Paul,

This would need aggressive testing as locks do not cause memory
synchronization; this in itself can break code that does not synchronize or
use volatile variables.  I use Semaphores often; always with a plan to
force object synchronization.  For example: ByteBuffer position/limit etc
are not volatile and must be synchronized to ensure that the positioning is
correct.

Semaphore's current count is also non volatile; that is how it quickly
decides in the tryLock() phase.

To recap, it may work but should be tested extensively.

-Jon


On Mon, Mar 10, 2014 at 7:26 PM, Mondain <[email protected]> wrote:

> Is https://issues.apache.org/jira/browse/DIRMINA-934 dead or just
> forgotten? I'm still using a patched version in Red5 and would love to see
> it committed. Lastly, whats the current status of a 3.0 release? I'd like
> to get started porting the our I/O code to work with it.
>
> Regards,
> Paul
>
> --
> http://gregoire.org/
> http://code.google.com/p/red5/
>

Reply via email to