I see.
I would suggest posting the system property workaround now. Whether or
not a re-evaluation changes the settings, your workaround always works
because it overrides the default settings.
Thanks,
Joe
On 5/22/2013 2:53 PM, Robert Strout wrote:
We've encountered the problem with SPECjbb2005.
We can add the system property setting to our runs to move ahead,
but SPECjbb2005
is a public benchmark. Everyone who runs it will need to apply the
same workaround.
Admittedly, SPEC performance benchmarks are only used by limited
cross-section of
the community.
I understand you are using the jaxp.properties option. Is there any
chance you could update the benchmark itself, e.g setting the system
properties at the entry of the tests and clear at the exit?
That's not possible.
The benchmark is owned and distributed by the SPEC org
(http://www.spec.org/jbb2005/).
They would be the ones that would need to update the benchmark, which
is not likely to
happen given that it's replacement SPECjbb2013 has been recently
released. And yet,
users (like us) will continue to run and use SPECjbb2005 as a
performance measure.
We may be able to get SPEC to post the system property work-around on
their web-site.
I think that's the best that could happen. Before we push that, we'll
wait until the
re-evaluation is completed. If the settings change, it may change
what we do.
Thx!
--Resii
