On Fri, 18 Dec 2020 22:33:11 GMT, Brent Christian <bchri...@openjdk.org> wrote:
> This change adds some extra test output for NativeLibraryTest, primarily via
> an update to the ForceGC utility class.
>
> It was observed that there was nothing preventing the Cleaner from cleaning
> the short-lived Object that ForceGC registers before
> await()/doit()/System.gc() is even called.
>
> The new 'o' reference is kept alive until FoceGC.await() has been called.
>
> We should find out a little more the next time NativeLibraryTest fails (or
> perhaps it won't fail anymore!)
This looks okay.
test/lib/jdk/test/lib/util/ForceGC.java line 49:
> 47: System.gc();
> 48: System.out.format("doit %d: gc %d%n", iter, i);
> 49: if (cleanerInvoked.await(1L, TimeUnit.SECONDS)) {
If the object that the test waits for garbage collected takes more GC cycles
than the object registered in ForceGC's cleaner, this will return immediately.
If you are concerned, maybe if the count down latch already becomes zero, it
should call System.sleep before the next System.gc().
-------------
Marked as reviewed by mchung (Reviewer).
PR: https://git.openjdk.java.net/jdk16/pull/53
