Hi David, Yes, the failure happened because Graal/JVMCI compiler threads were returned by allThreads() and sometimes a new Graal thread got squeezed between allThreads() calls.
Best regards, Daniil On 3/11/19, 12:31 AM, "David Holmes" <[email protected]> wrote: On 6/03/2019 7:59 am, Daniil Titov wrote: > Please review a fix for this test that intermittently fails with Graal on. > > The problem here is that the test does two consequent calls to VirtualMachine.allThreads() and checks that the number of threads returned by these calls is the same. With Graal on there is a chance that a new JVMCI compiler thread could be started between these calls that makes test to fail. The fix temporary suspends the debuggee VM while these two calls to VirtualMachine.allThreads() are made. Is this because Graal/JVMCI compiler threads are returned by allThreads() while regular compiler threads are not? Suspending the VM seems safest regardless. Thanks, David ----- > Webrev: http://cr.openjdk.java.net/~dtitov/8218464/webrev.01 > Bug: https://bugs.openjdk.java.net/browse/JDK-8218464 > > Thanks! > --Daniil > > >
