8241336: Some java.net tests failed with NoRouteToHostException on MacOS with special network configuration

[Daniel Fuchs]

Hi Alan.

On 26/03/2020 17:29, Alan Bateman wrote:
> NetworkConfiguration.isTestable - is the hardcoded check for "awdl" 
> needed? I don't think I've seen it with non-link local address so I 
> can't tell if it is needed. I'm

It might no longer be needed, but we know that awdl is always
the wrong choice, so excluding it right away won't harm.
That was part of Chris's fix for 7155591 [1].

I've applied your other comments:

http://cr.openjdk.java.net/~dfuchs/webrev_8241336/webrev.02/

best regards,

-- daniel

[1] 
http://cr.openjdk.java.net/~msheppar/7155591/webrev/test/java/net/MulticastSocket/SetOutgoingIf.java.frames.html 


> Promiscuous - can "addr" be renamed to toSocketAddress so the usages are 
> a bit clearer? >
> SetOutgoingIf - the initializer at L38 may not be obvious to some 
> readers that aren't used to that syntax, can it use a regular 
> constructor instead?
> 
> AdaptorMulticasting - I don't mind if it included here or separated into 
> its own issue as the changes aren't needed for JDK-8241336. If you are 
> including it then the patch should be cleaned up a bit so that it is 
> consistent with the existing code. It was an oversight (by me) when I 
> added this test. testSendNoReceive should have just ignored datagrams 
> that are sent by an unexpected sender or with unexpected bytes. I think 
> it complicates things to try to decode the bytes of a random datagram 
> and prefer to see that part removed. It's okay to lot a message that an 
> expected message has been received but its content it not interesting 
> for the test. One thing that we should do some time is change some of 
> the multicasting tests to use other multicast groups. I suspect some of 
> the tests have copied the multicast address from other tests so we end 
> up with several tests using the same group.
>