Re: [PATCHv2 0/6] git clone: Marry --recursive and --reference

From: Junio C Hamano <gitster@pobox.com>
Date: Aug 6, 2016, 17:29 UTC
Message-ID: <xmqqd1lliz5k.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <20160806012318.17968-1-sbeller@google.com>

Stefan Beller <sbeller@google.com> writes:

>  Some submodules in the referenced superproject may not be there, 
>  (they are just not initialized/cloned/checked out), which yields
>  an error for now.

Perhaps you can teach "git clone --reference" an new option (--reference-if-able) to do this? Then

    When `--reference` is given together with `--recursive`,
    the reference repository is assumed to contain the submodules
    as well and the submodules are setup as alternates of the
    submodules in the given reference project.
 
in which "assumed" is a horrible wording (leave the reader
wondering: "so what happens to my data when the assumption does not
hold") can become a lot more reasonable

    When using --reference with --recursive, the --reference is used
    to specify a repository that has a copy of the superproject.  If
    that copy has submodules cloned for itself in its $GIT_DIR/modules,
    they are used as --reference when cloning submodules in the
    resulting clone.

and readers expectation would match with the reality. Their submodules would be cloned in a regular fashion if the central mirror does not have it, and would take advantage of it if there is already a clone.

Come to think of it, do we even need --super-reference? "git clone --reference --recursive" is a two step process, in that first the superproject is cloned while creating objects/info/alternates, and then submodules are cloned (via "update --init"). Can we make the procedure to clone a submodule always look at the reference of the superproject (i.e. objects/info/alternates) and try to borrow from the place in it that corresponds to the submodule? That way, not just "git clone --reference --recursive" would take advantage of the existing mirrors of submodules, a user who does this:

    $ git clone --reference $URL super
    $ cd super
    $ git submodule update --init ...

would be able to take advantage of the "what the mirror the superproject uses already has" when cloning the submodules, no?