[U-Boot] [GIT PULL] Pull request: u-boot-staging

[Igor Grinberg]

Hi Wolfgang,

On 11/28/11 20:52, Wolfgang Denk wrote:
> Dear Igor Grinberg,
> 
> In message <4ECE39A9.4080201 at compulab.co.il> you wrote:
>>
>>>> As for my understanding, the delivery path ends with the repository
>>>> from which the pull process starts.
>>>
>>> That makes no sense either.  What about the case where the author
>>> provides (say, for convenience) a git repository where we can pull
>>> from, instead of applying the posting from the ML or PW?
>>
>> Well, I don't see how this is different from the custodian case,
>> except that contributor is not a custodian.
> 
> So would the custodian have to sign then, or not?  If yes, how would
> he do in a convenient way (git pull does not provide a "-s" option or
> similar to auto-sign all pulled commits [which would not make sense to
> me anyway])?  If no, then what is the difference between the git pull
> and - say - a git am ?

No, custodian should not sign it this way, because the patch has been
applied already and does not change (neither code, nor the meta data -
commit id and description) by the git pull command.

> 
>>> Hm... what's the difference then between pulling from a tree or
>>> cherry picking from it?  What about rebasing a tree?
>>
>> Cherry picking changes the commit id even if there are no conflicts
>> and can be done only on one commit at a time.
> 
> I cannot see what is special with a commit id in the context of
> signing a patch, or not.  My understanding is that the SCM is just an
> unrelated tool here, andw e could use SVN or even (ick!) CVS instead,
> and still the concept of Signed-off-by:s should and would work.

I understand your point here.
I see it a bit different though, signing is a part of "meta data"
of the commit as much as the commit id is, so AFAIK, the commit id
is the only unique identifier that should be and if you add the
signing information then the id must change.
That is also the reason, you don't have '-s' flag for git pull.

> 
>> Let me make it clear, I'm not arguing that one way is better
>> then another. I just thought that we are following the Linux
>> model, as there is no document describing, how this thing
>> should be done.
> 
> Is there a document that describes exactly how this is be done for
> Linux?  So far, all our conclusions appear to be derived from
> observations.

That is correct.
I've found the below link [1] with some correspondence involving Linus
about this stuff, but I don't think is really sheds any light on
the problem.

> 
>> Again, no rule - is kind of rule, but it would be nice to have
>> a document describing this decision, so less questions will arise.
> 
> Agreed. Does Linux have such a thing? [I could not find anything that
> goes into more detail than the SP file, which IMO is not really clear
> about that.]

Either me...
We could probably ask for some clarification on the
linux-kernel at vger.kernel.org to understand better the Linux model,
or can define something U-Boot specific.


[1] http://kerneltrap.org/node/3929

-- 
Regards,
Igor.