Re: [PATCH] commit: teach --gpg-sign option

From: Michael J Gruber <git@drmicha.warpmail.net>
Date: Oct 7, 2011, 08:40 UTC
Message-ID: <4E8EBAFE.8020805@drmicha.warpmail.net>
In-Reply-To: <robbat2-20111006T221637-481195848Z@orbis-terrarum.net>

[readding JCH to cc whom you dropped] Robin H. Johnson venit, vidit, dixit 07.10.2011 00:24:

Show 13 quoted lines

> On Wed, Oct 05, 2011 at 05:56:55PM -0700,  Junio C Hamano wrote:
>> And this uses the gpg-interface.[ch] to allow signing the commit, i.e.
>>
>>     $ git commit --gpg-sign -m foo
>>     You need a passphrase to unlock the secret key for
>>     user: "Junio C Hamano <gitster@pobox.com>"
>>     4096-bit RSA key, ID 96AFE6CB, created 2011-10-03 (main key ID 713660A7)
>>
>>     [master 8457d13] foo
>>      1 files changed, 1 insertions(+), 0 deletions(-)
> I like it, but I have a couple of questions: 
> 1. Are the sig lines used in computed SHA1/commitid of a given commit (I
>    see examples w/ --amend and that would usually change the SHA1)?

Yes, just like with tag objects.

> 2. Can we allow more than one person sign a commit?

I don't think we support it now (tags) but we could allow concatenating signatures since they are detached.

There's a somewhat delicate issue here: The signature (tag/commit) is a signature on the contents of the object, and is itself not part of the contents (or else we would have a chicken-egg-problem).

The sha1 of the object is determined by the content+header, i.e. including the signature.

So, by adding a signature, you change the sha1, but any existing signature remains valid.

This is also how you can try to achieve a specific sha1 for a given object content...

> 3. If I have prepared a series on a local branch, and I want to sign all
>    of them, is this a variant of rebase or?

If you really want to sign all you can rebase-i and use "exec" to do that automatically, but there's no point: signing the top-most commit serves the same purpose.

> I think this isn't a replacement for push certificates, but has value in
> itself. It's certainly provides better integration than the
> signature-in-note variants.
>

I do think it's meant as an implementation of push certificates. I don't see any other value in it which could not be achieved by signed tags. Can you describe any?

Michael