RFR(S): 8058891: serviceability/dcmd/CodelistTest.java - fails on all platforms

[Nils Eliasson]

On 2014-09-24 19:30, Vladimir Kozlov wrote:
> On 9/24/14 4:25 AM, Nils Eliasson wrote:
>> Hi Vladimir,
>>
>> On 2014-09-23 23:24, Vladimir Kozlov wrote:
>>> On 9/23/14 6:14 AM, Nils Eliasson wrote:
>>>> Hi,
>>>>
>>>> Please review this small fix. It contains:
>>>>
>>>> 1) Adding compiler dcmd tests to TEST.groups/hotspot_compiler_3
>>>
>>> I think they belong to hotspot_serviceability group. Please, move it 
>>> there.
>>
>> ok moved them.
>>
>>>
>>>> 2) Moving compiler dcmd tests to subfolder compiler and adding 
>>>> @library
>>>> tag to find and build DcmdUtil
>>>
>>> Was MethodIdentifierParser.java just moved and no change was done in 
>>> it?
>>
>> Yes
>>
>>>
>>>> 3) Fixing CompilerQueueTest to work in client, vintage server (no
>>>> tiered), and with interpreter only
>>>
>>> Could you explain 3) changes? I see only new throws which will 
>>> causes test failure.
>>
>> Are you sure you are looking at the right file?
>> http://cr.openjdk.java.net/~neliasso/8058891/webrev.03/test/serviceability/dcmd/compiler/CompilerQueueTest.java.udiff.html 
>>
>> The while loop will now match zero or many queue outputs.
>
> What is output with -Xint? Are you saying there will be no output and 
> as result the loop will not be executed (str == null)?

Yes

>
>>
>>>
>>> Should we also test interpreter mode in CodeCacheTest.java?:
>> Yes, I will add that. Also the CompilerQueueTest should be testen 
>> with interpreter since it failed there earlier. I will
>> add that too.
>>
>>>
>>> @run main/othervm -Xint -XX:+SegmentedCodeCache CodeCacheTest
>>>
>>>
>>>> 4) Preventing CodelistTest from trying to reflect MethodHandles
>>>
>>> Explain change.
>>
>> CodelistTest validates the printed methods by reflecting them. 
>> MethodHandle.invokeToStatic can not be reflected. The
>> tests skips all methods of MethodHandle now. This bug was not caught 
>> earlier because then we saw no methodHandles during
>> start up.
>
> Okay.
>
>>>
>>> Thanks,
>>> Vladimir
>>
>> New webrev: http://cr.openjdk.java.net/~neliasso/8058891/webrev.04/
>
> Looks good.
>
> Thanks,
> Vladimir
>

Thanks!
Nils