samldb and associated patches

Kamen Mazdrashki kamenim at
Fri Sep 24 03:28:49 MDT 2010

Hi Matthias,

On Fri, Sep 24, 2010 at 08:59, Matthias Dieter Wallnöfer <mdw at>wrote:

> Hi Kamen,
> Kamen Mazdrashki wrote:
>> I think we should leave Python tests with just "show_deleted:1" control.
>> We are deleting a normal objects and thus the result should be either
>> Tombstone
>> or Deleted-object, but no Recycled object.
> I will drop this.
> Please, leave "dsscync" part of this commit, it should be ok.

>> The test checks only repl_meta_data.c behavior, not subtree_delete.
>> Do we have a test for subtree_delete module? I can't find one.
>> I think it is ok to go 'in' this way, but it is good for us to have
>> a subtree_delete tests.
> Isn't "test_delete" under enough?
> Yeah. It tests if repl_meta_data handles delete of leaf objects correctly.
I worried about interaction between subtree_delete and repl_meta_data
in case we delete not a leaf object - for instance Server object
under Servers DN. In this case we have a mixture of objects with
DISALLOW_MOVE_ON_DELETE flag set and objects w/o this
flag. Thus, just few of those objects are moved in Deleted Objects
Do you have time to write such a test?
If not, I can do it next wee presumably and ask you for review

>>    I've also not pushed the timeout change for the reasons Kamen
>>    indicated.
>> This commit really solves the problem temporary.
>> If you happen to browse a heavy javascript-driven by the time
>> you are running 'make test', you may hit this problem again :)
>> We either should implement this wait with several retries
>> or implement role transfer operation to be synchronous from
>> client-side point of view.
>> Dropped
> Thanks. I will try to come up with something to solve the
problem in more generic way.



More information about the samba-technical mailing list