RFR: 8076613: gc/TestSmallHeap.java failed with OOME

Bengt Rutisson bengt.rutisson at oracle.com
Tue Jun 2 06:26:55 UTC 2015


Hi Kim,

On 01/06/15 22:01, Kim Barrett wrote:
> Please review this small test fix for a nightly test failure.
>
> The test attempts to verify that Java can be started with the
> documented minimum heap size of 2M. By default, such a small heap size
> will enable the UseCompressedOops option. However, there are nightly
> test configurations that forcibly disable compressed oops, and with
> such a configuration this test for small heap usability sometimes
> fails.
>
> Since it seems unreasonable to require small heap usage to work with
> the default compressed oop usage forcibly disabled, the fix for this
> test failure is to disable the test in such a configuration.
>
> CR:
> https://bugs.openjdk.java.net/browse/JDK-8076613

Thanks for putting so many details in the CR. Nice investigation.

The fix seems reasonable to me. However, the requirement to be able to 
start with a 2M heap does not seem that reasonable anymore in my mind. 
Maybe it is time that we update the documentation to something more 
reasonable.

As you probably know we already don't even respect the -Xmx2m setting 
due to page alignment and such. So, we should probably reconsider the 
minimum requirements. But that is definitely out of scope for this test fix.

Reviewed.

Bengt

>
> Webrev:
> http://cr.openjdk.java.net/~kbarrett/8076613/webrev.00/
>
> Testing:
> Local jtreg testing.
>



More information about the hotspot-gc-dev mailing list