kernel hacker’s guide to git

from :http://linux.yyz.us/git-howto.html

This tutorial is a cookbook of recipes getting up and running with Linus's source code management (SCM) software, "git." Its targetted mainly at Linux kernel hackers, though others may find it useful.

Table of Contents

  • Getting Started
  • Basic tasks
  • Branches
  • Misc. debris
  • Further reading

    Getting Started

    Installing git

    git requires bootstrapping, since you must have git installed in order to check out git.git (git repository), and linux-2.6.git (kernel repository). You may find that your distribution already provides a usable version of git. If so, try that first.

    • Fedora 7 and later: The git-core package is available through the standard package repositories. Fedora Core 3 through 6: git-core package is in Fedora Extras.

      yum install git-core

    If your distro does not package git, you may download the latest stable release from:
    http://www.kernel.org/pub/software/scm/git/

    tarball build-deps: zlib, libcurl, libcrypto (openssl)

    install tarball:

    unpack && make prefix=/usr/local && sudo make prefix=/usr/local install

    After reading the rest of this document, come back and update your copy of git to the latest: git://git.kernel.org/pub/scm/git/git.git

    Download a linux kernel tree for the very first time
    $ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux-2.6

    NOTE: The kernel tree is very large. This constitutes downloading just over 300 megabytes of compressed data (as of Jun 2008).


    Basic Tasks

    Update local kernel tree to latest 2.6.x upstream ("fast-forward merge")
    $ cd linux-2.6
    $ git pull git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

    or more simply, to pull from the location from which you cloned:

    $ cd linux-2.6
    $ git pull

    to pull from the origin repository from which you originally cloned the tree.

    Undo all local modifications:
    $ git checkout -f
    Check in your own modifications (e.g. do some hacking, or apply a patch)
    # go to repository
    $ cd linux-2.6
    
    # make some modifications
    $ vi drivers/net/sk98lin/skdim.c
    
    # NOTE: Run 'git add' and 'git rm' if adding or removing files.
    
    # check in all modifications
    $ git commit -a
    Undo recent commits:

    Sometimes you have made a few commits, or just pulled a change, and simply want those commits to go away.

    $ cd my-kernel-tree-2.6
    $ git reset HEAD~2	# make last 2 commits disappear

    will "disappear" the top two commits. DO NOT do this, if anyone has downloaded a tree containing the commits you just eliminated.

    Note that this is quite different from git revert, which applies a reversed patch as an additional commit.

    List all changes in working dir, in diff format.

    Display changes since last 'git add' or 'git rm':

    $ git diff

    Display changes since last commit:

    $ git diff HEAD
    Obtain summary of all changes in working dir
    $ git status
    List all changeset descriptions
    $ git log

    The 'git log' option "-p" shows diffs in addition to changeset text. The option "--stat" shows the diffstat in addition to the changeset text.

    List all changesets belonging to a specific file
    (in this case, net/ieee80211/ieee80211_module.c)
    $ git log net/ieee80211/ieee80211_module.c

    Branches

    List all branches
    $ git branch
    Make desired branch current in working directory
    $ git checkout $branch
    Create a new branch, and make it current
    $ git checkout -b my-new-branch-name master
    Examine which branch is current
    $ git status

    ('git branch' also shows you the current branch, using a "*" to indicate this)

    Obtain a diff between current branch, and master branch

    In most trees with branches, .git/refs/heads/master contains the current 'vanilla' upstream tree, for easy diffing and merging. (in trees without branches, 'master' simply contains your latest changes)

    $ git diff master..HEAD

    (this is equivalent to git diff HEAD, when used with HEAD branch)

    Obtain a list of changes between current branch, and master branch
    $ git log master..HEAD

    (this is equivalent to git log, when used with HEAD)

    or rather than full changeset descriptions, obtain a one-line summary of each changes:

    $ git shortlog master..HEAD
    Merge changes from one branch into another
    Let us suppose that you do work on branch A and branch B, and after work on those two branches is complete, you merge the work into mainline branch M.
    $ git checkout M	# switch to branch M
    $ git merge A		# merge A into M
    $ git merge B		# merge B into M

    Misc. Debris

    Optimize your repository

    git is heavily optimized for fast storage and retrieval on a per-command basis. However, over a long period of time, it can be useful to perform further optimizations, including packing all git objects into single "packfile" for fast retrieval and less wasted disk space.

    $ cd my-kernel-tree-2.6
    $ git gc

    will optimize your repository. You don't need to run this frequently — git is quite fast even without it. See the 'git gc' man page for more details.

    Check out an older kernel version
    $ cd my-kernel-tree-2.6
    $ git checkout -b tmp v2.6.22

    This creates a temporary branch 'tmp', with the contents of kernel version 2.6.22.

    Apply all patches in a Berkeley mbox-format file

    First, make sure that the tools subdirectory of the git-core repository is in your PATH.

    $ cd my-kernel-tree-2.6
    $ git am --utf8 --signoff /path/to/mbox

    The file /path/to/mbox is a Berkeley mbox file, containing one or more patches to be committed to the git repository. The --signoff option indicates that 'git am' should append the

    Signed-off-by: Your Name <your@email.com>

    line that is common to almost all kernel submissions. The name and email address are taken from the GIT_COMMITTER_NAME and GIT_COMMITTER_EMAIL environment variables (I recommend setting these in your .bash_profile or similar file).

    Don't forget to download tags from time to time.

    git pull only downloads sha1-indexed object data, and the requested remote head. This misses updates to the .git/refs/tags/ and .git/refs/heads/ directories. For tags, run git fetch --tags $URL.

    Tag a particular commit.

    For your own repositories, you may wish to give interesting or significant commits a name, known as a tag. The Linux kernel uses tags to for each kernel version: "v2.6.21", "v2.6.22", etc.

    $ cd my-kernel-tree-2.6
    $ git tag my-tag

    creates a new tag named "my-tag", based on the current commit. You can do a lot more with tagging, including GPG-signing, so read the man page for more details.


    Further reading

    Another good introduction is the official git tutorial, followed by the more in-depth git man page documentation.


posted @ 2011-03-27 16:00  冷月X  阅读(480)  评论(0编辑  收藏  举报