RFR: 8171045: gc/stress/TestStressG1Humongous.java fails to allocate the heap

[Michail Chernov]

Hi Thomas,

Thank you for response!

I agree that 1G and 512M is just a minimum size of heap to make this 
test reasonable. I filled RFE for test enhancement to make it more adequate.

I have other approach for this failure fixing:
http://cr.openjdk.java.net/~mchernov/8171045/webrev.01/

I can use this fix, but If you're argue against this fix, I will use 
previous.

I filed RFE for improvement of this test:
https://bugs.openjdk.java.net/browse/JDK-8171512

Thanks,
Michail

On 20/12/2016 19:30, Thomas Schatzl wrote:
> Hi Michail,
>
> On Tue, 2016-12-20 at 16:59 +0300, Michail Chernov wrote:
>> Hi,
>>
>> Could I have a reviews for this small change for one test, please?
>>
>> http://cr.openjdk.java.net/~mchernov/8171045/webrev.00/
>> https://bugs.openjdk.java.net/browse/JDK-8171045
>>
>> Sometimes test failed with error:
>>
>> Error occurred during initialization of VM
>> Could not reserve enough space for 1048576KB object heap
>>
>> This happens on 32 bit systems. Now the test limits heap size to 512
>> M.
> Not sure if that significantly changes the tests. Running with 32M
> regions with 512M of heap does not seem to make a lot of sense, but
> previously it has only been 1g either.
>
> One option could be to make two tests out of the one, one for 32 bits,
> one for 64 bits. But the previous 1g heap has already been a random
> guess too, so I do not know.
>
> Other than that, and unless anybody has a better idea, it looks good.
>
> Thanks,
>    Thomas
>