RFR 8237858: PlainSocketImpl.socketAccept() handles EINTR incorrectly

[Alan Bateman]

On 04/07/2020 15:44, Martin Buchholz wrote:
> On Fri, Jul 3, 2020 at 11:12 PM Alan Bateman <Alan.Bateman at oracle.com> wrote:
>
>> - "service" isn't a great name for the Executor. Also you can make use
>> of try-finally, e.g.
>> ExecutorService executor = Executors.newFixedThreadPool(1);
>> try { ... } finally { executor.shutdown(); }
> If you want to do this structured-concurrency-style, you should wait
> for all the threads you started to complete (they might block!).
> shutdown is not enough - you also want awaitTermination.
>
There is a future::get in the block so shutdown will be invoked when the 
task queue is empty, so it should lead to termination. However you are 
right that if the test throws before this then it may just shutdown 
while the "Server" task is still running. It could be make a lot more 
robust.

-Alan