git submodule when submodule is not a clone of a repository

git submodule when submodule is not a clone of a repository

[Sergio Callegari]

Hi,

I am trying to use submodules to keep a project organized.
I am using a workflow that is probably not the common one and I am running into
problems.

Basically, I have a superproject TOP, with subproject A and B. The thing is that
A and B are _not_ cloned from some existing repo.  They are two brand new
repositories created with git init.

Basically, I am doing this so that in the superproject I commit states with
specific versions of A and B (that are logically distinct things) that are known
to work fine together. Also I am doing this since A and B are actually just
started with TOP, but I have a provision of making them independently developed
things, so I like to have them separated from start.

The thing is that git-submodule makes very strong assumptions on A and B.
Particularly, if there is no url defined from them some commands (notably
update) will fail.

So I need to make a fake url for the submodule (e.g. .) and a fake origin for it
(again .)

Would it make sense to have a git submodule update not try to fetch if there is
no url defined, but still checkout the right commit for the submodule (rather
than skipping the submodule at all) ?

Sergio

Re: git submodule when submodule is not a clone of a repository

[Lars Hjemli]

On Jan 18, 2008 6:25 PM, Sergio Callegari <sergio.callegari@gmail.com> wrote:
> Would it make sense to have a git submodule update not try to fetch if there is
> no url defined, but still checkout the right commit for the submodule (rather
> than skipping the submodule at all) ?

Actually, we could (and probably should) teach git-submodule not to
fetch if the requested SHA1 is already available in the submodule, and
I guess this would solve the problem quite nicely for your usage.

But that will probably not happen before 1.5.4.

--
larsh

Re: git submodule when submodule is not a clone of a repository

[Junio C Hamano]

"Lars Hjemli" <hjemli@gmail.com> writes:

> Actually, we could (and probably should) teach git-submodule not to
> fetch if the requested SHA1 is already available in the submodule,..

I think that is a very sensible optimization.