RFC: update hook for GPG signed submission on secured branches

Dev & users:

I would like your feedback.

I am working on a continuous integration (CI) system for CipherShed.org and we want to allow any fork to submit their patch to our CI server farm which will do integration testing. We want it to be easy to submit code, but to not allow interference with other submitters branches or fetching of other people's (any for bandwidth and simplicity) branches.

What would you change? Any bugs that you see?

-Jason (no cc please, I am on the list)

Here are the 2 configs.

$ cat config
[core]
        repositoryformatversion = 0
        filemode = true
        bare = true
[http]
        receivepack = true
        uploadpack = false
        getanyfile = false

$ cat hooks/update #!/bin/bash

# (c) 2015 PD Inc. License found at http://www.apache.org/licenses/LICENSE-2.0 .

# --- Command line refname="$1" oldrev="$2" newrev="$3"

# --- Safety check
if [ -z "$GIT_DIR" ]; then
        echo "Don't run this script from the command line." >&2
        echo " (if you want, you could supply GIT_DIR then run" >&2
        echo "  $0 <ref> <oldrev> <newrev>)" >&2
        exit 1
fi

if [ -z "$refname" -o -z "$oldrev" -o -z "$newrev" ]; then
        echo "usage: $0 <ref> <oldrev> <newrev>" >&2
        exit 1
fi

sigkey=$(\
 git cat-file $newrev -p |\
 sed -e '/^ /{H;$!d;}; x;/^gpgsig /!d; s/^gpgsig//;' |\
 cut -c 2- |\
 gpg --list-packets --textmode |\
 sed '/keyid/!d; s/.*keyid \([0-9A-F]\{16\}\).*/\1/I' \
)

if [ -z "$sigkey" ]; then
        echo no GPG signature on commit $newrev
        exit 1
fi

if ! gpg -k "$sigkey" 2> /dev/null > /dev/null; then
        # "$sigkey" not known
        RES="$(gpg --keyserver hkp://pgp.mit.edu --recv-keys "$sigkey" 2>&1)"
        if [ $? -ne 0 ]; then
                echo "$RES"
                exit 1
        fi
fi

sigstatus=$(git log $newrev --pretty=format:%G? -n 1)

case "$sigstatus" in
        G)
                #ok, trusted
                ;;
        U)
                #ok, untrusted
                ;;
        *)
                #not ok
                echo sigstatus: $sigstatus
                git log $newrev --pretty=format:%GG -n 1
                exit 1;
                ;;
esac

if [[ $refname != refs/heads/* ]]; then
        echo only heads may be pushed, illegal ref: $refname
        exit 1;
fi

head="${refname:11}"

shopt -s nocasematch

case "$head" in
        ${sigkey}-*)
                #ok
                ;;
        ${sigkey}/*)
                #ok
                ;;
        ${sigkey:(-8)}-*)
                #ok
                ;;
        ${sigkey:(-8)}/*)
                #ok
                ;;
        *)
                #not your branch
                echo "you (a.k.a. $sigkey) are not authorized to push to branch: $head"
                echo "try making a branch like: $sigkey-... or $sigkey/*"
                echo "you can use a less secure ${sigkey:(-8)} too"
                exit 1
                ;;
esac

exit 0

--
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
-                                                               -
- Jason Pyeron                      PD Inc. http://www.pdinc.us -
- Principal Consultant              10 West 24th Street #100    -
- +1 (443) 269-1555 x333            Baltimore, Maryland 21218   -
-                                                               -
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
This message is copyright PD Inc, subject to license 20080407P00.

"Jason Pyeron" <jpyeron@pdinc.us> writes:

"-p" being a command line option should come before revision, but more importantly, because you accept pushes only to refs/heads/, you would want to explicitly require commit objects, no? i.e.

	git cat-file commit "$newrev" |

I am not sure if you need these unsightly backslashes. When you stop talking to it after saying "$(", or "$( git cat-file ... |", the shell _knows_ that you haven't stopped what you want to tell it.

It always makes me feel nervous to see people pipe sed output to another filter that is a mere s/.//;

Is this complex pipeline the same as this (I didn't understand the trailing I at the end)?

	git cat-file commit "$newrev" |
        sed -ne '/^gpgsig /,/^ -----END/{
        	s/^gpgsig //
                s/^ //p
	}' |
	gpg --list-packets --textmode |
        sed -ne '/^:signature packet:/s/.*keyid \([0-9A-F]*\).*/\1/p'

I am not sure if the design of this, to require signature only on the tip commit, is sound. That is not a -bug- in the script, though.

It is hard to tell where the magic number 11 comes from.  Perhaps

    head="${refname#refs/heads/}"

reads easier?

Junio C Hamano <gitster@pobox.com> writes:

Ehh, s/stopped/finished/, that is.

True.

It was a very quick Lego block build.

Case insensitive, could have used [0-9a-fA-F].

Will all future signature values end with a "^ -----END"? I was only going on the assumption that continuation lines start with a single space.

It is to handle the "all my devs worked on this, they do ________ GPG", so as long as the tip os signed, it is an implicit I am responsible for what is submitted.

Much.

Thanks!

-Jason

--
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
-                                                               -
- Jason Pyeron                      PD Inc. http://www.pdinc.us -
- Principal Consultant              10 West 24th Street #100    -
- +1 (443) 269-1555 x333            Baltimore, Maryland 21218   -
-                                                               -
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
This message is copyright PD Inc, subject to license 20080407P00. 

On Fri, Jan 16, 2015 at 02:47:25PM -0500, Jason Pyeron wrote:

Isn't this an ideal scenario for using the signed pushes introduced in Git 2.2.0?

"Jason Pyeron" <jpyeron@pdinc.us> writes:

Ahh, a GNU extension.

If it bothers you, alternatively, you can stop at the end of the commit object header, i.e.

	/^gpgsig /,/^$/{
        	s/^gigsig//
                s/^ //p
	}

but I do not think it matters that much ;-)