Unpleasent Git experience

Git and mercurial don’t need servers. They are distributed, thats the whole point. The most as far as a centralized repository most projects will need is a directory (or directories) on a file server that people can push to and pull from. Other options, for mercurial at least, I believe git offers similar choices, are accessing remote fileshares via ssh, running with a temporary builtin http server, or running from within apache or lighthttpd servers. Neither were for me particulalry more complicated than svn.

I would suggest that if you need a centralized repository why not use svn, and use some of the bridging options so that developers or teams can use distributed vcs for their branching and merging, but the main centralized repo (should you really want such a thing) runs svn.

“Git and mercurial don’t need servers. They are distributed, thats the whole point. ”

I know. Most projects have one gathering point though. See Github. Most companies need an official repository to integrate and draw releases from – I can’t think about a way to release a website from all distributed developers repos.

“Neither were for me particulalry more complicated than svn.”

Glad for you. See my post for a different experience with Apache2, Debian and git.

A combined Svn-Git setup looks strange to me. I’ll keep trying to setup git on the server for a repo.

Thanks for the insights.

Ouch. Sorry to hear you had a bad experience using git the first time.

How did you fix the error?
PUT error: curl result=22, HTTP code=403

Hi Stephan,

I have just installed git 1.6.2 on CentOS.

I am encountering the error:
fatal: pathspec ” did not match any files

when I run:
git add .

How did you fix it?

Thanks,
Jason

I’m exploring git too, and even if it is a big shift from centralized SCMs (I’ve been using SVN) I already see benefits of git, the main one being ability to check in revisions locally, thus maintaining history of changes, which in turn encourages experimenting new features in private branches.

My only concern is how will new developers, with no previous experience with SCMs of any kind, get their heads around it.

@Jason

Have you made any changes (created file/directory) in your working copy (directory in which you run “git add .” )?

@Divya I also had the PUT error: curl result=22, HTTP code=403 problem and it went away when I made sure that all files in the git reporistory where owned by the apache2 user (www-data in my case).

While I had chowned all the files before, I think I had introduced new files that did not belong to www-data when I run git update-server-info manually in the repository.

Bjorn got the answer, thanks it works.