Removal of two instances of synchronization on Boolean

[Andrew Hughes]

----- Original Message -----
> On 05/17/2012 03:38 PM, Adam Domurad wrote:
> > Hello all. Tiny patch here to remove two instances of
> > synchronization on
> > Boolean (considered bad practice as typically only two Boolean
> > objects
> > exist).
> > 
> > While I was not entirely sure what should be done about the Boolean
> > synchronization, Deepak's opinion was that the synchronization was
> > not
> > needed, so I submitted this small patch.
> 
> I could be wrong, but seems to me that read() and write() are running
> in
> different threads. Without some form of synchronization, writes from
> one
> thread may not be visible to the other thread at all. Making
> shuttingDown a 'volatile boolean' should be sufficient.
> 

Right.  I was thinking the same looking at the original patch just now.
The original synchronisation code is using the full weight of locking
to achieve the memory visibility effects than can also be achieved through
making the variable volatile.  As it's just a case of updating a flag in one
place and reading it elsewhere, this is sufficient.  For something more
complex, you'd need to look at java.util.concurrent.atomic.

> Cheers,
> Omair
> 

-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/)
Fingerprint = EC5A 1F5E C0AD 1D15 8F1F  8F91 3B96 A578 248B DC07