P4
I've extract this bug from bug 4476680; see that bug for more details on how to test & run. Basically, the bug is a form of runThese: 600+ JCK tests are started up and run in parallel.
Running with -Xprof reveals there's a race condition on thread creation and safepoints whereby a thread can get added to the threads list, but remain in state _thread_new forever when trying to bring the system to a safepoint. The system hangs. Basically, bug 4476680 starts a huge number of threads and -Xprof causes a huge number of safepoints and the combination of the 2 exposes the race condition.
James thinks it likely it could happen without -Xprof but we've not seen it.
Per James's request I filing this as a seperate bug.
cliff.click@eng 2001-07-24
Running with -Xprof reveals there's a race condition on thread creation and safepoints whereby a thread can get added to the threads list, but remain in state _thread_new forever when trying to bring the system to a safepoint. The system hangs. Basically, bug 4476680 starts a huge number of threads and -Xprof causes a huge number of safepoints and the combination of the 2 exposes the race condition.
James thinks it likely it could happen without -Xprof but we've not seen it.
Per James's request I filing this as a seperate bug.
cliff.click@eng 2001-07-24
- relates to
-
JDK-4476680 Two stress tests crash 64-bit HS VM with Error ID 4F530E43505002D7 01
-
- Closed
-