2007年5月19日星期六

Everyday GIT With 20 Commands

GIT suite 包含了 100 個以上的指令,而各指令的 manual page 中說明了該指令的功用以及使用的細節;但除非你知道用哪一個指令達成你想要做的事,不然壓根不知道要從哪一個 manual page 看起;反之,如果你已經知道要從哪個 manual 看起,那你大概也不需要看 manual 了。

難道這意指在使用 git 前,你必需瞭解每一個指令的用法嗎?其實不然。依據你的角色的不同,需要瞭解的指令也有些許的差異;但是不論扮演什麼樣的角色,你只要學一小部份的指令,就已足夠每日所需。這份文件的目的是成為一份密技表,以及為各種角色提供一個 git 的入口。

只要你有 repository,你就會需要 [Basic Repository] 中提到的指令 --- 其實這意指每個人都要看,因為 git 的每一個 working tree 都是一個 repository。

再來,只要有 commit 的需求,[Individual Developer (Standalone)] 中提到的指令就是你需要的,不論這個 repository 是否有提供他人存取。

如果你和他人合作,你也會需要 [Individual Developer (Participant)] 中提到的指令。

擔任 [Integrator] 角色的朋友,除了上面提到的以外,還需要多學一點。

[Repository Administration] 中提到的,則是給那些負責協助開發人員維護及交流 git repositories 的系統管理員所需要知道的。

Basic Repository

我們用這些指令來維護及操作 git repositories。

Examples

檢查 repository 是否有問題,並移除無用的部份
$ git fsck-objects (1)
$ git prune
$ git count-objects (2)
$ git repack (3)
$ git prune (4)
  1. 這裡我們沒有使用 "—full",一般來說這比較節省執行時使用的資源,並且可以保證 repository 的健康程度在合理的範圍內。

  2. 檢查散落的 objects 數量,並且回報如果不把它們打包會浪費多少資源。

  3. 因為沒有用 "-a",git 會使用累進式打包。依照經驗,以累積的方式每 4-5MB 打包散落的 objects 可以有不錯的成效。

  4. 在打包後,使用 prune 來移除重覆的散落 objects。

Repack a small project into single pack.
$ git repack -a -d (1)
$ git prune
  1. pack all the objects reachable from the refs into one pack and remove unneeded other packs

Individual Developer (Standalone)

A standalone individual developer does not exchange patches with other people, and works alone in a single repository, using the following commands.

Examples

Extract a tarball and create a working tree and a new repository to keep track of it.
$ tar zxf frotz.tar.gz
$ cd frotz
$ git-init-db
$ git add . (1)
$ git commit -m 'import of frotz source tree.'
$ git tag v2.43 (2)
  1. add everything under the current directory.

  2. make a lightweight, unannotated tag.

Create a topic branch and develop.
$ git checkout -b alsa-audio (1)
$ edit/compile/test
$ git checkout -- curses/ux_audio_oss.c (2)
$ git add curses/ux_audio_alsa.c (3)
$ edit/compile/test
$ git diff (4)
$ git commit -a -s (5)
$ edit/compile/test
$ git reset --soft HEAD^ (6)
$ edit/compile/test
$ git diff ORIG_HEAD (7)
$ git commit -a -c ORIG_HEAD (8)
$ git checkout master (9)
$ git pull . alsa-audio (10)
$ git log --since='3 days ago' (11)
$ git log v2.43.. curses/ (12)
  1. create a new topic branch.

  2. revert your botched changes in "curses/ux_audio_oss.c".

  3. you need to tell git if you added a new file; removal and modification will be caught if you do "commit -a" later.

  4. to see what changes you are committing.

  5. commit everything as you have tested, with your sign-off.

  6. take the last commit back, keeping what is in the working tree.

  7. look at the changes since the premature commit we took back.

  8. redo the commit undone in the previous step, using the message you originally wrote.

  9. switch to the master branch.

  10. merge a topic branch into your master branch

  11. review commit logs; other forms to limit output can be combined and include —max-count=10 (show 10 commits), —until=2005-12-10.

  12. view only the changes that touch what's in curses/ directory, since v2.43 tag.

Individual Developer (Participant)

A developer working as a participant in a group project needs to learn how to communicate with others, and uses these commands in addition to the ones needed by a standalone developer.

  • git-clone(1) from the upstream to prime your local repository.

  • git-pull(1) and git-fetch(1) from "origin" to keep up-to-date with the upstream.

  • git-push(1) to shared repository, if you adopt CVS style shared repository workflow.

  • git-format-patch(1) to prepare e-mail submission, if you adopt Linux kernel-style public forum workflow.

Examples

Clone the upstream and work on it. Feed changes to upstream.
$ git clone git://git.kernel.org/pub/scm/.../torvalds/linux-2.6 my2.6
$ cd my2.6
$ edit/compile/test; git commit -a -s (1)
$ git format-patch origin (2)
$ git pull (3)
$ git whatchanged -p ORIG_HEAD.. arch/i386 include/asm-i386 (4)
$ git pull git://git.kernel.org/pub/.../jgarzik/libata-dev.git ALL (5)
$ git reset --hard ORIG_HEAD (6)
$ git prune (7)
$ git fetch --tags (8)
  1. repeat as needed.

  2. extract patches from your branch for e-mail submission.

  3. "pull" fetches from "origin" by default and merges into the current branch.

  4. immediately after pulling, look at the changes done upstream since last time we checked, only in the area we are interested in.

  5. fetch from a specific branch from a specific repository and merge.

  6. revert the pull.

  7. garbage collect leftover objects from reverted pull.

  8. from time to time, obtain official tags from the "origin" and store them under .git/refs/tags/.

Push into another repository.
satellite$ git clone mothership:frotz/.git frotz (1)
satellite$ cd frotz
satellite$ cat .git/remotes/origin (2)
URL: mothership:frotz/.git
Pull: master:origin
satellite$ echo 'Push: master:satellite' >>.git/remotes/origin (3)
satellite$ edit/compile/test/commit
satellite$ git push origin (4)

mothership$ cd frotz
mothership$ git checkout master
mothership$ git pull . satellite (5)
  1. mothership machine has a frotz repository under your home directory; clone from it to start a repository on the satellite machine.

  2. clone creates this file by default. It arranges "git pull" to fetch and store the master branch head of mothership machine to local "origin" branch.

  3. arrange "git push" to push local "master" branch to "satellite" branch of the mothership machine.

  4. push will stash our work away on "satellite" branch on the mothership machine. You could use this as a back-up method.

  5. on mothership machine, merge the work done on the satellite machine into the master branch.

Branch off of a specific tag.
$ git checkout -b private2.6.14 v2.6.14 (1)
$ edit/compile/test; git commit -a
$ git checkout master
$ git format-patch -k -m --stdout v2.6.14..private2.6.14 |
git am -3 -k (2)
  1. create a private branch based on a well known (but somewhat behind) tag.

  2. forward port all changes in private2.6.14 branch to master branch without a formal "merging".

Integrator

A fairly central person acting as the integrator in a group project receives changes made by others, reviews and integrates them and publishes the result for others to use, using these commands in addition to the ones needed by participants.

Examples

My typical GIT day.
$ git status (1)
$ git show-branch (2)
$ mailx (3)
& s 2 3 4 5 ./+to-apply
& s 7 8 ./+hold-linus
& q
$ git checkout master
$ git am -3 -i -s -u ./+to-apply (4)
$ compile/test
$ git checkout -b hold/linus && git am -3 -i -s -u ./+hold-linus (5)
$ git checkout topic/one && git rebase master (6)
$ git checkout pu && git reset --hard master (7)
$ git pull . topic/one topic/two && git pull . hold/linus (8)
$ git checkout maint
$ git cherry-pick master~4 (9)
$ compile/test
$ git tag -s -m 'GIT 0.99.9x' v0.99.9x (10)
$ git fetch ko && git show-branch master maint 'tags/ko-*' (11)
$ git push ko (12)
$ git push ko v0.99.9x (13)
  1. see what I was in the middle of doing, if any.

  2. see what topic branches I have and think about how ready they are.

  3. read mails, save ones that are applicable, and save others that are not quite ready.

  4. apply them, interactively, with my sign-offs.

  5. create topic branch as needed and apply, again with my sign-offs.

  6. rebase internal topic branch that has not been merged to the master, nor exposed as a part of a stable branch.

  7. restart "pu" every time from the master.

  8. and bundle topic branches still cooking.

  9. backport a critical fix.

  10. create a signed tag.

  11. make sure I did not accidentally rewind master beyond what I already pushed out. "ko" shorthand points at the repository I have at kernel.org, and looks like this:

    $ cat .git/remotes/ko
    URL: kernel.org:/pub/scm/git/git.git
    Pull: master:refs/tags/ko-master
    Pull: maint:refs/tags/ko-maint
    Push: master
    Push: +pu
    Push: maint

    In the output from "git show-branch", "master" should have everything "ko-master" has.

  12. push out the bleeding edge.

  13. push the tag out, too.

Repository Administration

A repository administrator uses the following tools to set up and maintain access to the repository by developers.

  • git-daemon(1) to allow anonymous download from repository.

  • git-shell(1) can be used as a restricted login shell for shared central repository users.

update hook howto has a good example of managing a shared central repository.

Examples

Run git-daemon to serve /pub/scm from inetd.
$ grep git /etc/inetd.conf
git stream tcp nowait nobody \
/usr/bin/git-daemon git-daemon --inetd --syslog --export-all /pub/scm

The actual configuration line should be on one line.

Run git-daemon to serve /pub/scm from xinetd.
$ cat /etc/xinetd.d/git-daemon
# default: off
# description: The git server offers access to git repositories
service git
{
disable = no
type = UNLISTED
port = 9418
socket_type = stream
wait = no
user = nobody
server = /usr/bin/git-daemon
server_args = --inetd --syslog --export-all --base-path=/pub/scm
log_on_failure += USERID
}

Check your xinetd(8) documentation and setup, this is from a Fedora system. Others might be different.

Give push/pull only access to developers.
$ grep git /etc/passwd (1)
alice:x:1000:1000::/home/alice:/usr/bin/git-shell
bob:x:1001:1001::/home/bob:/usr/bin/git-shell
cindy:x:1002:1002::/home/cindy:/usr/bin/git-shell
david:x:1003:1003::/home/david:/usr/bin/git-shell
$ grep git /etc/shells (2)
/usr/bin/git-shell
  1. log-in shell is set to /usr/bin/git-shell, which does not allow anything but "git push" and "git pull". The users should get an ssh access to the machine.

  2. in many distributions /etc/shells needs to list what is used as the login shell.

CVS-style shared repository.
$ grep git /etc/group (1)
git:x:9418:alice,bob,cindy,david
$ cd /home/devo.git
$ ls -l (2)
lrwxrwxrwx 1 david git 17 Dec 4 22:40 HEAD -> refs/heads/master
drwxrwsr-x 2 david git 4096 Dec 4 22:40 branches
-rw-rw-r-- 1 david git 84 Dec 4 22:40 config
-rw-rw-r-- 1 david git 58 Dec 4 22:40 description
drwxrwsr-x 2 david git 4096 Dec 4 22:40 hooks
-rw-rw-r-- 1 david git 37504 Dec 4 22:40 index
drwxrwsr-x 2 david git 4096 Dec 4 22:40 info
drwxrwsr-x 4 david git 4096 Dec 4 22:40 objects
drwxrwsr-x 4 david git 4096 Nov 7 14:58 refs
drwxrwsr-x 2 david git 4096 Dec 4 22:40 remotes
$ ls -l hooks/update (3)
-r-xr-xr-x 1 david git 3536 Dec 4 22:40 update
$ cat info/allowed-users (4)
refs/heads/master alice\|cindy
refs/heads/doc-update bob
refs/tags/v[0-9]* david
  1. place the developers into the same git group.

  2. and make the shared repository writable by the group.

  3. use update-hook example by Carl from Documentation/howto/ for branch policy control.

  4. alice and cindy can push into master, only bob can push into doc-update. david is the release manager and is the only person who can create and push version tags.

HTTP server to support dumb protocol transfer.
dev$ git update-server-info (1)
dev$ ftp user@isp.example.com (2)
ftp> cp -r .git /home/user/myproject.git
  1. make sure your info/refs and objects/info/packs are up-to-date

  2. upload to public HTTP server hosted by your ISP.

没有评论: