Re: [Httpunit-develop] Upload 1.7.1 to central maven repository - Was: Re: Minor Bug in class ServletUnitHttpResponse

[Wolfgang F. <wf...@bi...>]

Dear Kai,

revision 1101 has your fix.

Yours
  Wolfgang
Am 03.10.12 11:18, schrieb Kai:
> I solved the bug!
>
> The final step was an upgrade from JUnit 4.7 to JUnit 4.10.
>
> JUnit 4.7 does not order the execution of Rules and @Before-annnotated
> methods correctly. So, my fix lead to several wired failures in
> Test-Cases with @Before-annotated methods, when executed with JUnit
> 4.7, becaus the PseudoServer wasn't started, when it first was
> accessed in the @Before-Method.
>
> You find my fix in the attached patch.
>
> I'm runnig the test in an endless-loop now, to be sure, that there is
> no other race-condition-issue:
>
> while (mvn test); do echo date ; done
>
> I will report, if that exposes any further race-condition-issues.
> Otherwise, everything might be ready for an upload :)
>
>
> Regards kai
>
> Am 28.09.2012 17:05, schrieb Kai:
>> I thin, I'm one step closer to a solution!
>>
>> First of all, I found a simple way, to reproduce the problem on all my
>> machines. That was necessary, becaus the test-case
>> WebPageTest.testTitle() only freezes on one of my machines (my old solw
>> laptop). And lately, it even works there.
>>
>> The following simple line of shell-code reruns the test-case forever.
>> This leads to a freeze after some time on all my machines:
>>
>> while (mvn test -Dtest=WebPageTest); do echo date; done
>>
>> So, I hope, now you are able to reproduce the error that I am
>> witnessing :)
>>
>>
>> Because I believe, that the root of the error is a race-condition in the
>> test-case execution, my next step was, to isolate the single
>> test-excecutions even more. Therefor, I changed PseudoServerTestSupport
>> into an ExternalRule.
>>
>> See:
>> http://blog.mycila.com/2009/11/writing-your-own-junit-extensions-using.html
>>
>> and:
>> http://kentbeck.github.com/junit/javadoc/4.10/org/junit/rules/ExternalResource.html
>>
>>
>>
>>
>> The good news:
>> Since I've done that, WebPageTest never freezes. The simple loop is
>> running on my fastest machine for over an hour now. Before the change,
>> it took only some minutes to freeze the test-execution.
>>
>> The bad news:
>> Now a lot of other tests fail :/
>>
>> But I think, they are only failing, because they share a common
>> preconfigured instance of PseudoServer, which was configured in the
>> @BeforeClass-annotated method, that I removed in my quick hack, because
>> the PseudoServer-Instance is recreated for every test-run.
>> So one only needs two Base-Test-Classes. One for independent test-runs
>> (like now after my quick hack) and one in the "old" style with a single
>> shared instance of PseudoServer for the whole class.
>>
>> Unfortunatly, I'm lacking the time to implement these changes now.
>> But I will try to find some time next week, to implment the changes.
>>
>>
>> Another question is:
>> Perhaps all this just works around a race-condition-bug, which might be
>> located in PseudoServer. Another solution to the problem above might be,
>> to rework the implementation of PseudoServer. IMHO it might be better,
>> for example, to implement the listening threads with
>> ExecutorService.execute() and wait() for them when the server is teared
>> down...
>>
>>
>> Regards,
>>
>> kai
>>
>> Am 23.09.2012 12:45, schrieb Kai Moritz:
>>> Am 20.09.2012 13:28, schrieb Russell Gold:
>>>> Sounds as though at times PseudoServer has nothing to send. That's the
>>>> only way that should happen. But that's kind of odd. Is it the first
>>>> read for the page?
>>>>
>>> Yes, I think so.
>>> It is the first property of the WebResponse, which is checked in the
>>> Test WebPageTest.testTitle() on line 189:
>>> assertEquals("Title", "A Sample Page", simplePage.getTitle());
>>>
>>> Regards,
>>> kai
>>>
>>>
>>>> On Sep 20, 2012, at 4:38 AM, Kai wrote:
>>>>>>
>>>>>> Testing always stops at that test-case on my box, sleeping for ever.
>>>>>> 100% reproducable. That also happens, if I execute that test
>>>>>> standalone
>>>>>> with "mvn test -Dtest=WebPageTest".
>>>>>>
>>>>> I wasn't able to fix that error yet.
>>>>> But I can provide some information, which may help tracking it down:
>>>>>
>>>>> This error surely is caused by some kind of a race-condition.
>>>>>
>>>>> The test-execution only freezes in one Test: WebPageTest.testTitle(),
>>>>> when it tries to fetch the title for comparison in the line
>>>>> "assertEquals("Title", "A Sample Page", simplePage.getTitle());"
>>>>>
>>>>> I debuged my way into the call stack up to
>>>>> WebResponse.readFromStream():1157. There, a call to
>>>>> inputStream.read()
>>>>> is made:
>>>>> "count = inputStream.read( buffer, 0, Math.min( maxBytes,
>>>>> buffer.length ) );"
>>>>> That call never returns.
>>>>>
>>>>> It may, again, be a race-condition in PseudoServer.
>>>>> But this is only a guess.
>>>>>
>>>>> I will try to find out more later...
>>>
>>
>

-- 

BITPlan - smart solutions
Wolfgang Fahl
Pater-Delp-Str. 1, D-47877 Willich Schiefbahn
Tel. +49 2154 811-480, Fax +49 2154 811-481
Web: http://www.bitplan.de
BITPlan GmbH, Willich - HRB 6820 Krefeld, Steuer-Nr.: 10258040548, Geschäftsführer: Wolfgang Fahl