RFR (XS): 8009408: gc/metaspace/ClassMetaspaceSizeInJmapHeap.java fails with "exit code 1"

Christian Törnqvist christian.tornqvist at oracle.com
Tue Mar 12 05:14:02 PDT 2013


Hi Erik,

The change looks good :)

Thanks,
Christian

-----Original Message-----
From: Erik Helin 
Sent: den 11 mars 2013 16:26
To: Staffan Larsen
Cc: Bengt Rutisson; hotspot-gc-dev openjdk.java.net; Christian Törnqvist
Subject: Re: RFR (XS): 8009408: gc/metaspace/ClassMetaspaceSizeInJmapHeap.java fails with "exit code 1"

All,

I've updated the change quite a bit based on feedback from Bengt and Christian.

The class JmapLauncher has moved to the newly created gc testlibrary. 
This gc testlibrary, com.oracle.java.testlibrary.gc, is meant to contain utilities that can be used by all the gc jtreg tests. If there are code that is of interest for additional tests, then code can be easily be moved from the gc testlibrary to the general testlibrary.

I have updated the ClassMetaspaceSizeInJmapHeap.java to make use of this new gc testlibrary.

Please see new webrev at:
http://cr.openjdk.java.net/~ehelin/8009408/webrev.02/

Thanks,
Erik

On 03/08/2013 01:34 PM, Staffan Larsen wrote:
> Looks good to me.
>
> /Staffan
>
> On 8 mar 2013, at 13:36, Erik Helin <erik.helin at oracle.com> wrote:
>
>> Bengt and Staffan,
>>
>> thanks for he feedback!
>>
>> I've updated the change to make use of the method getPlatformSpecificVMArgs() and also abstracted the jmap command generation slightly.
>>
>> I don't know if this new little class, JMapLauncher, should be part of the testlibrary or not, or if we should put parts of it in the testlibrary. Christian, maybe you can comment on this?
>>
>> Please see new webrev at:
>> http://cr.openjdk.java.net/~ehelin/8009408/webrev.01/
>>
>> Thanks,
>> Erik
>>
>> On 03/06/2013 10:05 AM, Bengt Rutisson wrote:
>>>
>>> Erik and Staffan,
>>>
>>> The ProcessTools class has the method getPlatformSpecificVMArgs() 
>>> that returns "-d64" if necessary. You can't use this as is since you 
>>> need to get "J-d64" but I think we should do something to make your 
>>> solution more general.
>>>
>>> Either adding a possible prefix to the getPlatformSpecificVMArgs() 
>>> or adding a separate method that returns "JDK-tools formatted" args. 
>>> It seems a bit too limited to fix this only for your particular 
>>> test. Would be nice to get it in to the testlibrary somehow.
>>>
>>> Bengt
>>>
>>>
>>> On 3/5/13 5:55 PM, Staffan Larsen wrote:
>>>> Looks good. I wish we could abstract this away so that not every 
>>>> test needs to do this work.
>>>>
>>>> /Staffan (mobile)
>>>>
>>>> On 5 mar 2013, at 17:39, Erik Helin <erik.helin at oracle.com> wrote:
>>>>
>>>>> Hi all,
>>>>>
>>>>> this change adds the option "-J-d64" or "-J-d32" (depending on 
>>>>> arch) when running jmap in the test 
>>>>> hotspot/test/gc/metaspace/ClassMetaspaceSizeInJmapHeap.java.
>>>>>
>>>>> Webrev:
>>>>> http://cr.openjdk.java.net/~ehelin/8009408/webrev.00/
>>>>>
>>>>> Bug:
>>>>> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8009408
>>>>>
>>>>> Thanks,
>>>>> Erik
>>>
>>
>



More information about the hotspot-gc-dev mailing list