[jboss-as7-dev] TS: Bad approach used as fallback mechanism for gathering IP address

[Brian Stansberry]

On 3/6/12 7:58 AM, Pavel Janousek wrote:
> ----- Original Message -----
>> From: "Brian Stansberry"<brian.stansberry at redhat.com>
>>
>> On 3/2/12 8:04 AM, Stuart Douglas wrote:
>>>
>>> On 03/03/2012, at 12:31 AM, Pavel Janousek wrote:
>>>
>>>> Hi,
>>>>
>>>> I've investigated some issue now and seen there are several places
>>>> in TS where we can see similar code like this:
>>>>
>>>> static final String someIPproperty = System.getProperty("<some
>>>> key>", "localhost");
>>>>
>>>> or
>>>>
>>>> static final String someIPproperty = System.getProperty("<some
>>>> key>", "127.0.0.1");
>>>>
>>>> As you can see above, this fallback mechanism is bound in this
>>>> case (key value isn't found) to IPv4 only IP address at all. This
>>>> approach is bad because we can run in other network stack - IPv6
>>>> is presented in these days - and in this case the such testcase
>>>> fails with some strange error due to this issue.
>>>
>>> The fallbacks are there for cases when no other stack is specified.
>>> As of today the test suite should honour all setting specified by
>>> the node0 and node1 addresses. If you are running under another
>>> network stack and have configured the build properly then these
>>> fallbacks should never be hit, they will only be used in the
>>> normal mvn install case with not other addresses set.
>>>
>>> To be honest this sort of fallback should probably be implemented
>>> in the pom, so that the system properties are always defined. Feel
>>> free to volunteer to fix it.
>>>
>>
>> +1; it would be nice to know these are always set in the pom. Then
>> when
>> reviewing a patch, the presence of "localhost" or "127.0.0.1" is
>> grounds
>> for kicking it back, with no need to even think about it.
>
> So could be a conclusion from related topic - the JIRA about request to testsuite structure maintainer(s) to accommodate and guarantee this request/feature set-up in every place(module) plus clean-up testcases java code to remove this fallback mechanism from java source code at all?
>
> If so, I'll rewrite one already filled placeholder...
>

I'm not sure I follow what you're saying. I think we need a JIRA to make 
sure the appropriate properties are always set by all the testsuite 
module poms, and then probably several JIRAs to get rid of the various 
places that use System.getProperty("<some key>", "localhost"). Maybe a 
single JIRA for the latter if it's a small enough task that one person 
can do it.


-- 
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat