On Jul 23, 2008, at 10:31 AM, Jerome Velociter wrote:

>>
>> On Jul 23, 2008, at 9:18 AM, Jerome Velociter wrote:
>>
>>> Hello devs,
>>>
>>> This is a small wondering I am having while reviewing selenium tests
>>> for
>>> XWiki Watch. Quoting my last remark on XWATCH-182 :
>>>
>>> I am still not sure whether or not tests that needs Internet
>>> connectivity
>>> are acceptable (such as the tests that add feeds from distant
>>> sources as
>>> Google Blog Search, etc.)
>>
>> They're not acceptable and usually it's very easy to not use them.
>> What's your use case?
>
> This test method is not very mandatory : it adds a feed from the  
> Google
> blog search engine. The "feed adding" feature is already tested via  
> direct
> (local) URL, which is sufficient IMO.

Ok this is an interesting use case and is true and valid for all cloud  
services. The system then becomes the internet.

I think we could have these types of tests but we need to make them a  
special test type and not mix them with the others which are running  
in a controlled environment since they might be subject to more outages.

Thanks
-Vincent

>>> I tend to think it is acceptable, as the only
>>> case one would try to run the tests offline, he would have to do it
>>> explicitly (with mvn -o), and thus would understand that a test  
>>> method
>>> with a name like "testAddRSSFeedFromGoogleBlogSearch" could indeed
>>> fail.
>>
>> The tests should all run in a self contained environment with
>> *minimal* dependencies on the environment.
>>
>> False positives are the plague of tests as we're currently
>> experiencing. 61 continuum mails this morning. We really need to all
>> have a look and fix that as otherwise real problems are hidden and
>> we're loosing faith in our tests.
>>
>> Since today is a bug fixing day we should completely fix our  
>> Continuum
>> tests today at the very minimum.
>
> Good, I'm dedicated to help doing this :)
>
> Jérôme.
_______________________________________________
devs mailing list
[email protected]
http://lists.xwiki.org/mailman/listinfo/devs

Reply via email to