JDK-8154083 : RedefineAddPrivateMethod.sh timeout
  • Type: Bug
  • Component: hotspot
  • Sub-Component: jvmti
  • Affected Version: 9
  • Priority: P3
  • Status: Closed
  • Resolution: Incomplete
  • Submitted: 2016-04-12
  • Updated: 2016-05-27
  • Resolved: 2016-04-21
Related Reports
Relates :  
Description
result: Error. Program `sh' timed out (timeout set to 1920000ms, elapsed time was 1920133ms).

VM Started: > No frames on the current call stack

main[1] Deferring breakpoint shtest:3.
It will be set after the class is loaded.
main[1] > Set deferred breakpoint shtest:3

Breakpoint hit: "thread=main", shtest.main(), line=3 bci=0
3            System.out.println("@1 breakpoint");

main[1] main[1] Set breakpoint shtest:4
main[1] > @1 breakpoint

Breakpoint hit: "thread=main", shtest.main(), line=4 bci=8
4            System.out.println("@2 breakpoint");

main[1] @2 breakpoint
Timeout signalled after 1,920 seconds
Comments
Closing as incomplete.
21-04-2016

The test does not crash on newer JVMs where the bug has been fixed, it only crashes on older unfixed JVMs. I couldn't get any further in finding the problem, I believe thread dumps are needed to progress. It looks like there are similar issues with other com/sun/jdi tests (JDK-8004127), might be the same problem.
15-04-2016

I'm removing the integration_blocker label since this likely is a known issue or will be closed due to lack of information.
15-04-2016

Trying to investigate this issue is a challenge. The test has timed out but there are no thread dumps or similar to help understand what the test was doing at the time. Since the test is supposed to crash and the log shows no signs of that I'd assume that we are hanging while creating the core dump. There is a known issue where we deadlock in the VM while writing coredumps on crashes on Windows. This could be a case of that. There is no way to verify this unfortunately. If this is something else I don't see any way to investigate it further. Andreas, since you know the test maybe you can get further? Does the test have to create a core dump? We have modified most other tests not to do that due to the deadlock problem.
15-04-2016

Andreas, can you take a look at this 'integration_blocker'?
13-04-2016