Re: [PATCH]: natThread.cc: Destroy join sync objects in finalizer

[Bryce McKinlay <mckinlay at redhat dot com>]

Tom Tromey wrote:

> Bryce> Right, with my scheme above, no extra synchronization would be
> Bryce> needed. It would still be safe to synchronize on the native state
> Bryce> (join_mutex, etc), if thread->nt is non-null. The native state would
> Bryce> not be destroyed until its finializer is run, so if a thread sees it
> Bryce> as non-null, there is no danger of a race.
>
> Perhaps you need to synchronize when changing the value of thread->nt.
> I suppose there are other places where we assume pointer writes are
> atomic.
Arn't pointer writes atomic on all platforms we support?

Bryce