Re: format-patch signoff argument no longer works

[Junio C Hamano <junkio@cox.net>]

Seth Falcon <sethfalcon@gmail.com> writes:

> Matthias Kestenholz <lists@spinlock.ch> writes:
>
>> * Geoff Russell (geoffrey.russell@gmail.com) wrote:
>>> It appears to have got lost when the shell script got converted to C.
>>> 
>> Yes, this was intentional. You should sign off your changes while
>> committing (git commit -s|--signoff)

A bit on this later, but first to clear one thing up...

> When should one commit _without_ signoff?  
>
> The obvious answer is: when one doesn't approve of the changes in the
> commit... But in my usual workflow, commit means
> works-for-me-I-think-it-is-good. :-)

Please, calm down and read Documentation/SubmittingPatches,
item (6), to understand what sign-off means.  It does not have
anything to do with the result of the commit "working".  I do
not use -s when making commits during my day-job, for example.

We do not want to make sign-off the default.  It has to be a
concious act on the signer's part to add one.  Otherwise it
would not carry much weight.

About the droppage of "format-patch -s", I have come to think of
it as a mistake (yes, I can change my mind).  Consider:

 * You are the leader of a group of people who hack on a part of
   the kernel, internally in your company.  You and other
   developers make improvements and make commits, with "git
   commit -s".

 * As the in-company integrator, you maintain the canonical
   "company tree" by pulling from others in your group.

 * It's time to send good pieces to Linus and/or Andrew and as
   the group lead you are responsible for sending them out.  The
   commits would have Sign-off's by the original committers, but
   as the contact person (representative) of your group, your
   name is better recognizable in the outside community, and as
   the leader of your group, it is a good practice for _you_ to
   vouch for what your group did.

In that scenario, in addition to what "commit -s" gives us, it
is handy for the person who is sending the patches out via
e-mail to add his own sign-off.

Now, we could do that by re-adding "format-patch -s" option, or
alternatively we could add that to "send-email".  We might want
to do both ;-)