Request for review: 8006629: NEED_TEST: need test for JDK-8001071

Tue Mar 19 10:31:39 PDT 2013

Hi all,

I'm still looking for a sponsor for this change. :)

webrev:
http://cr.openjdk.java.net/~kshefov/8001071/webrev.02/

reviews:
http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2013-March/005954.html
http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2013-February/005894.html

Thanks,
Filipp.

On 03/11/2013 07:02 PM, Gary Collins wrote:
> Hi Fillipp,
>
> I'm doing a bulk change for allot of tests.. I sent a webrev out last 
> week listing those.. See 8009152 RFR.
>
> I'll defer to Coleen for review of your current changeset.
>
> Gary
> On Mar 11, 2013, at 7:03 AM, Filipp Zhinkin wrote:
>
>> Gary,
>>
>> I suppose that your comment was about jtreg tests in general, but not 
>> about my particular test, which does not violate what you're wrote below.
>> If it is true, then I would like to ask someone to sponsor this change.
>>
>> Thanks,
>> Filipp.
>>
>> On 03/04/2013 10:16 PM, Gary Collins wrote:
>>> All these tests are a big mess.. I have a rather largish changeset 
>>> to hopefully add.. TESTVMOPTS TESTJAVACOPTS and make sure javac path 
>>> is used from
>>> a COMPILEJAVA path instead of TESTJAVA path.. When running profiles 
>>> the java under test is only a JRE.. thus no javac or jar,etc.. So, 
>>> we need
>>> to make sure the paths to tools are found and set correctly.. Allot 
>>> of the runtime tests have native components. In most tests we could 
>>> change ${BIT_TEST} to ${TESTVMOPTS}
>>> but appears that will not work for the most case.
>>>
>>>
>>>
>>> Gary
>>> On Mar 4, 2013, at 10:01 AM, Coleen Phillimore wrote:
>>>
>>>>
>>>> That's fine then. I was suggesting to investigate it but if it's 
>>>> too complicated, it's not worth doing.
>>>> Thanks,
>>>> Coleen
>>>>
>>>> On 03/04/2013 12:24 PM, Filipp Zhinkin wrote:
>>>>> I would like to take my words back. :)
>>>>>
>>>>> I want be able to run this test with any VM options passed to JTREG.
>>>>> While with .sh script the only one thing I need to do is just pass 
>>>>> value of TESTVMOPTS variable to JVM,
>>>>> with ProcessBuilder I'll have to correctly parse string with VM 
>>>>> options and it seems to be tricky.
>>>>>
>>>>> So I think that for one particular test it is unnecessary 
>>>>> complexity, but if there is a demand in such functionality
>>>>> in "testlibrary" it could be added there.
>>>>>
>>>>> Thanks,
>>>>> Filipp
>>>>>
>>>>> On 02/27/2013 03:16 PM, Filipp Zhinkin wrote:
>>>>>> Coleen,
>>>>>>
>>>>>> thank you for pointing my attention on it.
>>>>>> I can't see any reason not to use com/oracle/java/testlibrary and 
>>>>>> I agree that such implementation will be better.
>>>>>>
>>>>>> Thanks,
>>>>>> Filipp.
>>>>>>
>>>>>> On 02/26/2013 10:14 PM, Coleen Phillimore wrote:
>>>>>>>
>>>>>>> Filipp,
>>>>>>> Some new tests have been added that use process builder rather 
>>>>>>> than a .sh file. Can you use this instead?
>>>>>>> Otherwise, test looks good.
>>>>>>>
>>>>>>> Coleen
>>>>>>>
>>>>>>> On 02/26/2013 12:54 PM, Filipp Zhinkin wrote:
>>>>>>>> Vladimir,
>>>>>>>>
>>>>>>>> thank you!
>>>>>>>>
>>>>>>>> Filipp.
>>>>>>>>
>>>>>>>> On 02/26/2013 09:57 PM, Vladimir Kozlov wrote:
>>>>>>>>> Looks good.
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> Vladimir
>>>>>>>>>
>>>>>>>>> On 2/25/13 11:57 PM, Filipp Zhinkin wrote:
>>>>>>>>>> Hi all,
>>>>>>>>>>
>>>>>>>>>> I've updated test, so now it works correctly not only with 
>>>>>>>>>> fastdebug and
>>>>>>>>>> debug builds, but also with jvmg builds.
>>>>>>>>>> Also I've fixed issue with hs_err_pid*.log file detection 
>>>>>>>>>> (previously it
>>>>>>>>>> could use hs_err_pid*.mdmp instead of log file on windows).
>>>>>>>>>>
>>>>>>>>>> http://cr.openjdk.java.net/~kshefov/8001071/webrev.02/
>>>>>>>>>>
>>>>>>>>>> Thanks,
>>>>>>>>>> Filipp.
>>>>>>>>>>
>>>>>>>>>> On 01/31/2013 09:01 PM, Vladimir Kozlov wrote:
>>>>>>>>>>> On 1/31/13 5:34 AM, Filipp Zhinkin wrote:
>>>>>>>>>>>> Vladimir,
>>>>>>>>>>>>
>>>>>>>>>>>> thank you. I suppose that to run JPRT I have to be an 
>>>>>>>>>>>> OpenJDK committer?
>>>>>>>>>>>
>>>>>>>>>>> No, anyone can do test runs (-stree .) without commit.
>>>>>>>>>>>
>>>>>>>>>>> Vladimir
>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Thanks,
>>>>>>>>>>>> Filipp.
>>>>>>>>>>>>
>>>>>>>>>>>> On 01/31/2013 12:01 AM, Vladimir Kozlov wrote:
>>>>>>>>>>>>> Looks good. I would suggest to run it through JPRT to test 
>>>>>>>>>>>>> on all
>>>>>>>>>>>>> platforms: -rtests '*-*-*-runtime/8001071'
>>>>>>>>>>>>>
>>>>>>>>>>>>> Thanks,
>>>>>>>>>>>>> Vladimir
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 1/30/13 3:17 AM, Filipp Zhinkin wrote:
>>>>>>>>>>>>>> Here is an updated webrev:
>>>>>>>>>>>>>> http://cr.openjdk.java.net/~kshefov/8001071/webrev.01/
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> All code related to BIT_FLAG was removed and BIT_FLAG was 
>>>>>>>>>>>>>> replaced by
>>>>>>>>>>>>>> TESTVMOPTS.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thanks,
>>>>>>>>>>>>>> Filipp.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 01/23/2013 12:59 PM, Filipp Zhinkin wrote:
>>>>>>>>>>>>>>> Vladimir,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> thanks for reply!I'll fix it.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Filipp.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 01/22/2013 11:02 PM, Vladimir Kozlov wrote:
>>>>>>>>>>>>>>>> Filipp,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> You should never use BIT_FLAG! You should use 
>>>>>>>>>>>>>>>> ${TESTVMOPTS} to test
>>>>>>>>>>>>>>>> correct VM.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Vladimir
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On 1/21/13 3:19 AM, Filipp Zhinkin wrote:
>>>>>>>>>>>>>>>>> Hi all,
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Would someone review the following regression test please?
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Test checks that debug VM will crashed on assert if offset
>>>>>>>>>>>>>>>>> passed to Unsafe's access methods is bigger than 
>>>>>>>>>>>>>>>>> object size.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> To ensure that it is true, shell script execute 
>>>>>>>>>>>>>>>>> Test8001071
>>>>>>>>>>>>>>>>> which tries to get object from huge offset.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Only one Unsafe's method is checked because all other
>>>>>>>>>>>>>>>>> methods uses the same mechanism to get OOP from object
>>>>>>>>>>>>>>>>> using passed offset.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Test passes only if hs_err_pid* file created and it 
>>>>>>>>>>>>>>>>> contains
>>>>>>>>>>>>>>>>> tested assert. Test will also pass if product VM is 
>>>>>>>>>>>>>>>>> tested.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> http://cr.openjdk.java.net/~kshefov/8001071/webrev.00/
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Thanks,
>>>>>>>>>>>>>>>>> Filipp.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>