RFR: 8306467: Fix nsk/jdb/kill/kill001 to work with new JVMTI StopThread support for virtual threads.

[Alex Menkov]

On Fri, 12 May 2023 20:58:31 GMT, Chris Plummer <cjplummer at openjdk.org> wrote:

> Currently kill001 assumes that JVMTI StopThread (via JDI ThreadReference.stop) is not supported for virtual threads. [JDK-8306034](https://bugs.openjdk.org/browse/JDK-8306034) is adding support for StopThread on a virtual thread as long as it is suspended and mounted. This means, for example, it will work for virtual threads in the following conditions:
>  - Debuggee in a loop and suspended
>  - Debuggee at a breakpoint and suspended
> 
> But will continue to not work in the following situations:
> - Debuggee in a loop but not suspended
> - Debuggee suspended but unmounted, such as during a call the Thread.sleep()
> 
> kill001 suspends all threads at a breakpoint and then does a "jdb kill" on each thread, which translate to `ThreadReference.stop()`, so this is expected to work now.
> 
> Most of the changes involve undoing the virtual thread specific code that was added to the test as part of [JDK-8282385](https://bugs.openjdk.org/browse/JDK-8282385). However, there is an additional issue that also needs fixing. The test relies on the fact that the async exception is normally not caught, and that jdb normally stops when an uncaught exception is thrown. With virtual threads there ends up being an exception handler in `java.lang.VirtualThread.run()`, resulting in jdb not stopping when the async exception is thrown. This is fixed by having the test issue a jdb "catch all <classname>" command for each async exception type that the test throws.

test/hotspot/jtreg/vmTestbase/nsk/jdb/kill/kill001/kill001a.java line 165:

> 163:         // and therefore sleep the full time, resulting in a test timeout if too long.
> 164:         try {
> 165:             Thread.currentThread().sleep(kill001a.vthreadMode ? 10000 : kill001a.waitTime);

The PR description says:
`kill001 suspends all threads at a breakpoint and then does a "jdb kill" on each thread, which translate to ThreadReference.stop(), so this is expected to work now.`
As far as I see only main thread is suspended at "breakHere", all test threads wait in Thread.sleep().
I'd expect test threads to be unmounted in Thread.sleep(), so it's unclear how the test works in vthread mode

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/13967#discussion_r1192859692