I tease you from time to time about this topic so here it is. I’ll show you how i manage my dotfiles. It’s a recurring subject in the linuxian world and everybody use a different solution.
Some use a soup of symbolic links in every directions to centralize everything in one easily copy/pastable directory. Others rely on personal scripts or even software like GNU Stow. Some nerdier use Makefile as a way to get themselves out of this burden (and in fact it’s not so crazy). In my case I wanted to avoid having to learn a new soft. I wanted not to install a dedicated soft/script. It can’t be a PITA with links**. I want versionning to recover from mistakes.
Git to the rescue
A complete solution is to use Git. With this you have versionning, easy synchronization and you can have branches between multiple computers. But how do you set it up ? Do you put everything from /home in the repo but you’ll have to cope with files in your home you don’t want to add in your repo ? Moreover, you’ll have all git’s file in your home (description, HEAD, hooks, logs, refs, and so on).
The git killing command !
git –git-dir=/home/lord/.myconf –work-tree=/home/lord
Here’s the beast. It fits well in an alias (config for example). So every git’s files will be stored in /home/lord/.myconf so your /home stays clean.
Magnets how do they work ?
First you’ll have to add files you want git to track with config add .ssh/config will add a precise file or config add .ssh will add all files in the folder. BEWARE ! Don’t put sensitive files in your git repo if you’re not absolutely confident with all machines which will sync/store you repo ! (I don’t put my ssh keys in my repo).
Next you check everything is good with config status :
lord@hermes ~ %config status
On branch master
Your branch is ahead of ‘origin/master’ by 4 commits.
(use “git push” to publish your local commits)
Changes to be committed:
(use “git reset HEAD
We see that locally my repo have 4 commits that the “master server” haven’t and that i added files that aren’t commited yet. So let’s commit them config commit .ssh/config .ssh/known_hosts -m “[SSH] Adding .ssh/config and .ssh/known_hosts” .
lord@hermes ~ %config commit .ssh/config .ssh/known_hosts -m "[SSH] ajout de .ssh/config et .ssh/known_hosts" [master 35259ea] [SSH] ajout de .ssh/config et .ssh/known_hosts 2 files changed, 271 insertions(+) create mode 100644 .ssh/config create mode 100644 .ssh/known_hosts
Ok it’s done. If later you want to remove some files from your repo (but not the real files) you’ll have to config rm –cached .ssh/id_rsa . The important thing is –cached.
Hey ! It’s turning into a git tutorial ! Yes I know but there is just some more commands to learn. If you already know git you can skip this blog post now.
How to change the past ?
You need use config log to see all the past commits with their dates, messages and hash id. To see the content of a commit use config diff xxxxxx (you can limit to one file with config diff xxxxxx your/file) where xxxxx is the hash id of your commit.
Now you have found the faulty commit and want to cancel it ? The best is to config revert xxxxxxxx. You’ll quickly learn to split your commits in multiples smaller commits. The smaller the commit the easier to get back.
If you only need to temporarily go back to a previous state you just need to config checkout xxxxxxxx. You’ll be in the DETACHED HEAD state. You should keep you from editing anything in this mode, if not it’s quite hard to merge all your changes and go back to the current state. By the way, to go back to the current state enter config checkout master and there you are.
You should know that there is a reset and reset –hard command but those can be very destructive. RTFM heavily before trying to use them.
How do I sync elsewhere ?
I didn’t synchronized for a long time but now I do it mainly for a backup purpose. I now have a personnal gogs installation (inside a Sandstorm instance) where I keep a copy of many repo. So the repo exist locally and in gogs but also on a third location as a backup. You just have to add a remote to the repo with config remote add gogs https://xxxxxxx and config remote add backup ssh://xxxxxx .
And now after commits i must push with config push gogs and config push backup.
And now you can config fetch or config status and config pull. There you are.
Now if you want to add this on a new machine all you have to do it going to /home/user . You remove any file present (backup)and config init . then a nice config remote add origin xxxxxxx, followed by config pull origin (it can take some time for the first sync) and the final touch is config checkout master. And you’re done with all your files.
Not as simple as you said !
No. It looks hard but in fact you won’t have to do complex things everyday. You’ll mainly config commit and config push. If you already know git it’s very easy, if not it’s a very good reason to discover git.
The real magic is git –git-dir=/home/user/.myconf –work-tree=/home/user