A minimal, "single-serving" Git HTTP server for

1. emergencies
2. local, ad hoc experimentation
3. Git education

Normal, everyday use is not recommended unless security and performance are non-issues. Requires a normal git installation.

There's really no need to git clone or pip install since this thing is a single file. See --help for available options like basic auth, create-on-clone, create-on-push, SSL/TLS, etc.

Use case

Some VM or container in which SSH/NFS/SMB aren't desired and for which folder sharing or volume mapping aren't worth configuring

On the host:

laptop:~$ curl -L "$github_raw_url" | python3 - ./www &
Serving over port 8000 ...

On the client:

[my_vm]# git clone http://laptop.local:8000/repos/my_repo.git
Cloning ...

Some caveats

1. Your network setup might require that you provide an IP address (likely of a virtual bridge) in place of localhost or laptop.local. Export that (or 0.0.0.0) as _HOST to the server's environment.

2. The $GIT_DIR (.git folder or bare repo) being cloned must reside 2+ levels below the "document root" (~/www in the example above). If that's a problem, export _FIRST_CHILD_OK=1 or make a dummy tree and nest a link to the repo:

   ~/
   ├── www/
   │   └── repos/
   │       └── my_repo.git -> ../../my_repo/.git
   └── my_repo/
       └── .git/

3. Non-bare repos must have the receive.denyCurrentBranch option set to updateInstead in order to receive updates (pushes) and update the working directory:

   >$ git config --add receive.denyCurrentBranch "updateInstead"