8202252: (aio) Closed AsynchronousSocketChannel keeps completion handler alive

Alan Bateman Alan.Bateman at oracle.com
Tue Jul 3 07:33:37 UTC 2018


On 02/07/2018 19:55, Brian Burkhalter wrote:
>
> I changed it to use a ReferenceQueue per Roger’s suggestion:
>
> http://cr.openjdk.java.net/~bpb/8202252/webrev.03/ 
> <http://cr.openjdk.java.net/%7Ebpb/8202252/webrev.03/>
It's testing the right thing now and the switch to using a reference 
queue looks okay.

The bug may not have existed on Windows but I don't think we have other 
tests for this type of thing so my preference would be to drop the 
@requires os.family != "Windows".

One thing to double check in the test is whether it closes all the 
connections. In testConnect it doesn't close the client connection and 
also it seems to forget to accept and close the other end too. It looks 
to be correct in testRead and testWrite but the line breaks make it hard 
to read - maybe you could have slightly longer lines here to make this 
easier to read.

-Alan




More information about the nio-dev mailing list