On Wed, 20 Sep 2023 20:36:20 GMT, Chris Plummer <cjplum...@openjdk.org> wrote:
> It seems like the "IPv6-only system" part of the test should include more
> test cases, both passing and failing. I'm not sure why it was lmitted to just
> the one:
I have added additional test cases
> Does this test guarantee that "system" was honored, or only verify that there
> was no failure to attach? In other words, wouldn't this pass even if the
> debug agent ignored the use of "system"?
That is true, with `preferIPv4Stack(true)` I don't think there's a decision to
make about ordering. Would you like this test case to be removed?
> What happens if using "system" with `preferIPv4Stack(false)`? I'm not sure
> why you don't have a test case for this.
I have added additional test cases.
> I guess maybe the reason is we don't know if the test case would pass or not
> because we don't know if "system" will result in using v4 (pass) or v6
> (fail). Is there a way to determine that in advanced?
Right, it depends on what `getaddrinfo` is doing on the current system. One way
to determine that in advance is to run the test with
`-Djava.net.preferIPv6Address=system` and check which address
`InetAddress.getAllByName("localhost")` returns first, and then assert that is
consistent with what JDWP is doing. I'm didn't see a more direct way to check
what `getaddrinfo` is going to do, or avoid relying on its behavior on the
current system.
> test/jdk/com/sun/jdi/JdwpNetProps.java line 82:
>
>> 80: // - listen on IPv4
>> 81: new ListenTest("localhost", ipv6Address)
>> 82: .preferIPv6Addresses("false")
>
> It seems we should also have a test case that omits using
> `preferIPv6Addresses()`. I think like this test case, it should fail because
> the debug agent will default to v4.
Done
-------------
PR Comment: https://git.openjdk.org/jdk/pull/15796#issuecomment-1728600933
PR Review Comment: https://git.openjdk.org/jdk/pull/15796#discussion_r1332312895
PR Review Comment: https://git.openjdk.org/jdk/pull/15796#discussion_r1332312949
