RFR(S) : 8199265 : java/util/Arrays/TimSortStackSize2.java fails with OOM
david.holmes at oracle.com
Wed Jun 27 06:03:19 UTC 2018
Using 2x max heap seems a reasonable approach.
AFAICT the other changes are just cleanup and don't make any difference
to the way the test is run - right?
Only nit: ProcessTools.executeTestJvm is deprecated in favor of
On 27/06/2018 11:14 AM, Igor Ignatyev wrote:
>> 10 lines changed: 0 ins; 2 del; 8 mod;
> Hi all
> could you please review another attempt to make TimSortStackSize2 more robust?
> 8190679 changed the test to set both Xmx and Xms to avoid failures caused by Xmx being smaller than Xms (when Xmx is specified as external vm flag). it appears that current values of heap size might not be enough in some cases, therefore setting Xmx equal to Xms leads to OOM.
> The proposed fix sets Xmx two times more than Xms, this should be enough for the test to pass and also ensures that Xmx is always bigger than Xms (since the test specific flags are appended, external Xms/Xmx flags will be ignored).
> JBS: https://bugs.openjdk.java.net/browse/JDK-8199265
> webrev: http://cr.openjdk.java.net/~iignatyev/8199265/webrev.00/
> testing: java/util/Arrays/TimSortStackSize2.java multiple times on different platforms
> -- Igor
More information about the hotspot-dev