<Swing Dev> [12] Review Request: 8199072 Test javax/swing/GroupLayout/6613904/bug6613904.java is unstable

[Sergey Bylokhov]

Hi, Prasanta.

I did not add the new bugid for cleanup intentionally, because this test
is not a "regression test" for the current bug.

http://openjdk.java.net/jtreg/faq.html#when-should-i-update-the-bug-entry-in-a-test-description

It is quite useful to have the short list of bugs,
against which you need to verify the new change/fix
in the testcase(the updated test should fail before the
fix for "@bugid" and pass after).


On 08/11/2018 20:24, Prasanta Sadhukhan wrote:
> looks ok. Please add the bugid to the test.
> 
> Regards
> Prasanta
> On 09-Nov-18 5:28 AM, Sergey Bylokhov wrote:
>> Hello.
>> Please review the fix for jdk 12.
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8199072
>> Webrev: http://cr.openjdk.java.net/~serb/8199072/webrev.00
>>
>> Bug was found when our tests executed in the "agentvm" mode.
>> The reason of test failure was the jtreg not able to interrupt EDT,
>> on which the test was executing.
>>
>> The root cause is that the test uses "invokeLater()", which do
>> not wait when the runnable will be completed. And since jtreg will
>> assumes that completion of the main() is the end of the test it will
>> try to stop EDT.
>>
>> I have verified that if the fix for JDK-8199022(which
>> added "javax/swing" to the "othervm.dirs" in the TEST.ROOT) is reverted,
>> the test will start to fail again.
>>
>> Even if the test currently always passes, we have an opposite issue,
>> the jtreg may report "PASS" status for the test even before the "Runnable"
>> will start to execute.
>>
>> As a fix invokeLater() is replaced by the invokeAndWait().
>>
>>
> 


-- 
Best regards, Sergey.