I filed the following bug against this issue:
CR 7132378: Race in FutureTask if used with explicit set and get (
not Runnable )
It would appear to be a very specific issue confined to the FutureTask
implementation. I took a quick look at the latest FutureTask from Doug's
CVS and the changes (after a quick scan) would appear to be a candidate
for porting to a 7u update.
I think the best approach here would be to bring this specific change
into the JDK8 repository and let it bake there for a while. If there are
no problems we can then proceed with a request to backport to the 7u-dev
repo ( it will most likely be a candidate for 7u6 by this time ).
Sound reasonable?
As a matter of process, I'll follow up with a webrev for CR 7132378 (
the changes to sync up FutureTask ) on the core-lib-dev mailing list.
This is where most of the util_concurrent changes are reviewed.
Thanks,
-Chris.
On 22/01/2012 19:05, Jason Schroeder wrote:
For my information, would such a merge only reach jdk8? or would
other jdk versions receive the merge?
Thanks for the newer FutureTask.
Yes. This (and some related cases) were not originally intended
supported usages, but someone pointed out last spring that the spec
did not rule them out. So it is now explicitly covered in revised
base sources, that have been running successfully for months
(including in Android builds). We usually let changes settle for
a while before merging into openJDK . This should be among the
next ones to merge in.
