On 10/09/18 15:09, Alan Bateman wrote:
On 10/09/2018 14:53, Chris Hegarty wrote:
On 10/09/18 10:22, vyom tewari wrote:
...
Hi Andre,
i will apply your patch run our all network tests. I will update once
i will have result.
The patch in question is in the description of
https://bugs.openjdk.java.net/browse/JDK-8210493, correct?
I'm posting it in this thread for reference [*].
The proposed change is in the NET_InetAddressToSockaddr
function, which is used in many places. I wonder is this
change should rather be confined to PlainDatagramSocketImpl
bind ( since this is the only reported issue arising from
limiting the addition of the scope to link-local ) ?
If there is a DatagramSocket or DatagramChannel bound to a multicast
Ok, we absolutely need the DatagramChannel implementation too.
address then it has be possible to connect to that address too.
Hmmm... I get bind, but am less sure about `connect`. We will
need to test.
So we
may need it for the connect methods - that should become obvious once
some sets up an environment and starts to extend the existing tests to
bind to IPv6 multicast addresses.
Maybe the native implementation of these methods can be
updated to accept a scope, rather than poking around in
the InetSocketAddress from JNI code.
-Chris.
