Stephen Gallagher <sgallagh(a)redhat.com> wrote:
On 09/29/2010 09:09 AM, Jan Zelený wrote:
> Stephen Gallagher <sgallagh(a)redhat.com> wrote:
>> On 09/29/2010 07:24 AM, Jan Zelený wrote:
>>> Jan Zelený <jzeleny(a)redhat.com> wrote:
>>>> Jan Zelený <jzeleny(a)redhat.com> wrote:
>>>>> Based on Jakub's proof-of-concept patch I prepared a patch
which
>>>>> handles this option replacement. As it was discussed on the meeting
>>>>> couple weeks back, I didn't modify the update script (I hope I
>>>>> remember this requirement modification right) and in case both
>>>>> krb5_kdcip and krb5_server options are in the config file, the
latter
>>>>> one has bigger priority.
>>>>>
>>>>> Jan
>>>>
>>>> Sorry, wrong patch. Will send the most recent version later today.
>>>
>>> Here it is.
>>>
>>> Jan
>>
>> Nack.
>>
>> You can't remove krb5_kdcip from SSSDConfig API. It will break existing
>> configurations if they run it through authconfig. (SSSDConfig API will
>> throw an exception on the unknown value).
>>
>> If I'm reading this right, a user is going to see the deprecation error
>> message if they have neither krb5_kdcip nor krb5_server specified. This
>> is a bug, since if neither is specified, we use service discovery.
>>
>> confdb_get_string will return EOK and set krb5_servers to NULL if
>> krb5_kdcip is not found. You should handle this case accordingly.
>
> I'm sending an updated version. I left SSSDConfigTest.py unchanged, it
> should do the testing on the new values.
Nack. Please update SSSDConfigTest.py so that the tests do not fail.
I'm not 100% sure my approach is the right one, but I've corrected the script
and it doesn't fail any more.
Jan