Re: starting completly new repository

Hi,

First sorry if it is anwsered somewhere in the docs, but I am just (slowly) learning how to use git and reading the documentation. Any pointers to some article or tutorial that anwsers my questions will be appreciated.

What I am trying to do is to construct several repositories for the following work order:

1. There is some remote repo that is the main repository (call it main). 
It lives in some server and geneerally only one or two persons can push 
into it. All others can only pull from it.

2. Everybody has his own repo derived from main and she/he can make 
patches and send it to everybody else. One person is responsible for 
taking the patches and pushing them into main.

I need this set up fast and want to know how to do it. What I tried is basically:

(on the server) mkdir main git-init-db touch .git/git-daemon-export-ok

(on my computer) git-clone git://host/main main

But it looks like I must first do some commit on the server? But I can not make empty commit just to have things started? Or maybe there is some other way...

Also I wonder if I can do push over git protocol or I must use real ssh account on the server? This is not clear from the docs... At least not for me. How should I set up my repo (on my computer) to be able to push commits into main repo?

Also what should I set up additionally? How can I easily set author name and email for each repo? What is the difference between author and commiter and how should I set this up here?

Is there any documentation about git config file? Can I set author name, email and preffered editor in it or must I use environment?

Is there some irc channel for asking dumb questions as above and having them anwsered fast or should I use this mailing list?

Thanks in advance,

Grzegorz Kulewski

On Sat, 25 Mar 2006 22:06:42 +0100 (CET) Grzegorz Kulewski <kangur@polcom.net> wrote:

Did you actually start the git-daemon on the server?
 

You can't commit over the git protocol, you'll need to use ssh. No need to do anything to your repo in order to push.

See below.

Just what you might imagine, the person who created the patch and
the person who entered it into the repository respectively.
 

$ git repo-config user.email "user@email.com" $ git repo-config user.name "full name"

use the EDITOR environment variable to choose your desired editor.

 
#git on irc.freenode.net

Sean

Hi,

Dear diary, on Sat, Mar 25, 2006 at 10:06:42PM CET, I got a letter where Grzegorz Kulewski <kangur@polcom.net> said that...

well, there's Documentation/tutorial.txt, but it doesn't cover your questions, it looks.

You can of course make an empty commit, but it seems strange that you would want that - won't the people start working on some initial version of the project?

You must use real ssh account on the server. You can limit the account to be able to do only git pushes/pulls by setting its shell to git-shell.

 From Cogito's cg-commit documentation:

Each commit has two user identification fields - commit author and committer. By default, it is recorded that you authored the commit, but it is considered a good practice to change this to the actual author of the change if you are merely applying someone else's patch. It is always recorded that you were the patch committer.

See git-commit-tree(1), the COMMIT INFORMATION section.

One big problem with Git documentation (other than e.g. falling out-of-date occassionaly) is that information is scattered between porcelain and plumbing or even between several commands (e.g. git-whatchanged(1) which doesn't even directly mention where to gather the full list of available options).

Thankfully, the plumbing documentation is mostly coherent. ;-)

#git on Freenode (surprisingly ;)