Alan,
Thank you for your feedback!
I hope Michael will find time to take a look at the fix.
I also have an updated webrev with fixes in regression test:
http://cr.openjdk.java.net/~kshefov/8143554/webrev.01/
<http://cr.openjdk.java.net/%7Ekshefov/8143554/webrev.01/>
Thank you,
Svetlana
On 08.12.2015 16:51, Alan Bateman wrote:
On 08/12/2015 13:44, Svetlana Nikandrova wrote:
Hi Alan,
thank you for your replay. Please let me explain a little.
For example ServerSocket and Socket have different supported options
set, but the same SocketImpl under the hood.
Yes, SocketImpl's setOptions() and getOptions() can be modified to
add additional check for the actual socket type that encloses that
socket implementation, but I believe this will intricate method's
logic and tangle the dependencies. I like it how it was done in jdk 8
- clean and simple - so I think it's a good idea to maintain that
approach in jdk 9 as well.
I think this needs a closer look and I hope Michael will have cycles.
As things currently stands then the SocketImpl methods are specified
to throw UOE if called with unsupported operations. Yes, this might
mean the implementation needs to know whether it's for a connecting or
listening socket.
-Alan