[cc'ing net-dev]

Alan Bateman wrote:
Pavel Tisnovsky wrote:
Hi,

two-lines patch for regression test java/net/MulticastSocket/SetOutgoingIf.java is exposed at http://cr.openjdk.java.net/~ptisnovs/JTReg_SetOutgoinIf/

The test machine (where this test fails) has six network interfaces, but only two of them, namely eth2 and virbr0, can be used to send multicast packets. But the test does not check interface attributes and tries to use _ALL_ network interfaces (excluding loopback device, of course) to send and then receive multicast packets.

It fails, but the failure message depends on current network interface order (order is not guaranteed) returned by NetworkInterface.getNetworkInterfaces(). Patch check, if the given
network interface is able to send multicast packets.

I think, the patch could be applied to OpenJDK6 too.

Pavel Tisnovsky
Pavel - this looks good me, although I think you should send it to net-dev. To make it more reliable you could also check isUp as getNetworkInterface may enumerate network interfaces that are down.

I agree with Alan here, it is best to check isUp also.

Alan created the following bug for this:
6884837: "test/java/net/MulticastSocket/SetOutgoingIf.java should check if interfaces are up/support multicast"

I see you have push access, so if you make the appropriate changes (mentioned above) I can review the webrev and you can use the above bug number and description to push the changeset.

-Chris.

P.S. I wonder if we have a similar issue with some of the other multicasting tests?


-Alan.


Reply via email to