Re: [PATCH RFC] git-am: support any number of signatures

["Michael S. Tsirkin" <mst@redhat.com>]

On Fri, Jun 13, 2014 at 10:32:09AM -0700, Junio C Hamano wrote:
> "Michael S. Tsirkin" <mst@redhat.com> writes:
> 
> > On Thu, Jun 12, 2014 at 12:07:03PM -0700, Junio C Hamano wrote:
> >> "Michael S. Tsirkin" <mst@redhat.com> writes:
> >> ...
> >> > 1.  new parameter am.signoff can be used any number
> >> > 	of times:
> >> >
> >> > [am]
> >> > 	signoff = "Reviewed-by: Michael S. Tsirkin <mst@redhat.com>"
> >> > 	signoff = "Signed-off-by: Michael S. Tsirkin <mst@redhat.com>"
> >> >
> >> > 	if set all signatures are picked up when git am -s is used.
> >> 
> >> How does this interact with the logic to avoid appending the same
> >> Signed-off-by: line as the last one the incoming message already
> >> has?
> >
> > Not handled if you have multiple signatures.
> > That will have to be fixed.
> > Do we only care about the last line?
> >
> > Signed-off-by: A
> > Signed-off-by: B
> >
> > do we want to add
> >
> > Signed-off-by: A
> >
> > or would it be better to replace with
> > Signed-off-by: B
> > Signed-off-by: A
> >
> > ?
> >
> > Current git am will add A twice, I wonder if this is
> > a feature or a bug.
> 
> This is very much deliberate.
> 
> Appending A after existing A and B is meant to record that the patch
> originated from A, passed thru B possibly with changes by B, came
> back to A who wants to assert that the result is still under DCO.
> 
> The only case we can safely omit appending A's sign-off is when the
> last one in the chain is by A.  Imagine that you had a patch signed
> off by B, which A may have tweaked and forwarded under DCO with A's
> sign-off.  Such a patch would have sign-off chain B-A.
> 
> Now A makes further changes to the patch and says "the further
> change is also something I am authorized to release as open source"
> with the "-s" option or some other way.  It would not change that A
> can contribute under DCO if we did not add an extra A after existing
> B-A sign-off chain in that case.

OK imagine we have signatures:
A
B

Now A wants to sign this patch.

I think there are two reasonable ways to behave:
1. What you describe above:
A
B
A

2. For things like Tested-by: tags, removing tag from
where it was and adding it at the bottom:

B
A


This probably calls for a separate feature:
maybe adding "acks" along with "signoffs"?
acks would be unique, re-adding ack removes it from
the message and adds at the bottom.

-- 
MST