RFR: 8024718: Metaspace performance counters and memory pools should report the same data

[David Holmes]

On 17/09/2013 9:48 PM, Erik Helin wrote:
> On 2013-09-13, David Holmes wrote:
>> Hi Erik,
>> Just looking at the tests ...
>
> Hi David,
> thanks for reviewing the tests!
>
> On 2013-09-13, David Holmes wrote:
>> Can you please add -XX:+UsePerfData to the @run lines of the new test.
>
> Done.

Thanks.

> On 2013-09-13, David Holmes wrote:
>> Have you discussed the change to InputArguments.contains?
>
> I'm the one who added the class InputArguments to testlibrary and I'm
> also the only one that have been using it so far when writing tests :)
>
> However, I have CC:d Christian and Igor on this email, if they have an
> opinion on this.
>
> On 2013-09-13, David Holmes wrote:
>> If we have any args that are substrings of other args then this change
>> will cause incorrect matching.
>
> I actually do not know if any flag is a substring of any other flag, but
> you are right that it could potentially be confusing.

There are many, many VM options that are substrings. A few examples:

PrintAssembly
PrintAssemblyOptions

PrintCFG
PrintCFG0
PrintCFG1

PrintGC
PrintGCDetails


> On 2013-09-13, David Holmes wrote:
>> Would it not be better to add a new function that checks the prefix
>> rather than the whole arg?
>
> Agree. I suggest the following two methods:
> - InputArguments.contains, works the same way as today
> - InputArguments.containsPrefix, checks if one of the arguments contains
>    the given prefix
>
> I've uploaded a new webrev at:
> http://cr.openjdk.java.net/~ehelin/8024718/webrev.01/
>
> What do you think?

That works for me - the InputArguments part that is.

Thanks,
David

> Thanks,
> Erik
>
>> Thanks,
>> David
>>
>> On 13/09/2013 7:46 AM, Erik Helin wrote:
>>> Hi all,
>>>
>>> this patch fixes some issues with memory pools and the performance
>>> counter reporting slightly different numbers for metaspace. I've also
>>> updated two jtreg tests as well as added a new one to test the change.
>>>
>>> InputArguments.java in the testlibrary is updated as well.
>>> InputArguments.contains now takes a prefix as a parameter instead of the
>>> exact name of a flag. This makes it possible to check if a flag expecting
>>> a value has been set. For example, -XX:MaxMetaspaceSize=100m can be
>>> checked by running InputArguments.contains("-XX:MaxMetaspaceSize").
>>>
>>> Webrev:
>>> http://cr.openjdk.java.net/~ehelin/8024718/webrev.00/
>>>
>>> Testing:
>>> - JPRT
>>> - hotspot/test/gc (including updated and new tests)
>>>
>>> Thanks,
>>> Erik
>>>