-----Original Message-----
From: net-dev <net-dev-r...@openjdk.java.net> On Behalf Of Michael McMahon
Sent: Friday, 28 January 2022 16:14
To: net-dev@openjdk.java.net
Subject: Re: RFR: JDK-8280498: [aix]: jdk/java/net/Inet4Address/PingThis.java
fails
On Fri, 28 Jan 2022 14:46:32 GMT, Michael Felt <d...@openjdk.java.net> wrote:
>> If it defeats the purpose, then it needs to be skipped.
>>
>> * When I was trying to understand the test, it seemed to be that it
>> _assumed_ that "0.0.0.0" was 'converted' to 127.0.0.1.
>> * If there is an international standard (ISO, POSIX) that states 0.0.0.0
>> needs to be treated as 127.0.0.1 - AIX clearly has a bug. Same for `::)`
>> mapping to `::1` - AIX doesn't do that IPv4 (now), and, afaik, never has for
>> IPv6.
>> * While testing I also saw that AIX reacts differently (error message iirc)
>> when passed NULL, whereas it behaves "as expected" when given a null-string
>> `""`
>> * FYI: Adoptium has put this test on the excluded list, but if a fix using
>> `""` rather than "0.0.0.0" and/or "::1" rather than "::0", the test should
>> perhaps state the exclusion as a comment in the code, and return
>> immediately, as it does for Windows.
>
>> Or is it possible to change the implementation on AIX so the test passes
>> without change?
>
> * Digging into the java guts to map "0.0.0.0" to "127.0.0.1" seems too far to
> me - as I believe interfaces are not suppossed to be "translating" as well.
> * If the behavior "0.0.0.0" becomes "127.0.0.1" is an official standard - a
> bug needs to be filed with IBM (which I cannot do) - but we need to also be
> aware that any fix to AIX is not likely to ever be applied on AIX 7.1 TL4
> (maybe TL5) which are systems used to build distributions.
> * Again - being a noob (or fresh face; fresh meat) the test title is merely
> "PingThis" - where is "This" defined as "0.0.0.0" (if it had been named
> _PingHERE_ I could understand as the alias, iirc, for 0.0.0.0 is "Here".
Might be better to skip the test then by adding to the header:
*
* @requires os.family != "aix"
*
rather than the check in the test source
And maybe remove this:
```
if (System.getProperty("os.name").startsWith("Windows")) {
return;
}
```
And also state in header something similar for Windows.
a) consistency
b) a change is required to be pushed (and back-ported :please: either way
-------------
PR: https://git.openjdk.java.net/jdk/pull/7013
