[Freeipa-devel] [PATCH 0022][Tests] Prevent trust test failures cause by adding duplicate DNS forward zone

[Martin Babinsky]

On 06/27/2016 10:28 AM, Petr Spacek wrote:
> On 27.6.2016 10:26, Petr Spacek wrote:
>> On 27.6.2016 10:18, Martin Babinsky wrote:
>>> On 06/27/2016 10:04 AM, Petr Vobornik wrote:
>>>> On 06/27/2016 09:42 AM, Lenka Doudova wrote:
>>>>> Hi!
>>>>>
>>>>> With newly created AD machines in Brno lab, existing trust tests fail on
>>>>> 'ipa dnsforwardzone-add' command claiming the zone is already present,
>>>>> as new AD domain is dom-221.idm.lab.eng.brq.redhat.com.
>>>>>
>>>>> To prevent these failures I prepared attached patch, that will still
>>>>> attempt to add the forward zone, but in case of non-zero return code
>>>>> will check the message if it says that the forward zone is already
>>>>> configured, and lets the tests continue, if it is so.
>>>>>
>>>>>
>>>>> Lenka
>>>>>
>>>>
>>>>
>>>> Current approach expects that every error of ipa dnsforward-add here
>>>> will mean that the zone exists. So it might hide other issues - not very
>>>> good.
>>>>
>>>> On the other hand it is not very robust to parse error message.
>>>>
>>>> Question for general audience: What do you think if IPA client's exit
>>>> status would be the IPA error code instead of "1" for every error. E.g.
>>>> in DuplicateEntry case it's 4002.
>>>>
>>>> Btw, this is not a NACK.
>>>>
>>>
>>> Well AFAIK the exit status on POSIX systems is encoded into a single byte so
>>> you cannot have the return value greater that 255. We would have to devise
>>> some mapping between our XMLRPC status codes and subprocess return codes.
>>>
>>> Some of our exceptions have defined return values outside plain '1', e.g.
>>> NotFound has return value of 2. It would be possible to extend this concept on
>>> other common errors.
>>
>> Even more importantly, the forward zone is completely unnecessary because DNS
>> when DNS is set up properly. I would simply remove the dnsforwardzone-add.
>>
> Grr, I meant this:
> Even more importantly, the forward zone is completely unnecessary when DNS is
> set up properly. I would simply remove the dnsforwardzone-add.
>
+1, our tests should not fiddle with the provisioned environment as much 
as they sometimes do.

-- 
Martin^3 Babinsky