Re: Error fetching submodule from submodule
- Date: Fri, 8 Mar 2019 09:50:15 +0100
- From: Jesper Rønn-Jensen <jesperrr@xxxxxxxxx>
- Subject: Re: Error fetching submodule from submodule
You are completely right! It works as I expect if I remember the extra
parameter `--recursive` when doing `git submodule update --init
Thanks a lot for your feedback! This is really useful! I learned a
useful thing today :)
On Thu, Mar 7, 2019 at 7:08 PM Jeff King <peff@xxxxxxxx> wrote:
> On Thu, Mar 07, 2019 at 12:07:21PM +0100, Jesper Rønn-Jensen wrote:
> > Hi I think I may have found an error in the way git handles a
> > submodule's submodule. Read further for the example (extracted from a
> > real project).
> First off, thank you for the example script. It made understanding
> what's going on so much easier. :)
> > * I have a main repository which has some submodules defined.
> > * One of the submodules is a common submodule which is also included
> > in one of the other submodules
> > * When running `git fetch --recurse-submodules` I get an error.
> I think the presence of common_submodule in the main repo is actually a
> red herring. if you remove the last two lines of this stanza:
> > git setup main_repos
> > pushd main_repos
> > git submodule add ../common_submodule
> > git commit -m 'added submodule to main_repos'
> the outcome is the same.
> > # This line fails with error code 1 "Could not access submodule
> > 'common_submodule'"
> > git fetch --recurse-submodules
> It looks like "fetch" is smart enough to initialize a submodule when
> necessary, but not smart enough to do so recursively. If I replace that
> line with:
> git submodule update --init --recursive
> then it works as I'd expect. Likewise, cloning the repository with:
> git clone --recurse-submodules main_repos foo
> does what you'd want.
> After that, I think "git fetch --recurse-submodules" does what you want,
> because the submodule repository is already initialized.
> I'm not sure to what degree git-fetch intended to support initializing
> submodules. But it certainly seems like a bug that it handles the top
> layer but does not recurse (i.e., it should either handle all or none).
> Hopefully the commands above at least give you a workaround.
Mobile: +45 2989 1822
jesperrr@xxxxxxxxx (Private e-mail and Google Talk IM)