Re: [PATCH 0/2] git-am: add --message-id/--no-message-id options

Paolo Bonzini [off-list ref] writes:

From: Paolo Bonzini <pbonzini@redhat.com>

This series adds a --message-id option to git-mailinfo and git-am.
git-am also gets an am.messageid configuration key to set the default,
and a --no-message-id option to override the configuration key.
(I'm not sure of the usefulness of a mailinfo.messageid option, so
I left it out; this follows the example of -k instead of --scissors).

This option can be useful in order to associate commit messages with
mailing list discussions.

If both --message-id and -s are specified, the Signed-off-by goes
last.  This is coming out more or less naturally out of the git-am
implementation, but is also tested in t4150-am.sh.

Nice.  So if you apply a message whose last sign-off is yourself
with both of these options, what would we see?

    1. S-o-b: you and then M-id: and then another S-o-b: you?
    2. M-id: and then S-o-b: you?
    3. S-o-b: you and then M-id:?

I do not offhand know which one of the above possibilities to favor
more over others myself.  Just asking to find out more about the
thinking behind the design.

Thanks.

Paolo Bonzini (2):
  git-mailinfo: add --message-id
  git-am: add --message-id/--no-message-id

 Documentation/git-am.txt       | 11 +++++++++++
 Documentation/git-mailinfo.txt |  5 +++++
 builtin/mailinfo.c             | 22 +++++++++++++++++++++-
 git-am.sh                      | 21 +++++++++++++++++++--
 t/t4150-am.sh                  | 23 +++++++++++++++++++++++
 t/t5100-mailinfo.sh            |  4 ++++
 t/t5100/info0012--message-id   |  5 +++++
 t/t5100/msg0012--message-id    |  8 ++++++++
 t/t5100/patch0012--message-id  | 30 ++++++++++++++++++++++++++++++
 9 files changed, 126 insertions(+), 3 deletions(-)
 create mode 100644 t/t5100/info0012--message-id
 create mode 100644 t/t5100/msg0012--message-id
 create mode 100644 t/t5100/patch0012--message-id

On 25/11/2014 19:33, Junio C Hamano wrote:

quoted

quoted

If both --message-id and -s are specified, the Signed-off-by goes
last.  This is coming out more or less naturally out of the git-am
implementation, but is also tested in t4150-am.sh.

Nice.  So if you apply a message whose last sign-off is yourself
with both of these options, what would we see?

    1. S-o-b: you and then M-id: and then another S-o-b: you?
    2. M-id: and then S-o-b: you?
    3. S-o-b: you and then M-id:?

I do not offhand know which one of the above possibilities to favor
more over others myself.  Just asking to find out more about the
thinking behind the design.

You currently get (1), which is arguably the most precise but definitely
the ugliest.

In this case (posting as maintainer), I would probably not use "git am
--message-id"; instead I would use an alias to add the Message-Id (with
git interpret-trailers!) after posting to the mailing list, resulting in
either (2) or (3).

I think (but I am not sure) that git-am could use a hook to rewrite (1)
into (2) or (3).

Paolo