CTDB in master?

[Stefan (metze) Metzmacher]

Am 15.10.2013 16:15, schrieb Simo:
> On Tue, 2013-10-15 at 14:15 +0200, Michael Adam wrote:
>> On 2013-10-07 at 07:56 +0200, Stefan (metze) Metzmacher wrote:
>>> Am 06.10.2013 23:16, schrieb Andrew Bartlett:
>>>
>>>> https://gitweb.samba.org/?p=obnox/samba/samba-obnox.git;a=shortlog;h=refs/heads/master-ctdb
>>>>
>>>> It seems almost finished: What can I do to help?  Is there any
>>>
>>> I just need to know when I should do the import of the current ctdb master
>>> into samba master and make ctdb master read only.
>>
>> I was wondering wether the merge commit could be avoided and
>> a fast forward could be used instead: I doubt we will be able
>> to e.g. do proper bisects across that merge.
>>
>> What is more, the merge does not reference any officially
>> available code tree any more, due to the extra directory
>> level.  Or am I missing something here?
> 
> As much as I normally hate merges, I think this is one of those rare
> cases where a merge is the only thing that makes sense, even after a bit
> of history rewrite due to moving to a subdir.
> 
> It wouldn't make sense to rebase everything on top as older commits
> would have no working tests anyway so it is better to do bisects only

That's exactly why I used a merge.

Also we already some merge commit in the history,
e.g. a0130c662a40d6951fadae4bd9a6f7c230df05db

Git merges are a bit like talloc_reference(), you should try hard to
avoid them
but sometimes they're useful.

Also note 'git rebase -p' tries to recreate merge commits.

metze