How do I allow 2 people to work on the same local copy of a git repository? Or can I?

Question

This might be verbose, but I'm just starting out with git, so I'm still learning here.

Before now, I've had a team of developers all working off of the same hard drive with a local copy of all of our dev files. So we have 1 hard disk and 4 developers. All of us use sublime text and work on the projects together. We all work in the same room, so it's never been an issue to work on the same project at the same time. We just don't work on the same file at the same time. Not the greatest system, but it worked at the time.

Now we want to introduce git to the team for all of the reasons a VCS is important. The problem we're running into is files being locked by one user and no one else can use git on that repository.

Here's an example. I log into my mac in the morning and make some changes to files in Project X located at /Volumes/dev/projectx/. I open terminal and commit those changes to the local repository (stored also at /Volumes/dev/projectx/). My coworker gets in and logs into his mac. He opens up his terminal to check the status on the repo he's working on. So he moves into /Volumes/dev/projectx/ and runs git status. He gets an error message that says the index is locked. In order to allow him to run any git commands on the repo, I have to completely log out (maybe just kill some processes, but I don't know which ones). After I log out, he can work as though there's no problem.

Is there any way that we can both work on the same local repository at the same time? I've also discovered that, if I'm working in a project that has a git repository in it and anyone else even opens the project folder in a finder window, it completely locks me out from using that git repo (same index lock error).

We're willing to change the way we work on files, but since there are literally thousands of projects on the drive, it isn't really practical for each of us to have our own local copies of all of the files. Also, since many of the changes are a very simple text change of some kind, it seems tedious to host all of these repositories remotely and have to pull down all the files anytime we only want to update a single file.

I'm really looking for workflow suggestions here, but the question I asked is kind of the starting point here.

Answer 1

The whole point of using git is that you don't have to do this kind of crazy stuff.

I know what you've said about why you don't think you should all have complete copies. Here's the hard truth. You're wrong. Mostly. But that's ok, you said your willing to rethink how you work and that's good. Ill try to explain why its not that big a deal to have everyone use their own clones.

A assume all code is already in a remote repository - if its not, sign up on http://github.com or http://bitbucket.com and get a free repository, add it as a remote to your git repo, and push it up. Its really very simple.

Each of your developers should then make their own directory locally on their machines where they can clone the whole repository.

git clone http://github.com/yuoraccount/yourrepo ~/clones/localproject

The first time they clone, it will take a little time to download everything, but from then on, only each minor diff will need to be downloaded uploaded. Git is made to be efficient that way.

When you make a change, commit it, and push it up.

git commit -am "i made a small change"
git push origin master

Then everyone can pull it down.

git pull origin master

You can even all work on different branches, so your not just pushing to the same branch. This should all be really simple, and very easy to do.

You can also split your project into multiple repositories, but you don't really need to. Thousands of files is not a big deal, git can handle it without a problem. That's not to say that you won't have some challenges. Git is easy to use, but you'll eventually run into merge conflicts. They will be a little bit frustrating at first, but stack overflow has a thousand answers explaining how to deal with them - you should be fine.

This is what git is for.

Answer 2

Git is a decentralized version control system. The way you want to use git could be described as the opposite of decentralized.

There are a number of perfectly reasonable workflows possible with git, but all of them are going to involve each developer working in a local clone of some repo.

Answer 3

i guess that your "single harddisk" is really a single "network storage" that can be accessed concurrently. otherwise i don't see the problem (the harddisk can only be attached to a single computer at any time; whenever you unplug the disk, any locks should be released!)

anyhow, though you currently consider it impractical, you still might consider using per-user clones of the repositories.

the normal workflow of a multi-user repository would consist of a local clone of each repository on each developer machine (that is: a clone on your personal mac). then the trouble with concurrently accessing the locked central repository doesn't exist! it's exactly the use-case for which git was designed.

this would allow your users to work on the repository even without having access to your central network storage.

if you are concerned about disk-space, you might be happy to hear that git does a pretty decent job of compressing the data (just run git gc every once in a while in your repository).

if (for whatever reasons) you cannot have local copies (on the dev-machine) of the repositories, you might consider having per-user clones on the central storage. git can use hardlinks for local clones, so disk-space should not be an issue here.

Answer 4

Section 4 of the Pro Git book (version 2) is about "Git on the Server" and gives some information about what the OP is trying to achieve.

The OP's situation calls for the "Local Protocol".

Here is how this works:

1. Create a local remote from your project

cd into your project and create the remote in your preferred location (in your example /Volumes/dev/projectx):

git remote add origin /Volumes/dev/projectx.git

2. Set the permissions for the remote

git init --bare --shared /Volumes/dev/projectx.git

This will set the permissions properly so that different users can push and pull to/from the remote. This command does not affect your Git history and is safe to run.

The --bare flag is used because all that is needed for a remote is a bare repository, that is, a repository without a working directory (a repository with only the .git directory and nothing else).

Note: this step is not described in the Git Book for the Local Remote protocol and is only suggested when setting up a Git server, but in the experience of a colleague, without using this step, the permissions were not set properly.

3. Push the data to your new remote

Assuming you have a branch called main (do this for any branch you want your collaborators to have access to):

git push --set-upstream origin main

4. Have your collaborators clone the remote

This works as with any remote:

git clone /Volumes/dev/projectx.git

From here on, everybody can push and pull to/from the local remote.