OK, that makes sense - short-lived programs with -Xcomp are dominated by cost of compiling everything.
On Fri, Jul 28, 2017 at 2:30 PM, Leonid Mesnik <leonid.mes...@oracle.com> wrote: > Hi > > Currently there were no timeouts wit fastdebug were observed. I don’t know > exactly how Xint affects this test but don’t expect significant > performance degradation comparing with other tests. Usually timeoutfactor > is used to increase timeouts of all tests if VM or host are slow. However > this test launches a lot of VMs with small applications and significantly > depends on JVM startup performance. I don’t think that other ‘slow’ modes > affect startup performance so significantly as Xcomp. > > Leonid > > On Jul 28, 2017, at 2:17 PM, Martin Buchholz <marti...@google.com> wrote: > > Won't this test fail with any "slow" VM, e.g. fastdebug, -Xint, etc ? > > > On Fri, Jul 28, 2017 at 2:05 PM, Leonid Mesnik <leonid.mes...@oracle.com> > wrote: > >> Hi >> >> Please review following small fix which excludes test >> tools/launcher/modules/illegalaccess/IllegalAccessTest.java from >> execution in Xcomp mode. Test launches about 100 VMs and fails when -Xcomp >> is used. >> >> Tested locally with and without adding -Xcomp option. >> >> Webrev: http://cr.openjdk.java.net/~lmesnik/8184775/webrev.00/ < >> http://cr.openjdk.java.net/~lmesnik/8184775/webrev.00/> >> Bug: https://bugs.openjdk.java.net/browse/JDK-8184775 < >> https://bugs.openjdk.java.net/browse/JDK-8184775> >> Diff: >> --- old/test/tools/launcher/modules/illegalaccess/IllegalAccessTest.java >> 2017-07-27 17:15:37.000000000 -0700 >> +++ new/test/tools/launcher/modules/illegalaccess/IllegalAccessTest.java >> 2017-07-27 17:15:37.000000000 -0700 >> @@ -23,6 +23,7 @@ >> >> /** >> * @test >> + * @requires vm.compMode != "Xcomp" >> * @modules java.base/jdk.internal.misc >> * java.base/sun.security.x509 >> * java.activation >> >> Leonid > > > >
