[PATCHES] GPO support for the AD DC itself

[David Mulder]

Ah, there's the bug. In libgpo/pygpo.c on line 207, I loaded the default
smb.conf (/etc/samba/smb.conf) into the lp. I should be pulling the
settings from the smb.conf specified on the command line.
Hard coding the correct smb.conf allows the test to succeed.

On 06/26/2017 02:16 PM, David Mulder via samba-technical wrote:
> looks like the test hangs, I suppose because the service dies. It dies
> in pdb_get_methods() where it throws a smb_panic with the message
> "pdb_get_methods: failed to get pdb methods for backend samba_dsdb".
> This is drilled down inside of create_local_nt_token(), so I think this
> is the same issue you mentioned.
>
> I'd try running in testenv, but my box doesn't have a gui, and I can't
> figure out how the screen code works (tried passing the SCREEN=1 param,
> but it never attaches to my screen session). I'll need to setup another
> build box.
>
> On 06/25/2017 03:44 PM, Garming Sam wrote:
>> It's supposed to 'hang', it sets up a fresh AD test environment enclosed
>> inside an xterm where you can run any tests manually. It will run
>> gpoupdate in the background much like a standard setup, but it appears
>> to fail.
>>
>>
>> On 24/06/17 07:07, David Mulder wrote:
>>> Hmm, the test just hangs for me.
>>>
>>>
>>> On 06/21/2017 05:49 PM, Garming Sam wrote:
>>>> I'm also seeing segfaults during make test due to create_local_nt_token
>>>> inside samba_gpoupdate when it is run. It appears that it tries to open
>>>> my local database (to little avail and may be the cause of the
>>>> segfault), not the one intended for the test environment. 
>>>>
>>>> In my case, I was running 'make testenv', and watching the output.
>>>>
>>>>
>>>> Cheers,
>>>>
>>>> Garming
>>>>
>>>> On 21/06/17 14:15, Garming Sam wrote:
>>>>> Compiling the patches, I noticed that you haven't rebased on master
>>>>> recently (at least from what you've given). There are some interface
>>>>> changes which mean that there is a TALLOC_CTX * being supplied into some
>>>>> functions.
>>>>>
>>>>> There is also a compiler warning in torture/gpo/apply.c:
>>>>>
>>>>> execv(cmd[0], (char * const *)&(cmd[1]))
>>>>>
>>>>> Instead of casting, I think you should use a discard_const_p.
>>>>>
>>>>>
>>>>> Cheers,
>>>>>
>>>>> Garming
>>>>>
>>>>> On 21/06/17 01:12, David Mulder wrote:
>>>>>>> In the patch 'libgpo: Add libgpo python bindings' you remove a return
>>>>>>> case in finalize_local_nt_token. Can you explain why? Where were you
>>>>>>> using it and why is this necessary?
>>>>>> Oh, yes. I believe that happens inside the gp_get_machine_token() call
>>>>>> in py_ads_get_gpo_list(). This call always fails for a computer object
>>>>>> that's a DC. I'm not certain why, but I can look into that more.
>>>>>>> On the whole, it looks much improved. But there needs to be some more
>>>>>>> comments (either in the code or the commit message) on what you're
>>>>>>> actually doing. For instance, I notice that you've introduced GPO
>>>>>>> unapply and it uses an xml log file. How is this file actually formatted
>>>>>>> and used? Similarly, in applying krb5 settings, which settings does it
>>>>>>> alter and how does it alter them? There needs to be some more high level
>>>>>>> documentation, both for other developers and for users.
>>>>>>> docs-xml/smbdotconf/domain/gpoupdatecommand.xml probably needs an
>>>>>>> extensive list of what is and is not being applied for example because
>>>>>>> it's not obvious at a glance.
>>>>>> I'll add some better documentation in gpoupdatecommand.xml.
>>>>>> I suppose the best place to document the new xml unapply log would be in
>>>>>> comments.
>>>>>> I'll add more comments in general to make all of it more clear.
>>>>>>

-- 
David Mulder
SUSE Labs Software Engineer - Samba
dmulder at suse.com
SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)