RFR (S): 8025427 jstat tests fails on 32-bit platforms

[Staffan Larsen]

On 30 sep 2013, at 08:30, Erik Helin <erik.helin at oracle.com> wrote:

> Hi Staffan,
> 
> On 2013-09-27, Staffan Larsen wrote:
>> This change makes jstat output '-' instead of '\ufffd' when the result of a calculation is NaN. '\ufffd' is typically not displayable in a terminal. 
> 
> The changes to RowClosure.java looks good!

Thanks.

> 
> On 2013-09-27, Staffan Larsen wrote:
>> This change also updates the tests to allow '-' as the output int the CCS (Compressed Class Space Space - Percent Used) column from -gcutil and -gccause. It can be hard to see from the diff but what I have done is change the regexp for that column from:
>> 
>> [0-9]+\.[0-9]+
>> 
>> to:
>> 
>> ([0-9]+\.[0-9]+)|-
> 
> these tests were added to jdk/test/ProblemList.txt as part of [0]. Could
> you please remove the tests from jdk/test/ProblemList.txt as part of
> this change?

I'll do that.

> 
> The changes to the tests looks good.
> 
> On 2013-09-27, Staffan Larsen wrote:
>> Testing: Passes tests locally, currently running JPRT.
> 
> Please be aware of that the tests are only failing on 32-bit platforms.
> If you want to reproduce them easily locally, the fastest way I found
> was to add -XX:-UseCompressedClassPointers in the tests (they do not
> take the -vmoption argument to jtreg into account).

Yes, I found out... :-)

Thanks,
/Staffan

> 
> Thanks,
> Erik
> 
> [0]: http://hg.openjdk.java.net/jdk8/tl/jdk/rev/95f609fcb639.
> 
>> webrev: http://cr.openjdk.java.net/~sla/8025427/webrev.00/
>> bug: https://bugs.openjdk.java.net/browse/JDK-8025427
>> 
>> Thanks,
>> /Staffan
>