Code Review Request: 8017779: java/net/Authenticator/B4769350.java fails

Kurchi Hazra kurchi.subhra.hazra at oracle.com
Thu Jul 18 16:26:17 PDT 2013


Hi Michael,

    I added some comments as to what is the purpose of the latches and 
barriers. Those comments alongwith the
comments describing the purpose of the handlers should make the 
synchronization logic more clear. Let me know what
you think: http://cr.openjdk.java.net/~khazra/8017779/webrev.01/

Thanks,
Kurchi

On 7/17/2013 2:07 PM, Kurchi Hazra wrote:
>
> On 7/17/2013 12:27 AM, Michael McMahon wrote:
>> On 16/07/13 20:11, Kurchi Hazra wrote:
>>> Hi,
>>>      We have observed that test/java/net/Authenticator/B4769350.java 
>>> fails intermittently. Although not reproducible always,
>>> the bug could be in the test/sun/net/www/httptest library that this 
>>> test uses. I have rewritten the test to use
>>> com.sun.net.httpserver instead since we anyway want to move away 
>>> from using the httptest library.
>>>
>>> I have used CyclicBarriers to mimic TestHttpServer.rendezvous() and 
>>> CountDownLatches to
>>> mimic TestHttpServer.waitForCondition() and hopefully preserved the 
>>> rest of the logic in the test.
>>> I have not seen the test failing after these changes.
>>>
>>> Bug: http://bugs.sun.com/view_bug.do?bug_id=8017779
>>> Webrev: http://cr.openjdk.java.net/~khazra/8017779/webrev.00/
>>>
>>> Thanks,
>>> Kurchi
>>>
>>
>> Kurchi,
>>
>> Since this is a fairly complicated test, and it's great to see it 
>> being rewritten,
>> is there any possibility of adding some commentary that explains the 
>> purpose
>> of the synchronization code. For instance, I can't see the purpose of 
>> the call
>> on line 163 as it just blocks a thread that has already completed its 
>> work
>>
>> Michael
>
> Hi Michael,
>
>   I have just preserved whatever logic the test originally had. The 
> specific instance you point out waits
> for the T1b() handle to be executed for case 0 before moving forward. 
> But I guess it is hard to understand in a
> glance and I'll add some more comments and get back with a new webrev.
>
> Thanks,
> Kurchi

-- 
-Kurchi




More information about the net-dev mailing list