On 08/12/2015 05:45 AM, Vincent Ryan wrote:
On 11 Aug 2015, at 17:01, Sean Mullan <sean.mul...@oracle.com> wrote:
On 08/11/2015 09:14 AM, Vincent Ryan wrote:
The process exec for uname is only made when the test is run on Solaris SPARC
earlier than 11.2 but I take your point.
I’ll look at caching its value and re-using the util methods.
Actually thinking more about this, and I don't know for sure, but agentvm mode
may reuse VMs for different update versions of Solaris, so a cached value could
return the wrong value and cause unexpected results. So this is probably not
feasible.
OK I’ll leave the current code unchanged as there’s no significant benefit to
using the utility methods in jdk.testlibrary.Platform class either
since System.getProperty effectively caches the property values it reads.
Ok.
--Sean
--Sean
On 11 Aug 2015, at 13:28, Sean Mullan <sean.mul...@oracle.com> wrote:
There are bunch of methods for determining os.version, etc in
jdk/testlibrary/Platform that might be better to reuse.
Did you consider caching the value of "uname -v" so it can be reused by other
tests? It seems expensive to exec a process just to get that value every time one of
these tests is run.
--Sean
On 08/11/2015 07:34 AM, Vincent Ryan wrote:
Please review this change to omit several PKCS11 tests from test runs on
Solaris SPARC 11.1 and earlier.
The tests had been failing intermittently.
Webrev: http://cr.openjdk.java.net/~vinnie/8133318/webrev.00/
Bug: https://bugs.openjdk.java.net/browse/JDK-8133318
Thanks.
