Vampire procedure fails

[Matthias Dieter Wallnöfer]

Andrew, do you still work on this? Or would you like to push your efforts?

Matthias

Andrew Bartlett wrote:
> On Wed, 2010-06-09 at 19:22 +1000, Andrew Bartlett wrote:
>    
>> On Wed, 2010-06-09 at 14:17 +1000, Andrew Bartlett wrote:
>>      
>>> On Tue, 2010-06-08 at 17:06 +0200, Matthias Dieter Wallnöfer wrote:
>>>        
>>>> Hi Zahari,
>>>>
>>>> well, every time when we add some stricter constraint checking something
>>>> might fail. Here it's exactly like this.
>>>> I really don't know much about the vampire process, so maybe someone can
>>>> help to fix this: on the normal provision we import the schema, load it
>>>> and only then the "objectclass_attrs" module does work. How it works on
>>>> a vampire? Could we already have a schema when we load another one?
>>>> If this is true and not changeable then the only fix would be to
>>>> deactivate the attribute constraint checking when the "relax" flag is
>>>> specified.
>>>>          
>>> Firstly, I would say that 'relax' is not the correct solution here.
>>> There is a full schema loaded at this time, that has been imported over
>>> DRSUAPI.  It is a little different to that which we load from the DB or
>>> LDIF, and this sometimes causes bugs (I've tried to resolve this, but
>>> it's quite difficult).
>>>
>>> We need to see what state the offending entry is in at the time of the
>>> failed check.  We should also check to see that the list of valid
>>> attributes has been expanded correctly into strings - the DRS protocol
>>> only gives them as integers, and this is the cause of many of the issues
>>> here.
>>>
>>> This is also why it's critical to get the vampire test into 'make test'.
>>>        
>> Attached is an untested idea, to use the schema from our local LDIF as
>> the bootstrap, rather than the hardcoded conversion of DRS ->  Schema
>> objects.
>>      
> Like a dog with a bone, I've not been able to let this idea rest.  When
> git.samba.org is back up, I'll push my continued efforts in this area.
> I've got some trouble importing the converted schema, but once I sort
> that we should be able to use the ldb based schema builder, even for the
> first schema import over DRS.
>
> Andrew Bartlett
>
>