RFR: 8035618 : Four api/org_omg/CORBA TCK tests fail under plugin only
Alan Bateman
Alan.Bateman at oracle.com
Thu Feb 27 09:49:38 UTC 2014
On 27/02/2014 09:38, Seán Coffey wrote:
>
> On 27/02/2014 07:45, Alan Bateman wrote:
>> On 26/02/2014 21:47, Seán Coffey wrote:
>>> Good points Alan. Changes uploaded here :
>>> http://cr.openjdk.java.net/~coffeys/webrev.8035618.v2/webrev/
>> This looks much better. Are you sure that pmLock is needed? Is there
>> any reason not to synchronize on the AppContext when adding the
>> PresentationManager key?
> I was conscious of possible contention coming on the AppContext object
> if syncing on it. I figured a static lock should work (given that we
> had zero synchronization in the old code)
Assuming AppContext is mostly for applets and JNLP applications then I
wouldn't expect it should be a problem. I also wouldn't expect there are
many cases with multiple applets in the same VM all using CORBA at the
same time. So what you have it okay but equally okay to drop pmLock and
use ac instead.
> Yes - I'll get the direct AppContext reference removed for JDK 9 fix
> via JavaAWTAccess but will need to re-examine the code to see if we
> can avoid sun.awt code.
That would be great.
-Alan.
More information about the core-libs-dev
mailing list