0

I'm working with a framework that more or less forces me to mix my own code in with framework code--that is, neither can be isolated neatly into its own subdirectory. I have thus far kept them separate by using .gitignore to ignore all framework code, but when I needed to upgrade the framework I created an upgrade branch, and did the upgrade, not initially realizing that this would mean previously untracked/ignored files would be tracked in the new branch, and therefore would no longer exist when I switched back to the master branch.

After recovering from the massive facepalm moment, I thought I'd have to switch to using Git Submodules. The problem this poses, however, is that it appears that code belonging to a submodule must be kept in a subdirectory. From Pro Git:

Submodules allow you to keep a Git repository as a subdirectory of another Git repository.

Is this truly a limitation of git submodules, or is there any way around it?

Improve this question

1 Answer 1

Reset to default
1

I believe so (a submodule is a complete git repository on its own, and the same folder cannot contain two git repositories). I also believe that when there is a submodule in the folder Foo, the outer repository will ignore everything in Foo except that it will keep track of which revision the submodule is at.

(Warning: huge hack) However, you could have two separate git repositories, one inside the other, and "just" be very careful with the respective .gitignores - however, you won't be able to make one repository control which revision the other repository is at. Also, the outermost repository will have to be one folder level outside of where your project is located (if it needs to include top-level files from that folder and you already have your main repository there), so if your project is at /foo/bar/Project, you need to have the framework repository in /foo/bar, in addition to the main repository in /foo/bar/Project.

Improve this answer
Sign up to request clarification or add additional context in comments.

8 Comments

Yes, I thought of that hack. I'd like to avoid it at all costs, since I'm going to be sharing this repo with other developers, and they will undoubtedly ask WHAT I was thinking. They are totally new to Git, so putting them through that will make them curse me and Git both. I'd sooner just mix everything in with my code, though I'm hoping to avoid that.
Oh, actually (reading your answer more carefully) I thought of a worse hack, to allow two repos to exist in the same directory. This can be done by controlling GIT_* environment variables to specify which repo to use and what name you're using instead of the default .git. But in either case, something I'd like to avoid for work I'm sharing with others, and which is not merely experimental.
@iconoclast: Interesting, I didn't know that that was possible. And indeed, I would probably have thought "what an idiot" about any person who would suggest what I suggested - this is a last resort hack ;-)
@iconoclast: The best approach is probably to have everything in the same repository. How about writing a shell script that recreates all framework files, which you can run whenever you switch to an old version of your code?
if I write a shell script, then is it worth keeping the framework files in my repo at all? I'm picturing the shell script doing this: a file system backup or tar of my project, downloading the framework, copying all framework files that are safe to replace, but keeping my files for settings and such (these are included in the framework but mine should take precedence). Did you envision one that did something different?
|

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.