Using Git to Maintain Your Website

Is svn that bad? I used for a few projects (personal projects so it’s just me) and I find it easy to use. Never really had any problems. I have it on at my host (at antoniocs.org) and with tortoise I can commit or update easily.

I have started doing exactly this with Mercurial and was wondering why more people weren’t doing it. Is there any downside to it? None that I can see.

Another article using the same technique is … http://joemaller.com/2008/11/25/a-web-focused-g…
I'm working on something similar for Drupal sites. Do you have issues with server created content? I also wonder if versioning the DB's would be useful too using a separate repo. Git seems to be remarkably efficient with space.

I don't know of any downsides other than perhaps initial setup time (which is very low once you get comfortable with it). I am very happy with the system.

Thanks, that's useful to know.

It seems to work well on a test setup, I'm waiting for Git to be installed on the server (because it's easier than installing for multiple domains on one server).

Do you use Git excludes to avoid overwriting site specific files? I want to use it with Drupal which uses a settings.php for the DB & config. I think I would like them tracked in Git, but not pushed & pulled between different repos. I wonder if the hub could filter them?
Do you have this issue?

This helped me (git newbie) immensely. I set this up in combination with the EGit/JGit Eclipse plugin – for anyone else Eclipse-ifying a static website with Git, also look at: http://github.com/guides/using-the-egit-eclipse…
One note, you should add a default .project (copy one from another project & change the xml name tag) in the root of the to-be-repo'd folder before doing 'git add .', or else Eclipse won't be able to load the project by importing!

Rocku,

Checkout the script in the contrib/hooks section of Git source.

The script will read permissions into a file in the repo & write them out too. It just requires hooks to call the script.

I've not implemented this myself, but there is a .gitignore file that you can use for this. I'd have to check out the usage, though…

I've not implemented this myself, but there is a .gitignore file that you can use for this. I'd have to check out the usage, though…

You may want to remind users that when they issue the clone command (git clone ssh://yoursite.com/path/to/html.git), they need to specify the full path to their git repository. I was getting the “fatal: The remote end hung up unexpectedly” error until I realized I was not specifying the full path on the remote server filesystem.

I'm in the midst of converting to Git and looking around on the web, I see that everyone seem to do dev -> bare repo -> live site. I”m new to git and there's one thing I don't understand… Why not make the live site the repo, what's the use of the bare repo in the middle?

Any clarification on that much appreciated :)

@ Yann
That's my thought too.
Shouldn't you simply branch your project at the main repo, instead of having an almost live copy on the server?
Also it seems counter to the distributed model of Git, that is, each developer has the whole project, so they can pull from each other to see code and collaborate.
That second point may be quite misinformed.
Anyone know the answer?

I've been looking around at different topics on using Git to maintain a website. My Problem is that I don't understand how with all the branching and everything else to actually make it effective. For example, I have my main branch, which is my live site, set it up so that when it's pushed it updates the live site. Then I can have a development branch for trying out changes and such, this is awesome, but after I make the changes how do I try them out for like a PHP dynamic site? I assume I would create another repo for development that updates a development server or space, so I make my changes, push to the dev, then I can test them there? Okay then with git again it's really easy to branch again for maybe a quick patch, or to try another little change out. But where do I push that to test those changes? Do I need to create yet another repo that updates another web space somewhere, which would only be short lived, so seems like a lot of work. I guess I really really like the idea of Git, but don't understand how to utilize the branching and still be able to test changes. Or do you just not use branching so much in this kind of situation? Or is it really like a few comments I've seen online where people say “it's stupid to update a website with git”?

Thank you for your post; it has been very helpful! I am in the process of making this set-up for myself, and it seems that the post-update hook you give has changed yesterday or today; is that correct? The old was:

WORKDIR=”/your/live/html/directory/”
export GIT_DIR=”$WORKDIR/.git”
pushd $WORKDIR > /dev/null
git pull
popd > /dev/null

And the new is:

cd ../htdocs
env -i git pull

It seems so, but just to clarify, is the new/replacement simply a more efficient way of doing the same thing? Thanks again!

hey

thanks for the tip. I'll try it

With git you have a history of all changes, and you can roll back to
specific points.

You could do 'git clone you@yoursite.com:rel/to/home.git'. 'you@' optional if usernames are the same, or you set up something in .ssh/config; I regularly just have 'devbox:somerepo.git' as remote 'url'.