On Tue, 21 Feb 2023 18:59:35 GMT, Stuart Marks <sma...@openjdk.org> wrote:
>> I don't think all of those code paths ever executed. If `System.exit()` is >> called, regardless of the exit code, `make` treats it as an error: >> >> `TEST RESULT: Failed. Unexpected exit from test [exit code: 0]` >> >> I added a try/catch in `main` so that if an exception is thrown during the >> test, `System.exit` will be called. > > My bad, I was confused about the execution environments. In the old test, > jtreg invoked the shell script, which in turn invoked the Java test as a > standalone Java program. In that environment it's paramount that every code > path call System.exit() in order to avoid stale JVMs building up. > > The new test is invoked directly by jtreg, which disallows calling > System.exit(). Instead, every code path needs to result in the main() method > completing somehow, either via a normal return or by throwing an exception > directly or letting one propagate from some called method. So, you need to > pull out the try/catch and any calls to System.exit(). Sorry about that. > > If we're going to stick with a scheme where we're running separate JVM per > test, there's no need to unexport any RMI service. When main() returns or > throws an exception, jtreg will detect this and terminate the JVM because of > /othervm, and any exported RMI services will be destroyed at that time. Thus, > the finally-block in `testRmiCommunication` can simply be removed. I removed the finally block. ------------- PR: https://git.openjdk.org/jdk/pull/11910
