This is for those who haven't used git before and need a crash-course
on basic operations. I'll keep this as simple as possible, and focus
specifically on working on Stone Soup, rather than git in general.
I've also added links to the official git docs at the end, which you
can read instead of, or in addition to this, if you're inclined.
Installing git
--------------
I strongly recommend using at least git 1.6 or later. While you can
use older versions, the newer versions are much more user-friendly.
This guide assumes you're using 1.6.
Linux: Install git using your package manager. The git package is
usually called git-core.
Mac: Install git using Mac ports or Fink. The MacPorts port is
called git-core (sudo port install git-core).
Windows: Install msysgit (https://github.com/msysgit/msysgit/releases).
Be sure to download the 'msysGit-netinstall', as this also includes the
necessary tools to build Crawl - other installers include only Git itself.
TortoiseGit is apparently pretty usable now for those who want Windows
explorer integration, but I have not used it myself.
Using git for Stone Soup development
------------------------------------
0. Basic git settings:
Set your name for git to use when committing changes:
$ git config --global user.name "John Doe"
Set your email address:
$ git config --global user.email "jdoe@users.sourceforge.net"
This sets your default e-mail id for all git work on your system. You
can use any e-mail address you have. In the above, John Doe is still
using his old sourceforge id as a spam-catching measure.
1. Clone the crawl repository from gitorious (analogous to svn checkout):
Developers (anyone with commit rights to the git repo):
$ git clone git@gitorious.org:crawl/crawl.git
Other users:
$ git clone git://gitorious.org/crawl/crawl.git
"git clone" clones Crawl's entire git repository to your machine. When
it's done, you have a full local copy of the crawl repository with all
its history.
If you want to build the contributing libraries rather than using the
versions already installed on your system, you can fetch the submodules:
$ cd crawl
$ git submodule update --init
2. Sanity-check your cloned repository (optional):
When you come back to a git repository after a while, you may not
recall where you cloned it from. You can check with:
$ git remote -v
For me, this reports:
origin git@gitorious.org:crawl/crawl.git (fetch)
origin git@gitorious.org:crawl/crawl.git (push)
See all available branches in the repository:
$ git branch -a
See all tags:
$ git tag
3. SSH keys (for core developers):
If you want to actually push to the gitorious repository, you will
need to create an ssh key and upload your public key to gitorious.org.
Let's create an ssh key (if you already have a key, skip this
step):
$ ssh-keygen
You can accept all the default options and use an empty passphrase
for convenience (don't do this on an account you share with other
people, or they can commit to the repository too :P)
Once the key is generated, you'll have two files id_rsa, and
id_rsa.pub in your ~/.ssh (the .ssh directory in your home
directory). id_rsa is your private key, and should not be shared or
given to anyone else (or they can pretend to be you); id_rsa.pub is
your public key, and this is what you'll upload to gitorious.
Go to gitorious.org, login, and hit the "Dashboard" link to access
your account settings. Click on "Manage SSH keys" and then on
"Add SSH key".
Copy the one line in your id_rsa.pub, and paste it into the
public SSH keys text area. Hit Save when you're done.
Within a few minutes of uploading your key, you should be able to
push/pull from gitorious.org.
Windows users (msysgit) can follow these exact same steps in a git
bash prompt.
You can also create a DSA key instead of an RSA key (or use an
existing DSA key). gitorious accepts both.
4. Updating your repository with the latest changes from gitorious.org:
Use
$ git pull
to grab the latest commits from gitorious.org.
git pull assumes your working tree is clean; it will refuse to
overwrite any files that you've modified locally (unlike svn's
"svn update", which will happily try to modify the file anyway,
and add conflict markers if there are conflicts).
If git pull fails because you have local changes, you have two
options:
1. Complete your local changes and create a commit, then pull
again.
2. Temporarily save (stash) your local changes, pull changes from
gitorious, and reapply your local changes:
$ git stash
(this saves your local changes)
$ git pull
(grab changes from gitorious)
$ git stash apply
(reinstate your local changes)
If your local changes conflict with the changes from gitorious,
git stash apply will warn you of the conflict and add conflict
markers to the relevant files.
5. Committing a change to the 'master' branch:
git's master branch is the equivalent of svn trunk. Immediately
after cloning a repository, the cloned crawl repo will be on the
master branch. You can check what branch you're working on at any
time with:
$ git branch
The active branch will be asterisked.
Before starting work, let's make sure our working copy is not
dirty:
$ git status
# On branch master
nothing to commit (working directory clean)
Looking good. For our first change, let's assume we're fixing a
bug in, say, stuff.cc.
$ vim stuff.cc
<hackhackhack>
$ make
<test; that bugfix rocks the world>
Let's check how git sees things now:
$ git status
# On branch master
# Changed but not updated:
# (use "git add <file>..." to update what will be committed)
# (use "git checkout -- <file>..." to discard changes in working directory)
#
# modified: stuff.cc
#
no changes added to commit (use "git add" and/or "git commit -a")
git sees that we've modified stuff.cc. You can see what changes
you've made with:
$ git diff
Time to actually commit the change:
$ git commit -a
git will bring up your preferred editor for you to enter your
commit message. On Windows, you may have to set your EDITOR
environment variable; alternatively, you can specify your commit
message on the command line:
$ git commit -a -m "Fitted stuff.cc with warp drive"
When you're done, your change has been committed to your local
repository. Let's try git status again:
$ git status
# On branch master
# Your branch is ahead of 'origin/master' by 1 commit.
#
nothing to commit (working directory clean)
So git's telling us that we have one local commit that we haven't
sent to gitorious' git repository yet. Let's send in our fix
to gitorious.org (core devs only):
$ git push
You will be prompted for your password again. git should then
push your commit, producing a last line that looks like:
54ea5f1..ec2e15e master -> master
The exact commit ids will differ, but a successful push looks
like this. Your push may fail if someone has pushed changes to
gitorious already (git will warn you about a non-fast-forward).
In this case, just pull and push again:
$ git pull
$ git push
6. Patching (for regular contributors, not core devs):
Once you have commited and tested a change, you can create a patch for it.
You can then send this patch to the devs or upload it to a mantis
issue for review and potential addition to the game.
To start, make and save your changes in whatever text editor you choose.
Then, as explained above, commit those changes:
$ git commit -a
In most cases, this is a good time to compile and test the change. Make
sure it works correctly, testing in Wizard Mode as appropriate.
When you're happy with your patch, you can make a patch file with the
format-patch command. For example:
$ git format-patch -1
This will create a patch from the most recent commit you've made. The
patch file will be placed in your base Crawl source directory.
For a more detailed patch creation guide, check out patch_guide.txt in the
/docs/develop/ directory.
7. Viewing revision history:
You can see a history of changes with
$ git log
git log by itself may make it appear that the change history is a
straight line, but we know that git can handle branching
histories. We can request that git log show us the branching
history with:
$ git log --graph
You can also view history using the gitk GUI (this is installed
by default; everyone should have it, and I recommend it):
$ gitk
The history commands normally show you the history of the current
branch, but you can view other branches/tags' histories by naming
the branch or tag:
$ gitk origin/stone_soup-0.2
$ gitk release-0.5.1
If you want to annotate a file with last author and commit to
change each line in the file, you can use git blame, which is
similar to svn blame:
$ git blame stuff.cc
8. Committing a change to the 0.5 branch:
So far we've restricted our attention to 'master', which is the
easiest branch to work with, since it's selected by default. Now
let's say we have a bug report with a 0.5.1 save, and we need the
0.5 code to test the save with:
Let's take a look at the branches we have locally:
$ git branch
* master
The only local branch in our repository is master. Let's look at
the branches in the remote repository (gitorious):
$ git branch -r
origin/HEAD -> origin/master
origin/master
origin/stone_soup
origin/stone_soup-0.1.3
origin/stone_soup-0.1.4
origin/stone_soup-0.1.5
origin/stone_soup-0.1.6
origin/stone_soup-0.1.7
origin/stone_soup-0.2
origin/stone_soup-0.3
origin/stone_soup-0.4
origin/stone_soup-0.5
So our local repository has a "master" branch corresponding to
gitorious' "master" branch. We do not yet have a branch corresponding
to gitorious' "stone_soup-0.5", so let's create the branch and switch
to it:
$ git checkout -b stone_soup-0.5 origin/stone_soup-0.5
Branch stone_soup-0.5 set up to track remote branch stone_soup-0.5
from origin.
Switched to a new branch 'stone_soup-0.5'
master and stone_soup-0.5 are both local branches now:
$ git branch
master
* stone_soup-0.5
git status will also confirm that we're on 0.5 now:
$ git status
# On branch stone_soup-0.5
nothing to commit (working directory clean)
A quick peek at the history to make really sure we're on 0.5:
$ git log
To grab the latest changes for 0.5:
$ git pull
Ok, now we compile 0.5, test the 0.5 save and verify that a bug
exists. Once we've fixed the bug, we create a commit:
$ git commit -a
Check git status:
$ git status
# On branch stone_soup-0.5
# Your branch is ahead of 'origin/stone_soup-0.5' by 1 commit.
#
nothing to commit (working directory clean)
Right, we're ready to push our fix to gitorious. But now that we
have multiple local branches, let's first ask git what it plans to
do when we push:
$ git push --dry-run -v
Pushing to git@gitorious.org:crawl/crawl.git
Enter passphrase for key '...':
To git@gitorious.org:crawl/crawl.git
= [up to date] master -> master
b05bb66..976e722 stone_soup-0.5 -> stone_soup-0.5
So git wants to push the local master and stone_soup-0.5 branches
to the corresponding branches on gitorious; the master branch has
no new local changes, whereas the 0.5 branch does (the new change
is 976e722).
By default, git-push will push *all* your local branches to the
corresponding branches on gitorious. This is important to
remember; if you had local commits on master, they would also be
pushed to gitorious. This behaviour can be changed (the option
is called "push.default") if it bothers you, or you can do:
git push <remote> <local branch>:<remote branch>
So for instance, if I am working on 'master' locally, but want to push
to a remote branch called 'fixes-for-upstream', I could do:
git push github master:fixes-for-upstream
Once you're done working on 0.5, you can switch back to "master"
with:
$ git checkout master
The next time you need to work on 0.5 again, you can return to it
with:
$ git checkout stone_soup-0.5
9. Common operations:
Reverting Changes
-----------------
It often happens that you make a change to a file that you didn't
want, or accidentally delete files that you did want. You can
revert a file to its pristine version as:
$ git checkout stuff.cc
This also works to bring back files you accidentally deleted. If
you forget these commands, "git status" will remind you:
$ git status
# On branch master
# Changed but not updated:
# (use "git add/rm <file>..." to update what will be committed)
# (use "git checkout -- <file>..." to discard changes in working directory)
#
# deleted: mt19937ar.cc
#
no changes added to commit (use "git add" and/or "git commit -a")
Adding New Files
----------------
When committing changes with "git commit -a", newly created files
won't be added to the commit unless you request it with "git add".
Let's take an example:
$ git status
# On branch master
# Untracked files:
# (use "git add <file>..." to include in what will be committed)
#
# dwim.cc
nothing added to commit but untracked files present (use "git add" to track)
Before we commit, we must add the new file:
$ git add dwim.cc
$ git status
# On branch master
# Changes to be committed:
# (use "git reset HEAD <file>..." to unstage)
#
# new file: dwim.cc
$ git commit -a
In general, "git status" is your friend. It will usually tell you
exactly what you need to do.
10. Branching:
Let's say it's time to create a stable 0.6 branch. Here's how you'd
do it:
a) Branch 0.6 from master:
$ git checkout -b stone_soup-0.6 master
Let's check our local branches now:
master
stone_soup-0.5
* stone_soup-0.6
stone_soup-0.6 is currently a *local* branch. gitorious' git
repo doesn't have it yet.
b) Push the local branch to gitorious with a --dry-run first to
make sure we're not lousing up anything:
$ git push --dry-run -v origin stone_soup-0.6
Pushing to git@gitorious.org:crawl/crawl.git
Enter passphrase for key '...':
To git@gitorious.org:crawl/crawl.git
* [new branch] stone_soup-0.6 -> stone_soup-0.6
Now for the real push:
$ git push origin stone_soup-0.6
c) Point your local branch at the new branch in gitorious:
Now that we've created the branch on gitorious, we want to set
up our local "stone_soup-0.6" to get changes from gitorious'
"stone_soup-0.6" when we do a git pull. We do this as:
Tell git that stone_soup-0.6's remote repository is on gitorious:
$ git config branch.stone_soup-0.6.remote origin
And that the corresponding branch in gitorious is stone_soup-0.6:
$ git config branch.stone_soup-0.6.merge refs/heads/stone_soup-0.6
Now you can pull changes from gitorious' 0.6 branch:
$ git pull
This is only necessary for new branches that you create locally
and push to gitorious. This configuration is automatically set
up for you when you work with branches that already exist on
gitorious.
11. Tagging:
Once 0.6 is fit for a release, you'll need to tag it:
$ git checkout stone_soup-0.6
$ git tag -a release-0.6 -m "0.6: Now with extra kangaroos"
12. Committing changes from one major branch of development to another:
Once we create a stable branch (say 0.5), we usually make all
changes to trunk, but apply bugfixes to the 0.5 branch as well.
You can apply changes from master to stone_soup-0.5 with git
cherry-pick:
$ git checkout master
[ make the changes for the bugfix ]
$ git commit -a
$ git checkout stone_soup-0.5
And cherry-pick the tip commit of the master branch (the bugfix
we just committed there):
$ git cherry-pick master
If the commit you want to cherry-pick is not the tip of a branch, just
use the commit's id:
$ git cherry-pick b0ed1449
git push will then send in the commits from both branches.
Additional Reading:
------------------
I've intentionally kept my examples very simple and glossed over a
lot of details. I've covered the common operations in the svn
workflow, but git can do much more for you if you spend a little
time learning it.
git documentation central:
http://git-scm.com/documentation
git tutorial:
http://www.kernel.org/pub/software/scm/git/docs/gittutorial.html
Another crash course for svn users:
http://git-scm.org/course/svn.html
