

*From:*Semyon Sadetsky
*Sent:* Friday, November 17, 2017 12:22 AM

*To:* Sreeprakash Sreedharan <sreeprakas...@oracle.com>; awt-dev@openjdk.java.net; swing-...@openjdk.java.net *Subject:* Re: <Swing Dev> [10] RFR JDK-8176072: READING attributes are not available on TSF

Hi Sreeprakash,

Shouldn't the AwtInputTextInfor::GetAttributeInfor be readdressed to m_pResultTextInfor as well?

--Semyon

On 11/16/2017 08:36 AM, Sreeprakash Sreedharan wrote:

    Hello All,

    Please review the following fix in JDK10.

    Bug : https://bugs.openjdk.java.net/browse/JDK-8176072

    Webrev :
    http://cr.openjdk.java.net/~ssreedharan/8176072/jdk10/webrev.00/
    <http://cr.openjdk.java.net/%7Essreedharan/8176072/jdk10/webrev.00/>

    The bug describes a scenario wherein reading
    (AttributedCharacterIterator.Attribute.READING) attributes are not
    obtained from the InputMethodEvent text. (Japanese IME)

    (Detailed summary is provided in the JBS-Comment
    
<https://bugs.openjdk.java.net/browse/JDK-8176072?focusedCommentId=14131714&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14131714>)

    Issue:

    This issue is seen when IMM sends WM_IME_COMPOSITION with both
    GCS_COMPSTR and GCS_RESULTSTR. The scenario is handled in
    AwtInputTextInfor::GetContextData, where we extract the result
    string using the GCS_RESULTSTR data and store it in
    m_pResultTextInfor, member variable inside the main
    AwtInputTextInfor. Along with the result string the READING
    attributes of the first portion of the string is also stored in
    m_pResultTextInfor. The main AwtInputTextInfor stores the data
    represented by GCS_COMPSTR. When GetClauseInfor is called on the
    main AwtInputTextInfor, it just checks whether the read clause is
    present within it. If it’s not present, it just returns NULL even
    though the GCS_RESULTSTR data present within m_pResultTextInfor
    has READING attributes present.

    Solution :

    In GetClauseInfor, made sure that even if the main
    AwtInputTextInfor doesn't have any clause and Reading information
    it still checks for the GCS_RESULTSTR data held in m_pResultTextInfor.

    Also in WInputMethod.java, there was a condition which checks
    whether the last element in clauseBoundary is equal to length of
    the text entered. This is an incorrect assumption, as proven in
    the test case.

    Here, the last element of clauseBoundary comes as 2 (length of the
    Japanese String formed after typing 'a','b','e','space','space'),
    but the final text now has the new character 's' added and the
    total length will be 3.

    The test case added requires the user to change to Japanese IME
    and hence its manual.

    I have run all relevant JTREG test cases and Mach5 jobs and didn't
    see any failures.

    Regards,

    Sreeprakash