On Thu, 14 Jul 2022 12:54:43 GMT, Jaikiran Pai <j...@openjdk.org> wrote:
>> Roger Riggs has updated the pull request incrementally with one additional
>> commit since the last revision:
>>
>> Cleanup waiting for library unload
>
> test/jdk/java/lang/ClassLoader/loadLibraryUnload/libloadLibraryUnload.c line
> 48:
>
>> 46: if (ref == NULL) {
>> 47: // Only create a single GlobalRef
>> 48: ref = (*env)->NewGlobalRef(env, obj);
>
> Hello Roger, do we not use any synchronization here because even if both
> winning threads end up concurrently here, the `Object` instance they will be
> setting as a global ref will be the same one? Ofcourse, there will then be
> two calls to `NewGlobalRef` for the same instance. I am unaware what the
> semantics of that call will be in such cases and if it is of any concern.
Its a racy bug, each NewGlobalRef will have a reference to the object and only
the global ref in the last call to `setRef` will be cleared.
The easiest fix is to synchronize on `Class1.class` in Class1.loadLibrary().
-------------
PR: https://git.openjdk.org/jdk/pull/9474
