user-manual: start revising "internals" chapter
[git/git.git] / Documentation / user-manual.txt
CommitLineData
d19fbc3c
BF
1Git User's Manual
2_________________
3
4This manual is designed to be readable by someone with basic unix
79c96c57 5command-line skills, but no previous knowledge of git.
d19fbc3c 6
ef89f701 7Chapter 1 gives a brief overview of git commands, without any
b181d57f 8explanation; you may prefer to skip to chapter 2 on a first reading.
ef89f701
BF
9
10Chapters 2 and 3 explain how to fetch and study a project using
d5cd5de4
BF
11git--the tools you'd need to build and test a particular version of a
12software project, to search for regressions, and so on.
6bd9b682 13
ef89f701 14Chapter 4 explains how to do development with git, and chapter 5 how
d5cd5de4 15to share that development with others.
6bd9b682
BF
16
17Further chapters cover more specialized topics.
18
d19fbc3c
BF
19Comprehensive reference documentation is available through the man
20pages. For a command such as "git clone", just use
21
22------------------------------------------------
23$ man git-clone
24------------------------------------------------
25
ef89f701
BF
26Git Quick Start
27===============
28
29This is a quick summary of the major commands; the following chapters
30will explain how these work in more detail.
31
32Creating a new repository
33-------------------------
34
35From a tarball:
36
37-----------------------------------------------
38$ tar xzf project.tar.gz
39$ cd project
40$ git init
41Initialized empty Git repository in .git/
42$ git add .
43$ git commit
44-----------------------------------------------
45
46From a remote repository:
47
48-----------------------------------------------
49$ git clone git://example.com/pub/project.git
50$ cd project
51-----------------------------------------------
52
53Managing branches
54-----------------
55
56-----------------------------------------------
57$ git branch # list all branches in this repo
58$ git checkout test # switch working directory to branch "test"
59$ git branch new # create branch "new" starting at current HEAD
60$ git branch -d new # delete branch "new"
61-----------------------------------------------
62
63Instead of basing new branch on current HEAD (the default), use:
64
65-----------------------------------------------
66$ git branch new test # branch named "test"
67$ git branch new v2.6.15 # tag named v2.6.15
68$ git branch new HEAD^ # commit before the most recent
69$ git branch new HEAD^^ # commit before that
70$ git branch new test~10 # ten commits before tip of branch "test"
71-----------------------------------------------
72
73Create and switch to a new branch at the same time:
74
75-----------------------------------------------
76$ git checkout -b new v2.6.15
77-----------------------------------------------
78
79Update and examine branches from the repository you cloned from:
80
81-----------------------------------------------
82$ git fetch # update
83$ git branch -r # list
84 origin/master
85 origin/next
86 ...
04483524 87$ git checkout -b masterwork origin/master
ef89f701
BF
88-----------------------------------------------
89
90Fetch a branch from a different repository, and give it a new
91name in your repository:
92
93-----------------------------------------------
94$ git fetch git://example.com/project.git theirbranch:mybranch
95$ git fetch git://example.com/project.git v2.6.15:mybranch
96-----------------------------------------------
97
98Keep a list of repositories you work with regularly:
99
100-----------------------------------------------
101$ git remote add example git://example.com/project.git
b181d57f 102$ git remote # list remote repositories
ef89f701
BF
103example
104origin
b181d57f 105$ git remote show example # get details
ef89f701
BF
106* remote example
107 URL: git://example.com/project.git
108 Tracked remote branches
109 master next ...
b181d57f
BF
110$ git fetch example # update branches from example
111$ git branch -r # list all remote branches
ef89f701
BF
112-----------------------------------------------
113
114
115Exploring history
116-----------------
117
118-----------------------------------------------
119$ gitk # visualize and browse history
120$ git log # list all commits
121$ git log src/ # ...modifying src/
122$ git log v2.6.15..v2.6.16 # ...in v2.6.16, not in v2.6.15
123$ git log master..test # ...in branch test, not in branch master
124$ git log test..master # ...in branch master, but not in test
125$ git log test...master # ...in one branch, not in both
126$ git log -S'foo()' # ...where difference contain "foo()"
127$ git log --since="2 weeks ago"
128$ git log -p # show patches as well
129$ git show # most recent commit
130$ git diff v2.6.15..v2.6.16 # diff between two tagged versions
131$ git diff v2.6.15..HEAD # diff with current head
132$ git grep "foo()" # search working directory for "foo()"
133$ git grep v2.6.15 "foo()" # search old tree for "foo()"
134$ git show v2.6.15:a.txt # look at old version of a.txt
135-----------------------------------------------
136
b181d57f 137Search for regressions:
ef89f701
BF
138
139-----------------------------------------------
140$ git bisect start
141$ git bisect bad # current version is bad
142$ git bisect good v2.6.13-rc2 # last known good revision
143Bisecting: 675 revisions left to test after this
144 # test here, then:
145$ git bisect good # if this revision is good, or
146$ git bisect bad # if this revision is bad.
147 # repeat until done.
148-----------------------------------------------
149
150Making changes
151--------------
152
153Make sure git knows who to blame:
154
155------------------------------------------------
156$ cat >~/.gitconfig <<\EOF
157[user]
04483524
JN
158 name = Your Name Comes Here
159 email = you@yourdomain.example.com
ef89f701
BF
160EOF
161------------------------------------------------
162
163Select file contents to include in the next commit, then make the
164commit:
165
166-----------------------------------------------
167$ git add a.txt # updated file
168$ git add b.txt # new file
169$ git rm c.txt # old file
170$ git commit
171-----------------------------------------------
172
173Or, prepare and create the commit in one step:
174
175-----------------------------------------------
b181d57f 176$ git commit d.txt # use latest content only of d.txt
ef89f701
BF
177$ git commit -a # use latest content of all tracked files
178-----------------------------------------------
179
180Merging
181-------
182
183-----------------------------------------------
184$ git merge test # merge branch "test" into the current branch
185$ git pull git://example.com/project.git master
186 # fetch and merge in remote branch
187$ git pull . test # equivalent to git merge test
188-----------------------------------------------
189
e4add70c
BF
190Sharing your changes
191--------------------
ef89f701
BF
192
193Importing or exporting patches:
194
195-----------------------------------------------
196$ git format-patch origin..HEAD # format a patch for each commit
197 # in HEAD but not in origin
04483524 198$ git am mbox # import patches from the mailbox "mbox"
ef89f701
BF
199-----------------------------------------------
200
ef89f701
BF
201Fetch a branch in a different git repository, then merge into the
202current branch:
203
204-----------------------------------------------
205$ git pull git://example.com/project.git theirbranch
206-----------------------------------------------
207
208Store the fetched branch into a local branch before merging into the
209current branch:
210
211-----------------------------------------------
212$ git pull git://example.com/project.git theirbranch:mybranch
213-----------------------------------------------
214
e4add70c
BF
215After creating commits on a local branch, update the remote
216branch with your commits:
217
218-----------------------------------------------
219$ git push ssh://example.com/project.git mybranch:theirbranch
220-----------------------------------------------
221
222When remote and local branch are both named "test":
223
224-----------------------------------------------
225$ git push ssh://example.com/project.git test
226-----------------------------------------------
227
228Shortcut version for a frequently used remote repository:
229
230-----------------------------------------------
231$ git remote add example ssh://example.com/project.git
232$ git push example test
233-----------------------------------------------
234
b181d57f
BF
235Repository maintenance
236----------------------
237
238Check for corruption:
239
240-----------------------------------------------
04e50e94 241$ git fsck
b181d57f
BF
242-----------------------------------------------
243
244Recompress, remove unused cruft:
245
246-----------------------------------------------
247$ git gc
248-----------------------------------------------
249
d19fbc3c
BF
250Repositories and Branches
251=========================
252
253How to get a git repository
254---------------------------
255
256It will be useful to have a git repository to experiment with as you
257read this manual.
258
259The best way to get one is by using the gitlink:git-clone[1] command
260to download a copy of an existing repository for a project that you
261are interested in. If you don't already have a project in mind, here
262are some interesting examples:
263
264------------------------------------------------
265 # git itself (approx. 10MB download):
266$ git clone git://git.kernel.org/pub/scm/git/git.git
267 # the linux kernel (approx. 150MB download):
268$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
269------------------------------------------------
270
271The initial clone may be time-consuming for a large project, but you
272will only need to clone once.
273
274The clone command creates a new directory named after the project
275("git" or "linux-2.6" in the examples above). After you cd into this
276directory, you will see that it contains a copy of the project files,
277together with a special top-level directory named ".git", which
278contains all the information about the history of the project.
279
d5cd5de4 280In most of the following, examples will be taken from one of the two
d19fbc3c
BF
281repositories above.
282
283How to check out a different version of a project
284-------------------------------------------------
285
286Git is best thought of as a tool for storing the history of a
287collection of files. It stores the history as a compressed
288collection of interrelated snapshots (versions) of the project's
289contents.
290
81b6c950
BF
291A single git repository may contain multiple branches. It keeps track
292of them by keeping a list of <<def_head,heads>> which reference the
293latest version on each branch; the gitlink:git-branch[1] command shows
294you the list of branch heads:
d19fbc3c
BF
295
296------------------------------------------------
297$ git branch
298* master
299------------------------------------------------
300
4f752407
BF
301A freshly cloned repository contains a single branch head, by default
302named "master", with the working directory initialized to the state of
303the project referred to by that branch head.
d19fbc3c 304
81b6c950
BF
305Most projects also use <<def_tag,tags>>. Tags, like heads, are
306references into the project's history, and can be listed using the
d19fbc3c
BF
307gitlink:git-tag[1] command:
308
309------------------------------------------------
310$ git tag -l
311v2.6.11
312v2.6.11-tree
313v2.6.12
314v2.6.12-rc2
315v2.6.12-rc3
316v2.6.12-rc4
317v2.6.12-rc5
318v2.6.12-rc6
319v2.6.13
320...
321------------------------------------------------
322
fe4b3e59 323Tags are expected to always point at the same version of a project,
81b6c950 324while heads are expected to advance as development progresses.
fe4b3e59 325
81b6c950 326Create a new branch head pointing to one of these versions and check it
d19fbc3c
BF
327out using gitlink:git-checkout[1]:
328
329------------------------------------------------
330$ git checkout -b new v2.6.13
331------------------------------------------------
332
333The working directory then reflects the contents that the project had
334when it was tagged v2.6.13, and gitlink:git-branch[1] shows two
335branches, with an asterisk marking the currently checked-out branch:
336
337------------------------------------------------
338$ git branch
339 master
340* new
341------------------------------------------------
342
343If you decide that you'd rather see version 2.6.17, you can modify
344the current branch to point at v2.6.17 instead, with
345
346------------------------------------------------
347$ git reset --hard v2.6.17
348------------------------------------------------
349
81b6c950 350Note that if the current branch head was your only reference to a
d19fbc3c 351particular point in history, then resetting that branch may leave you
81b6c950
BF
352with no way to find the history it used to point to; so use this command
353carefully.
d19fbc3c
BF
354
355Understanding History: Commits
356------------------------------
357
358Every change in the history of a project is represented by a commit.
359The gitlink:git-show[1] command shows the most recent commit on the
360current branch:
361
362------------------------------------------------
363$ git show
364commit 2b5f6dcce5bf94b9b119e9ed8d537098ec61c3d2
365Author: Jamal Hadi Salim <hadi@cyberus.ca>
366Date: Sat Dec 2 22:22:25 2006 -0800
367
368 [XFRM]: Fix aevent structuring to be more complete.
369
370 aevents can not uniquely identify an SA. We break the ABI with this
371 patch, but consensus is that since it is not yet utilized by any
372 (known) application then it is fine (better do it now than later).
373
374 Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
375 Signed-off-by: David S. Miller <davem@davemloft.net>
376
377diff --git a/Documentation/networking/xfrm_sync.txt b/Documentation/networking/xfrm_sync.txt
378index 8be626f..d7aac9d 100644
379--- a/Documentation/networking/xfrm_sync.txt
380+++ b/Documentation/networking/xfrm_sync.txt
381@@ -47,10 +47,13 @@ aevent_id structure looks like:
382
383 struct xfrm_aevent_id {
384 struct xfrm_usersa_id sa_id;
385+ xfrm_address_t saddr;
386 __u32 flags;
387+ __u32 reqid;
388 };
389...
390------------------------------------------------
391
392As you can see, a commit shows who made the latest change, what they
393did, and why.
394
35121930
BF
395Every commit has a 40-hexdigit id, sometimes called the "object name" or the
396"SHA1 id", shown on the first line of the "git show" output. You can usually
397refer to a commit by a shorter name, such as a tag or a branch name, but this
398longer name can also be useful. Most importantly, it is a globally unique
399name for this commit: so if you tell somebody else the object name (for
400example in email), then you are guaranteed that name will refer to the same
401commit in their repository that it does in yours (assuming their repository
402has that commit at all). Since the object name is computed as a hash over the
403contents of the commit, you are guaranteed that the commit can never change
404without its name also changing.
405
406In fact, in <<git-internals>> we shall see that everything stored in git
407history, including file data and directory contents, is stored in an object
408with a name that is a hash of its contents.
d19fbc3c
BF
409
410Understanding history: commits, parents, and reachability
411~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
412
413Every commit (except the very first commit in a project) also has a
414parent commit which shows what happened before this commit.
415Following the chain of parents will eventually take you back to the
416beginning of the project.
417
418However, the commits do not form a simple list; git allows lines of
419development to diverge and then reconverge, and the point where two
420lines of development reconverge is called a "merge". The commit
421representing a merge can therefore have more than one parent, with
422each parent representing the most recent commit on one of the lines
423of development leading to that point.
424
425The best way to see how this works is using the gitlink:gitk[1]
426command; running gitk now on a git repository and looking for merge
427commits will help understand how the git organizes history.
428
429In the following, we say that commit X is "reachable" from commit Y
430if commit X is an ancestor of commit Y. Equivalently, you could say
431that Y is a descendent of X, or that there is a chain of parents
432leading from commit Y to commit X.
433
3dff5379
PR
434Understanding history: History diagrams
435~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
d19fbc3c
BF
436
437We will sometimes represent git history using diagrams like the one
438below. Commits are shown as "o", and the links between them with
439lines drawn with - / and \. Time goes left to right:
440
1dc71a91
BF
441
442................................................
d19fbc3c
BF
443 o--o--o <-- Branch A
444 /
445 o--o--o <-- master
446 \
447 o--o--o <-- Branch B
1dc71a91 448................................................
d19fbc3c
BF
449
450If we need to talk about a particular commit, the character "o" may
451be replaced with another letter or number.
452
453Understanding history: What is a branch?
454~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
455
81b6c950
BF
456When we need to be precise, we will use the word "branch" to mean a line
457of development, and "branch head" (or just "head") to mean a reference
458to the most recent commit on a branch. In the example above, the branch
459head named "A" is a pointer to one particular commit, but we refer to
460the line of three commits leading up to that point as all being part of
d19fbc3c
BF
461"branch A".
462
81b6c950
BF
463However, when no confusion will result, we often just use the term
464"branch" both for branches and for branch heads.
d19fbc3c
BF
465
466Manipulating branches
467---------------------
468
469Creating, deleting, and modifying branches is quick and easy; here's
470a summary of the commands:
471
472git branch::
473 list all branches
474git branch <branch>::
475 create a new branch named <branch>, referencing the same
476 point in history as the current branch
477git branch <branch> <start-point>::
478 create a new branch named <branch>, referencing
479 <start-point>, which may be specified any way you like,
480 including using a branch name or a tag name
481git branch -d <branch>::
482 delete the branch <branch>; if the branch you are deleting
483 points to a commit which is not reachable from this branch,
484 this command will fail with a warning.
485git branch -D <branch>::
486 even if the branch points to a commit not reachable
487 from the current branch, you may know that that commit
488 is still reachable from some other branch or tag. In that
489 case it is safe to use this command to force git to delete
490 the branch.
491git checkout <branch>::
492 make the current branch <branch>, updating the working
493 directory to reflect the version referenced by <branch>
494git checkout -b <new> <start-point>::
495 create a new branch <new> referencing <start-point>, and
496 check it out.
497
72a76c95
BF
498The special symbol "HEAD" can always be used to refer to the current
499branch. In fact, git uses a file named "HEAD" in the .git directory to
500remember which branch is current:
501
502------------------------------------------------
503$ cat .git/HEAD
504ref: refs/heads/master
505------------------------------------------------
506
507Examining an old version without creating a new branch
508------------------------------------------------------
509
510The git-checkout command normally expects a branch head, but will also
511accept an arbitrary commit; for example, you can check out the commit
512referenced by a tag:
513
514------------------------------------------------
515$ git checkout v2.6.17
516Note: moving to "v2.6.17" which isn't a local branch
517If you want to create a new branch from this checkout, you may do so
518(now or later) by using -b with the checkout command again. Example:
519 git checkout -b <new_branch_name>
520HEAD is now at 427abfa... Linux v2.6.17
521------------------------------------------------
522
523The HEAD then refers to the SHA1 of the commit instead of to a branch,
524and git branch shows that you are no longer on a branch:
525
526------------------------------------------------
527$ cat .git/HEAD
528427abfa28afedffadfca9dd8b067eb6d36bac53f
529git branch
530* (no branch)
531 master
532------------------------------------------------
533
534In this case we say that the HEAD is "detached".
535
536This can be an easy way to check out a particular version without having
537to make up a name for a new branch. However, keep in mind that when you
538switch away from the (for example, by checking out something else), you
539can lose track of what the HEAD used to point to.
d19fbc3c
BF
540
541Examining branches from a remote repository
542-------------------------------------------
543
544The "master" branch that was created at the time you cloned is a copy
545of the HEAD in the repository that you cloned from. That repository
546may also have had other branches, though, and your local repository
547keeps branches which track each of those remote branches, which you
548can view using the "-r" option to gitlink:git-branch[1]:
549
550------------------------------------------------
551$ git branch -r
552 origin/HEAD
553 origin/html
554 origin/maint
555 origin/man
556 origin/master
557 origin/next
558 origin/pu
559 origin/todo
560------------------------------------------------
561
562You cannot check out these remote-tracking branches, but you can
563examine them on a branch of your own, just as you would a tag:
564
565------------------------------------------------
566$ git checkout -b my-todo-copy origin/todo
567------------------------------------------------
568
569Note that the name "origin" is just the name that git uses by default
570to refer to the repository that you cloned from.
571
572[[how-git-stores-references]]
f60b9642
BF
573Naming branches, tags, and other references
574-------------------------------------------
d19fbc3c
BF
575
576Branches, remote-tracking branches, and tags are all references to
f60b9642
BF
577commits. All references are named with a slash-separated path name
578starting with "refs"; the names we've been using so far are actually
579shorthand:
d19fbc3c 580
f60b9642
BF
581 - The branch "test" is short for "refs/heads/test".
582 - The tag "v2.6.18" is short for "refs/tags/v2.6.18".
583 - "origin/master" is short for "refs/remotes/origin/master".
d19fbc3c 584
f60b9642
BF
585The full name is occasionally useful if, for example, there ever
586exists a tag and a branch with the same name.
d19fbc3c 587
f60b9642
BF
588As another useful shortcut, if the repository "origin" posesses only
589a single branch, you can refer to that branch as just "origin".
d19fbc3c 590
f60b9642
BF
591More generally, if you have defined a remote repository named
592"example", you can refer to the branch in that repository as
593"example". And for a repository with multiple branches, this will
594refer to the branch designated as the "HEAD" branch.
d19fbc3c
BF
595
596For the complete list of paths which git checks for references, and
f60b9642
BF
597the order it uses to decide which to choose when there are multiple
598references with the same shorthand name, see the "SPECIFYING
599REVISIONS" section of gitlink:git-rev-parse[1].
d19fbc3c
BF
600
601[[Updating-a-repository-with-git-fetch]]
602Updating a repository with git fetch
603------------------------------------
604
605Eventually the developer cloned from will do additional work in her
606repository, creating new commits and advancing the branches to point
607at the new commits.
608
609The command "git fetch", with no arguments, will update all of the
610remote-tracking branches to the latest version found in her
611repository. It will not touch any of your own branches--not even the
612"master" branch that was created for you on clone.
613
d5cd5de4
BF
614Fetching branches from other repositories
615-----------------------------------------
616
617You can also track branches from repositories other than the one you
618cloned from, using gitlink:git-remote[1]:
619
620-------------------------------------------------
621$ git remote add linux-nfs git://linux-nfs.org/pub/nfs-2.6.git
04483524 622$ git fetch linux-nfs
d5cd5de4
BF
623* refs/remotes/linux-nfs/master: storing branch 'master' ...
624 commit: bf81b46
625-------------------------------------------------
626
627New remote-tracking branches will be stored under the shorthand name
628that you gave "git remote add", in this case linux-nfs:
629
630-------------------------------------------------
631$ git branch -r
632linux-nfs/master
633origin/master
634-------------------------------------------------
635
636If you run "git fetch <remote>" later, the tracking branches for the
637named <remote> will be updated.
638
639If you examine the file .git/config, you will see that git has added
640a new stanza:
641
642-------------------------------------------------
643$ cat .git/config
644...
645[remote "linux-nfs"]
923642fe
BF
646 url = git://linux-nfs.org/pub/nfs-2.6.git
647 fetch = +refs/heads/*:refs/remotes/linux-nfs/*
d5cd5de4
BF
648...
649-------------------------------------------------
650
fc90c536
BF
651This is what causes git to track the remote's branches; you may modify
652or delete these configuration options by editing .git/config with a
653text editor. (See the "CONFIGURATION FILE" section of
654gitlink:git-config[1] for details.)
d5cd5de4 655
d19fbc3c
BF
656Exploring git history
657=====================
658
659Git is best thought of as a tool for storing the history of a
660collection of files. It does this by storing compressed snapshots of
661the contents of a file heirarchy, together with "commits" which show
662the relationships between these snapshots.
663
664Git provides extremely flexible and fast tools for exploring the
665history of a project.
666
aacd404e 667We start with one specialized tool that is useful for finding the
d19fbc3c
BF
668commit that introduced a bug into a project.
669
670How to use bisect to find a regression
671--------------------------------------
672
673Suppose version 2.6.18 of your project worked, but the version at
674"master" crashes. Sometimes the best way to find the cause of such a
675regression is to perform a brute-force search through the project's
676history to find the particular commit that caused the problem. The
677gitlink:git-bisect[1] command can help you do this:
678
679-------------------------------------------------
680$ git bisect start
681$ git bisect good v2.6.18
682$ git bisect bad master
683Bisecting: 3537 revisions left to test after this
684[65934a9a028b88e83e2b0f8b36618fe503349f8e] BLOCK: Make USB storage depend on SCSI rather than selecting it [try #6]
685-------------------------------------------------
686
687If you run "git branch" at this point, you'll see that git has
688temporarily moved you to a new branch named "bisect". This branch
689points to a commit (with commit id 65934...) that is reachable from
690v2.6.19 but not from v2.6.18. Compile and test it, and see whether
691it crashes. Assume it does crash. Then:
692
693-------------------------------------------------
694$ git bisect bad
695Bisecting: 1769 revisions left to test after this
696[7eff82c8b1511017ae605f0c99ac275a7e21b867] i2c-core: Drop useless bitmaskings
697-------------------------------------------------
698
699checks out an older version. Continue like this, telling git at each
700stage whether the version it gives you is good or bad, and notice
701that the number of revisions left to test is cut approximately in
702half each time.
703
704After about 13 tests (in this case), it will output the commit id of
705the guilty commit. You can then examine the commit with
706gitlink:git-show[1], find out who wrote it, and mail them your bug
707report with the commit id. Finally, run
708
709-------------------------------------------------
710$ git bisect reset
711-------------------------------------------------
712
713to return you to the branch you were on before and delete the
714temporary "bisect" branch.
715
716Note that the version which git-bisect checks out for you at each
717point is just a suggestion, and you're free to try a different
718version if you think it would be a good idea. For example,
719occasionally you may land on a commit that broke something unrelated;
720run
721
722-------------------------------------------------
04483524 723$ git bisect visualize
d19fbc3c
BF
724-------------------------------------------------
725
726which will run gitk and label the commit it chose with a marker that
727says "bisect". Chose a safe-looking commit nearby, note its commit
728id, and check it out with:
729
730-------------------------------------------------
731$ git reset --hard fb47ddb2db...
732-------------------------------------------------
733
734then test, run "bisect good" or "bisect bad" as appropriate, and
735continue.
736
737Naming commits
738--------------
739
740We have seen several ways of naming commits already:
741
d55ae921 742 - 40-hexdigit object name
d19fbc3c
BF
743 - branch name: refers to the commit at the head of the given
744 branch
745 - tag name: refers to the commit pointed to by the given tag
746 (we've seen branches and tags are special cases of
747 <<how-git-stores-references,references>>).
748 - HEAD: refers to the head of the current branch
749
eb6ae7f4 750There are many more; see the "SPECIFYING REVISIONS" section of the
aec053bb 751gitlink:git-rev-parse[1] man page for the complete list of ways to
d19fbc3c
BF
752name revisions. Some examples:
753
754-------------------------------------------------
d55ae921 755$ git show fb47ddb2 # the first few characters of the object name
d19fbc3c
BF
756 # are usually enough to specify it uniquely
757$ git show HEAD^ # the parent of the HEAD commit
758$ git show HEAD^^ # the grandparent
759$ git show HEAD~4 # the great-great-grandparent
760-------------------------------------------------
761
762Recall that merge commits may have more than one parent; by default,
763^ and ~ follow the first parent listed in the commit, but you can
764also choose:
765
766-------------------------------------------------
767$ git show HEAD^1 # show the first parent of HEAD
768$ git show HEAD^2 # show the second parent of HEAD
769-------------------------------------------------
770
771In addition to HEAD, there are several other special names for
772commits:
773
774Merges (to be discussed later), as well as operations such as
775git-reset, which change the currently checked-out commit, generally
776set ORIG_HEAD to the value HEAD had before the current operation.
777
778The git-fetch operation always stores the head of the last fetched
779branch in FETCH_HEAD. For example, if you run git fetch without
780specifying a local branch as the target of the operation
781
782-------------------------------------------------
783$ git fetch git://example.com/proj.git theirbranch
784-------------------------------------------------
785
786the fetched commits will still be available from FETCH_HEAD.
787
788When we discuss merges we'll also see the special name MERGE_HEAD,
789which refers to the other branch that we're merging in to the current
790branch.
791
aec053bb 792The gitlink:git-rev-parse[1] command is a low-level command that is
d55ae921
BF
793occasionally useful for translating some name for a commit to the object
794name for that commit:
aec053bb
BF
795
796-------------------------------------------------
797$ git rev-parse origin
798e05db0fd4f31dde7005f075a84f96b360d05984b
799-------------------------------------------------
800
d19fbc3c
BF
801Creating tags
802-------------
803
804We can also create a tag to refer to a particular commit; after
805running
806
807-------------------------------------------------
04483524 808$ git tag stable-1 1b2e1d63ff
d19fbc3c
BF
809-------------------------------------------------
810
811You can use stable-1 to refer to the commit 1b2e1d63ff.
812
813This creates a "lightweight" tag. If the tag is a tag you wish to
814share with others, and possibly sign cryptographically, then you
815should create a tag object instead; see the gitlink:git-tag[1] man
816page for details.
817
818Browsing revisions
819------------------
820
821The gitlink:git-log[1] command can show lists of commits. On its
822own, it shows all commits reachable from the parent commit; but you
823can also make more specific requests:
824
825-------------------------------------------------
826$ git log v2.5.. # commits since (not reachable from) v2.5
827$ git log test..master # commits reachable from master but not test
828$ git log master..test # ...reachable from test but not master
829$ git log master...test # ...reachable from either test or master,
830 # but not both
831$ git log --since="2 weeks ago" # commits from the last 2 weeks
832$ git log Makefile # commits which modify Makefile
833$ git log fs/ # ... which modify any file under fs/
834$ git log -S'foo()' # commits which add or remove any file data
835 # matching the string 'foo()'
836-------------------------------------------------
837
838And of course you can combine all of these; the following finds
839commits since v2.5 which touch the Makefile or any file under fs:
840
841-------------------------------------------------
842$ git log v2.5.. Makefile fs/
843-------------------------------------------------
844
845You can also ask git log to show patches:
846
847-------------------------------------------------
848$ git log -p
849-------------------------------------------------
850
851See the "--pretty" option in the gitlink:git-log[1] man page for more
852display options.
853
854Note that git log starts with the most recent commit and works
855backwards through the parents; however, since git history can contain
3dff5379 856multiple independent lines of development, the particular order that
d19fbc3c
BF
857commits are listed in may be somewhat arbitrary.
858
859Generating diffs
860----------------
861
862You can generate diffs between any two versions using
863gitlink:git-diff[1]:
864
865-------------------------------------------------
866$ git diff master..test
867-------------------------------------------------
868
869Sometimes what you want instead is a set of patches:
870
871-------------------------------------------------
872$ git format-patch master..test
873-------------------------------------------------
874
875will generate a file with a patch for each commit reachable from test
876but not from master. Note that if master also has commits which are
877not reachable from test, then the combined result of these patches
878will not be the same as the diff produced by the git-diff example.
879
880Viewing old file versions
881-------------------------
882
883You can always view an old version of a file by just checking out the
884correct revision first. But sometimes it is more convenient to be
885able to view an old version of a single file without checking
886anything out; this command does that:
887
888-------------------------------------------------
889$ git show v2.5:fs/locks.c
890-------------------------------------------------
891
892Before the colon may be anything that names a commit, and after it
893may be any path to a file tracked by git.
894
aec053bb
BF
895Examples
896--------
897
898Check whether two branches point at the same history
2f99710c 899~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
aec053bb
BF
900
901Suppose you want to check whether two branches point at the same point
902in history.
903
904-------------------------------------------------
905$ git diff origin..master
906-------------------------------------------------
907
69f7ad73
BF
908will tell you whether the contents of the project are the same at the
909two branches; in theory, however, it's possible that the same project
910contents could have been arrived at by two different historical
d55ae921 911routes. You could compare the object names:
aec053bb
BF
912
913-------------------------------------------------
914$ git rev-list origin
915e05db0fd4f31dde7005f075a84f96b360d05984b
916$ git rev-list master
917e05db0fd4f31dde7005f075a84f96b360d05984b
918-------------------------------------------------
919
69f7ad73
BF
920Or you could recall that the ... operator selects all commits
921contained reachable from either one reference or the other but not
922both: so
aec053bb
BF
923
924-------------------------------------------------
925$ git log origin...master
926-------------------------------------------------
927
928will return no commits when the two branches are equal.
929
b181d57f
BF
930Find first tagged version including a given fix
931~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
aec053bb 932
69f7ad73
BF
933Suppose you know that the commit e05db0fd fixed a certain problem.
934You'd like to find the earliest tagged release that contains that
935fix.
936
937Of course, there may be more than one answer--if the history branched
938after commit e05db0fd, then there could be multiple "earliest" tagged
939releases.
940
941You could just visually inspect the commits since e05db0fd:
942
943-------------------------------------------------
944$ gitk e05db0fd..
945-------------------------------------------------
946
b181d57f
BF
947Or you can use gitlink:git-name-rev[1], which will give the commit a
948name based on any tag it finds pointing to one of the commit's
949descendants:
950
951-------------------------------------------------
04483524 952$ git name-rev --tags e05db0fd
b181d57f
BF
953e05db0fd tags/v1.5.0-rc1^0~23
954-------------------------------------------------
955
956The gitlink:git-describe[1] command does the opposite, naming the
957revision using a tag on which the given commit is based:
958
959-------------------------------------------------
960$ git describe e05db0fd
04483524 961v1.5.0-rc0-260-ge05db0f
b181d57f
BF
962-------------------------------------------------
963
964but that may sometimes help you guess which tags might come after the
965given commit.
966
967If you just want to verify whether a given tagged version contains a
968given commit, you could use gitlink:git-merge-base[1]:
969
970-------------------------------------------------
971$ git merge-base e05db0fd v1.5.0-rc1
972e05db0fd4f31dde7005f075a84f96b360d05984b
973-------------------------------------------------
974
975The merge-base command finds a common ancestor of the given commits,
976and always returns one or the other in the case where one is a
977descendant of the other; so the above output shows that e05db0fd
978actually is an ancestor of v1.5.0-rc1.
979
980Alternatively, note that
981
982-------------------------------------------------
4a7979ca 983$ git log v1.5.0-rc1..e05db0fd
b181d57f
BF
984-------------------------------------------------
985
4a7979ca 986will produce empty output if and only if v1.5.0-rc1 includes e05db0fd,
b181d57f 987because it outputs only commits that are not reachable from v1.5.0-rc1.
aec053bb 988
4a7979ca
BF
989As yet another alternative, the gitlink:git-show-branch[1] command lists
990the commits reachable from its arguments with a display on the left-hand
991side that indicates which arguments that commit is reachable from. So,
992you can run something like
993
994-------------------------------------------------
995$ git show-branch e05db0fd v1.5.0-rc0 v1.5.0-rc1 v1.5.0-rc2
996! [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
997available
998 ! [v1.5.0-rc0] GIT v1.5.0 preview
999 ! [v1.5.0-rc1] GIT v1.5.0-rc1
1000 ! [v1.5.0-rc2] GIT v1.5.0-rc2
1001...
1002-------------------------------------------------
1003
1004then search for a line that looks like
1005
1006-------------------------------------------------
1007+ ++ [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
1008available
1009-------------------------------------------------
1010
1011Which shows that e05db0fd is reachable from itself, from v1.5.0-rc1, and
1012from v1.5.0-rc2, but not from v1.5.0-rc0.
1013
1014
d19fbc3c
BF
1015Developing with git
1016===================
1017
1018Telling git your name
1019---------------------
1020
1021Before creating any commits, you should introduce yourself to git. The
1022easiest way to do so is:
1023
1024------------------------------------------------
1025$ cat >~/.gitconfig <<\EOF
1026[user]
1027 name = Your Name Comes Here
1028 email = you@yourdomain.example.com
1029EOF
1030------------------------------------------------
1031
fc90c536
BF
1032(See the "CONFIGURATION FILE" section of gitlink:git-config[1] for
1033details on the configuration file.)
1034
d19fbc3c
BF
1035
1036Creating a new repository
1037-------------------------
1038
1039Creating a new repository from scratch is very easy:
1040
1041-------------------------------------------------
1042$ mkdir project
1043$ cd project
f1d2b477 1044$ git init
d19fbc3c
BF
1045-------------------------------------------------
1046
1047If you have some initial content (say, a tarball):
1048
1049-------------------------------------------------
1050$ tar -xzvf project.tar.gz
1051$ cd project
f1d2b477 1052$ git init
d19fbc3c
BF
1053$ git add . # include everything below ./ in the first commit:
1054$ git commit
1055-------------------------------------------------
1056
1057[[how-to-make-a-commit]]
ae25c67a 1058How to make a commit
d19fbc3c
BF
1059--------------------
1060
1061Creating a new commit takes three steps:
1062
1063 1. Making some changes to the working directory using your
1064 favorite editor.
1065 2. Telling git about your changes.
1066 3. Creating the commit using the content you told git about
1067 in step 2.
1068
1069In practice, you can interleave and repeat steps 1 and 2 as many
1070times as you want: in order to keep track of what you want committed
1071at step 3, git maintains a snapshot of the tree's contents in a
1072special staging area called "the index."
1073
01997b4a
BF
1074At the beginning, the content of the index will be identical to
1075that of the HEAD. The command "git diff --cached", which shows
1076the difference between the HEAD and the index, should therefore
1077produce no output at that point.
eb6ae7f4 1078
d19fbc3c
BF
1079Modifying the index is easy:
1080
1081To update the index with the new contents of a modified file, use
1082
1083-------------------------------------------------
1084$ git add path/to/file
1085-------------------------------------------------
1086
1087To add the contents of a new file to the index, use
1088
1089-------------------------------------------------
1090$ git add path/to/file
1091-------------------------------------------------
1092
eb6ae7f4 1093To remove a file from the index and from the working tree,
d19fbc3c
BF
1094
1095-------------------------------------------------
1096$ git rm path/to/file
1097-------------------------------------------------
1098
1099After each step you can verify that
1100
1101-------------------------------------------------
1102$ git diff --cached
1103-------------------------------------------------
1104
1105always shows the difference between the HEAD and the index file--this
1106is what you'd commit if you created the commit now--and that
1107
1108-------------------------------------------------
1109$ git diff
1110-------------------------------------------------
1111
1112shows the difference between the working tree and the index file.
1113
1114Note that "git add" always adds just the current contents of a file
1115to the index; further changes to the same file will be ignored unless
1116you run git-add on the file again.
1117
1118When you're ready, just run
1119
1120-------------------------------------------------
1121$ git commit
1122-------------------------------------------------
1123
1124and git will prompt you for a commit message and then create the new
3dff5379 1125commit. Check to make sure it looks like what you expected with
d19fbc3c
BF
1126
1127-------------------------------------------------
1128$ git show
1129-------------------------------------------------
1130
1131As a special shortcut,
1132
1133-------------------------------------------------
1134$ git commit -a
1135-------------------------------------------------
1136
1137will update the index with any files that you've modified or removed
1138and create a commit, all in one step.
1139
1140A number of commands are useful for keeping track of what you're
1141about to commit:
1142
1143-------------------------------------------------
1144$ git diff --cached # difference between HEAD and the index; what
1145 # would be commited if you ran "commit" now.
1146$ git diff # difference between the index file and your
1147 # working directory; changes that would not
1148 # be included if you ran "commit" now.
1149$ git status # a brief per-file summary of the above.
1150-------------------------------------------------
1151
ae25c67a 1152Creating good commit messages
d19fbc3c
BF
1153-----------------------------
1154
1155Though not required, it's a good idea to begin the commit message
1156with a single short (less than 50 character) line summarizing the
1157change, followed by a blank line and then a more thorough
1158description. Tools that turn commits into email, for example, use
1159the first line on the Subject line and the rest of the commit in the
1160body.
1161
ae25c67a 1162How to merge
d19fbc3c
BF
1163------------
1164
1165You can rejoin two diverging branches of development using
1166gitlink:git-merge[1]:
1167
1168-------------------------------------------------
1169$ git merge branchname
1170-------------------------------------------------
1171
1172merges the development in the branch "branchname" into the current
1173branch. If there are conflicts--for example, if the same file is
1174modified in two different ways in the remote branch and the local
1175branch--then you are warned; the output may look something like this:
1176
1177-------------------------------------------------
fabbd8f6
BF
1178$ git merge next
1179 100% (4/4) done
1180Auto-merged file.txt
d19fbc3c
BF
1181CONFLICT (content): Merge conflict in file.txt
1182Automatic merge failed; fix conflicts and then commit the result.
1183-------------------------------------------------
1184
1185Conflict markers are left in the problematic files, and after
1186you resolve the conflicts manually, you can update the index
1187with the contents and run git commit, as you normally would when
1188creating a new file.
1189
1190If you examine the resulting commit using gitk, you will see that it
1191has two parents, one pointing to the top of the current branch, and
1192one to the top of the other branch.
1193
1194In more detail:
1195
1196[[resolving-a-merge]]
1197Resolving a merge
1198-----------------
1199
1200When a merge isn't resolved automatically, git leaves the index and
1201the working tree in a special state that gives you all the
1202information you need to help resolve the merge.
1203
1204Files with conflicts are marked specially in the index, so until you
ef561ac7
BF
1205resolve the problem and update the index, gitlink:git-commit[1] will
1206fail:
d19fbc3c
BF
1207
1208-------------------------------------------------
1209$ git commit
1210file.txt: needs merge
1211-------------------------------------------------
1212
ef561ac7
BF
1213Also, gitlink:git-status[1] will list those files as "unmerged", and the
1214files with conflicts will have conflict markers added, like this:
1215
1216-------------------------------------------------
1217<<<<<<< HEAD:file.txt
1218Hello world
1219=======
1220Goodbye
1221>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1222-------------------------------------------------
1223
1224All you need to do is edit the files to resolve the conflicts, and then
1225
1226-------------------------------------------------
1227$ git add file.txt
1228$ git commit
1229-------------------------------------------------
1230
1231Note that the commit message will already be filled in for you with
1232some information about the merge. Normally you can just use this
1233default message unchanged, but you may add additional commentary of
1234your own if desired.
1235
1236The above is all you need to know to resolve a simple merge. But git
1237also provides more information to help resolve conflicts:
1238
1239Getting conflict-resolution help during a merge
1240~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
d19fbc3c
BF
1241
1242All of the changes that git was able to merge automatically are
1243already added to the index file, so gitlink:git-diff[1] shows only
ef561ac7 1244the conflicts. It uses an unusual syntax:
d19fbc3c
BF
1245
1246-------------------------------------------------
1247$ git diff
1248diff --cc file.txt
1249index 802992c,2b60207..0000000
1250--- a/file.txt
1251+++ b/file.txt
1252@@@ -1,1 -1,1 +1,5 @@@
1253++<<<<<<< HEAD:file.txt
1254 +Hello world
1255++=======
1256+ Goodbye
1257++>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1258-------------------------------------------------
1259
1260Recall that the commit which will be commited after we resolve this
1261conflict will have two parents instead of the usual one: one parent
1262will be HEAD, the tip of the current branch; the other will be the
1263tip of the other branch, which is stored temporarily in MERGE_HEAD.
1264
ef561ac7
BF
1265During the merge, the index holds three versions of each file. Each of
1266these three "file stages" represents a different version of the file:
1267
1268-------------------------------------------------
1269$ git show :1:file.txt # the file in a common ancestor of both branches
1270$ git show :2:file.txt # the version from HEAD, but including any
1271 # nonconflicting changes from MERGE_HEAD
1272$ git show :3:file.txt # the version from MERGE_HEAD, but including any
1273 # nonconflicting changes from HEAD.
1274-------------------------------------------------
1275
1276Since the stage 2 and stage 3 versions have already been updated with
1277nonconflicting changes, the only remaining differences between them are
1278the important ones; thus gitlink:git-diff[1] can use the information in
1279the index to show only those conflicts.
1280
1281The diff above shows the differences between the working-tree version of
1282file.txt and the stage 2 and stage 3 versions. So instead of preceding
1283each line by a single "+" or "-", it now uses two columns: the first
1284column is used for differences between the first parent and the working
1285directory copy, and the second for differences between the second parent
1286and the working directory copy. (See the "COMBINED DIFF FORMAT" section
1287of gitlink:git-diff-files[1] for a details of the format.)
1288
1289After resolving the conflict in the obvious way (but before updating the
1290index), the diff will look like:
d19fbc3c
BF
1291
1292-------------------------------------------------
1293$ git diff
1294diff --cc file.txt
1295index 802992c,2b60207..0000000
1296--- a/file.txt
1297+++ b/file.txt
1298@@@ -1,1 -1,1 +1,1 @@@
1299- Hello world
1300 -Goodbye
1301++Goodbye world
1302-------------------------------------------------
1303
1304This shows that our resolved version deleted "Hello world" from the
1305first parent, deleted "Goodbye" from the second parent, and added
1306"Goodbye world", which was previously absent from both.
1307
ef561ac7
BF
1308Some special diff options allow diffing the working directory against
1309any of these stages:
1310
1311-------------------------------------------------
1312$ git diff -1 file.txt # diff against stage 1
1313$ git diff --base file.txt # same as the above
1314$ git diff -2 file.txt # diff against stage 2
1315$ git diff --ours file.txt # same as the above
1316$ git diff -3 file.txt # diff against stage 3
1317$ git diff --theirs file.txt # same as the above.
1318-------------------------------------------------
1319
1320The gitlink:git-log[1] and gitk[1] commands also provide special help
1321for merges:
d19fbc3c
BF
1322
1323-------------------------------------------------
1324$ git log --merge
ef561ac7 1325$ gitk --merge
d19fbc3c
BF
1326-------------------------------------------------
1327
ef561ac7
BF
1328These will display all commits which exist only on HEAD or on
1329MERGE_HEAD, and which touch an unmerged file.
d19fbc3c 1330
ef561ac7 1331Each time you resolve the conflicts in a file and update the index:
d19fbc3c
BF
1332
1333-------------------------------------------------
1334$ git add file.txt
d19fbc3c
BF
1335-------------------------------------------------
1336
ef561ac7
BF
1337the different stages of that file will be "collapsed", after which
1338git-diff will (by default) no longer show diffs for that file.
d19fbc3c
BF
1339
1340[[undoing-a-merge]]
ae25c67a 1341Undoing a merge
d19fbc3c
BF
1342---------------
1343
1344If you get stuck and decide to just give up and throw the whole mess
1345away, you can always return to the pre-merge state with
1346
1347-------------------------------------------------
1348$ git reset --hard HEAD
1349-------------------------------------------------
1350
1351Or, if you've already commited the merge that you want to throw away,
1352
1353-------------------------------------------------
1c73bb0e 1354$ git reset --hard ORIG_HEAD
d19fbc3c
BF
1355-------------------------------------------------
1356
1357However, this last command can be dangerous in some cases--never
1358throw away a commit you have already committed if that commit may
1359itself have been merged into another branch, as doing so may confuse
1360further merges.
1361
1362Fast-forward merges
1363-------------------
1364
1365There is one special case not mentioned above, which is treated
1366differently. Normally, a merge results in a merge commit, with two
1367parents, one pointing at each of the two lines of development that
1368were merged.
1369
1370However, if one of the two lines of development is completely
1371contained within the other--so every commit present in the one is
1372already contained in the other--then git just performs a
1373<<fast-forwards,fast forward>>; the head of the current branch is
1374moved forward to point at the head of the merged-in branch, without
1375any new commits being created.
1376
b684f830
BF
1377Fixing mistakes
1378---------------
1379
1380If you've messed up the working tree, but haven't yet committed your
1381mistake, you can return the entire working tree to the last committed
1382state with
1383
1384-------------------------------------------------
1385$ git reset --hard HEAD
1386-------------------------------------------------
1387
1388If you make a commit that you later wish you hadn't, there are two
1389fundamentally different ways to fix the problem:
1390
1391 1. You can create a new commit that undoes whatever was done
1392 by the previous commit. This is the correct thing if your
1393 mistake has already been made public.
1394
1395 2. You can go back and modify the old commit. You should
1396 never do this if you have already made the history public;
1397 git does not normally expect the "history" of a project to
1398 change, and cannot correctly perform repeated merges from
1399 a branch that has had its history changed.
1400
1401Fixing a mistake with a new commit
1402~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1403
1404Creating a new commit that reverts an earlier change is very easy;
1405just pass the gitlink:git-revert[1] command a reference to the bad
1406commit; for example, to revert the most recent commit:
1407
1408-------------------------------------------------
1409$ git revert HEAD
1410-------------------------------------------------
1411
1412This will create a new commit which undoes the change in HEAD. You
1413will be given a chance to edit the commit message for the new commit.
1414
1415You can also revert an earlier change, for example, the next-to-last:
1416
1417-------------------------------------------------
1418$ git revert HEAD^
1419-------------------------------------------------
1420
1421In this case git will attempt to undo the old change while leaving
1422intact any changes made since then. If more recent changes overlap
1423with the changes to be reverted, then you will be asked to fix
1424conflicts manually, just as in the case of <<resolving-a-merge,
1425resolving a merge>>.
1426
365aa199 1427[[fixing-a-mistake-by-editing-history]]
b684f830
BF
1428Fixing a mistake by editing history
1429~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1430
1431If the problematic commit is the most recent commit, and you have not
1432yet made that commit public, then you may just
1433<<undoing-a-merge,destroy it using git-reset>>.
1434
1435Alternatively, you
1436can edit the working directory and update the index to fix your
1437mistake, just as if you were going to <<how-to-make-a-commit,create a
1438new commit>>, then run
1439
1440-------------------------------------------------
1441$ git commit --amend
1442-------------------------------------------------
1443
1444which will replace the old commit by a new commit incorporating your
1445changes, giving you a chance to edit the old commit message first.
1446
1447Again, you should never do this to a commit that may already have
1448been merged into another branch; use gitlink:git-revert[1] instead in
1449that case.
1450
1451It is also possible to edit commits further back in the history, but
1452this is an advanced topic to be left for
1453<<cleaning-up-history,another chapter>>.
1454
1455Checking out an old version of a file
1456~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1457
1458In the process of undoing a previous bad change, you may find it
1459useful to check out an older version of a particular file using
1460gitlink:git-checkout[1]. We've used git checkout before to switch
1461branches, but it has quite different behavior if it is given a path
1462name: the command
1463
1464-------------------------------------------------
1465$ git checkout HEAD^ path/to/file
1466-------------------------------------------------
1467
1468replaces path/to/file by the contents it had in the commit HEAD^, and
1469also updates the index to match. It does not change branches.
1470
1471If you just want to look at an old version of the file, without
1472modifying the working directory, you can do that with
1473gitlink:git-show[1]:
1474
1475-------------------------------------------------
ed4eb0d8 1476$ git show HEAD^:path/to/file
b684f830
BF
1477-------------------------------------------------
1478
1479which will display the given version of the file.
1480
d19fbc3c
BF
1481Ensuring good performance
1482-------------------------
1483
1484On large repositories, git depends on compression to keep the history
1485information from taking up to much space on disk or in memory.
1486
1487This compression is not performed automatically. Therefore you
17217090 1488should occasionally run gitlink:git-gc[1]:
d19fbc3c
BF
1489
1490-------------------------------------------------
1491$ git gc
1492-------------------------------------------------
1493
17217090
BF
1494to recompress the archive. This can be very time-consuming, so
1495you may prefer to run git-gc when you are not doing other work.
d19fbc3c 1496
11e016a3
BF
1497Ensuring reliability
1498--------------------
1499
1500Checking the repository for corruption
1501~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1502
1191ee18
BF
1503The gitlink:git-fsck[1] command runs a number of self-consistency checks
1504on the repository, and reports on any problems. This may take some
21dcb3b7
BF
1505time. The most common warning by far is about "dangling" objects:
1506
1507-------------------------------------------------
04e50e94 1508$ git fsck
21dcb3b7
BF
1509dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1510dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1511dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1512dangling blob 218761f9d90712d37a9c5e36f406f92202db07eb
1513dangling commit bf093535a34a4d35731aa2bd90fe6b176302f14f
1514dangling commit 8e4bec7f2ddaa268bef999853c25755452100f8e
1515dangling tree d50bb86186bf27b681d25af89d3b5b68382e4085
1516dangling tree b24c2473f1fd3d91352a624795be026d64c8841f
1517...
1518-------------------------------------------------
1519
1191ee18
BF
1520Dangling objects are objects that are harmless, but also unnecessary;
1521you can remove them at any time with gitlink:git-prune[1] or the --prune
1522option to gitlink:git-gc[1]:
21dcb3b7
BF
1523
1524-------------------------------------------------
1525$ git gc --prune
1526-------------------------------------------------
1527
1191ee18
BF
1528This may be time-consuming. Unlike most other git operations (including
1529git-gc when run without any options), it is not safe to prune while
1530other git operations are in progress in the same repository.
21dcb3b7 1531
988b27d3 1532For more about dangling objects, see <<dangling-objects>>.
21dcb3b7 1533
11e016a3
BF
1534
1535Recovering lost changes
1536~~~~~~~~~~~~~~~~~~~~~~~
1537
559e4d7a
BF
1538Reflogs
1539^^^^^^^
1540
1541Say you modify a branch with gitlink:git-reset[1] --hard, and then
1542realize that the branch was the only reference you had to that point in
1543history.
1544
1545Fortunately, git also keeps a log, called a "reflog", of all the
1546previous values of each branch. So in this case you can still find the
1547old history using, for example,
1548
1549-------------------------------------------------
1550$ git log master@{1}
1551-------------------------------------------------
1552
1553This lists the commits reachable from the previous version of the head.
1554This syntax can be used to with any git command that accepts a commit,
1555not just with git log. Some other examples:
1556
1557-------------------------------------------------
1558$ git show master@{2} # See where the branch pointed 2,
1559$ git show master@{3} # 3, ... changes ago.
1560$ gitk master@{yesterday} # See where it pointed yesterday,
1561$ gitk master@{"1 week ago"} # ... or last week
1562-------------------------------------------------
1563
1564The reflogs are kept by default for 30 days, after which they may be
036be17e 1565pruned. See gitlink:git-reflog[1] and gitlink:git-gc[1] to learn
559e4d7a
BF
1566how to control this pruning, and see the "SPECIFYING REVISIONS"
1567section of gitlink:git-rev-parse[1] for details.
1568
1569Note that the reflog history is very different from normal git history.
1570While normal history is shared by every repository that works on the
1571same project, the reflog history is not shared: it tells you only about
1572how the branches in your local repository have changed over time.
1573
1574Examining dangling objects
1575^^^^^^^^^^^^^^^^^^^^^^^^^^
1576
1577In some situations the reflog may not be able to save you. For
1578example, suppose you delete a branch, then realize you need the history
79c96c57 1579it contained. The reflog is also deleted; however, if you have not
559e4d7a
BF
1580yet pruned the repository, then you may still be able to find
1581the lost commits; run git-fsck and watch for output that mentions
1582"dangling commits":
1583
1584-------------------------------------------------
1585$ git fsck
1586dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1587dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1588dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1589...
1590-------------------------------------------------
1591
aacd404e 1592You can examine
559e4d7a
BF
1593one of those dangling commits with, for example,
1594
1595------------------------------------------------
1596$ gitk 7281251ddd --not --all
1597------------------------------------------------
1598
1599which does what it sounds like: it says that you want to see the commit
1600history that is described by the dangling commit(s), but not the
1601history that is described by all your existing branches and tags. Thus
1602you get exactly the history reachable from that commit that is lost.
1603(And notice that it might not be just one commit: we only report the
1604"tip of the line" as being dangling, but there might be a whole deep
79c96c57 1605and complex commit history that was dropped.)
559e4d7a
BF
1606
1607If you decide you want the history back, you can always create a new
1608reference pointing to it, for example, a new branch:
1609
1610------------------------------------------------
1611$ git branch recovered-branch 7281251ddd
1612------------------------------------------------
1613
11e016a3 1614
d19fbc3c 1615Sharing development with others
b684f830 1616===============================
d19fbc3c
BF
1617
1618[[getting-updates-with-git-pull]]
1619Getting updates with git pull
b684f830 1620-----------------------------
d19fbc3c
BF
1621
1622After you clone a repository and make a few changes of your own, you
1623may wish to check the original repository for updates and merge them
1624into your own work.
1625
1626We have already seen <<Updating-a-repository-with-git-fetch,how to
1627keep remote tracking branches up to date>> with gitlink:git-fetch[1],
1628and how to merge two branches. So you can merge in changes from the
1629original repository's master branch with:
1630
1631-------------------------------------------------
1632$ git fetch
1633$ git merge origin/master
1634-------------------------------------------------
1635
1636However, the gitlink:git-pull[1] command provides a way to do this in
1637one step:
1638
1639-------------------------------------------------
1640$ git pull origin master
1641-------------------------------------------------
1642
1643In fact, "origin" is normally the default repository to pull from,
1644and the default branch is normally the HEAD of the remote repository,
1645so often you can accomplish the above with just
1646
1647-------------------------------------------------
1648$ git pull
1649-------------------------------------------------
1650
1651See the descriptions of the branch.<name>.remote and
9d13bda3 1652branch.<name>.merge options in gitlink:git-config[1] to learn
d19fbc3c
BF
1653how to control these defaults depending on the current branch.
1654
1655In addition to saving you keystrokes, "git pull" also helps you by
1656producing a default commit message documenting the branch and
1657repository that you pulled from.
1658
1659(But note that no such commit will be created in the case of a
1660<<fast-forwards,fast forward>>; instead, your branch will just be
79c96c57 1661updated to point to the latest commit from the upstream branch.)
d19fbc3c 1662
1191ee18
BF
1663The git-pull command can also be given "." as the "remote" repository,
1664in which case it just merges in a branch from the current repository; so
4c63ff45
BF
1665the commands
1666
1667-------------------------------------------------
1668$ git pull . branch
1669$ git merge branch
1670-------------------------------------------------
1671
1672are roughly equivalent. The former is actually very commonly used.
1673
d19fbc3c 1674Submitting patches to a project
b684f830 1675-------------------------------
d19fbc3c
BF
1676
1677If you just have a few changes, the simplest way to submit them may
1678just be to send them as patches in email:
1679
036be17e 1680First, use gitlink:git-format-patch[1]; for example:
d19fbc3c
BF
1681
1682-------------------------------------------------
eb6ae7f4 1683$ git format-patch origin
d19fbc3c
BF
1684-------------------------------------------------
1685
1686will produce a numbered series of files in the current directory, one
1687for each patch in the current branch but not in origin/HEAD.
1688
1689You can then import these into your mail client and send them by
1690hand. However, if you have a lot to send at once, you may prefer to
1691use the gitlink:git-send-email[1] script to automate the process.
1692Consult the mailing list for your project first to determine how they
1693prefer such patches be handled.
1694
1695Importing patches to a project
b684f830 1696------------------------------
d19fbc3c
BF
1697
1698Git also provides a tool called gitlink:git-am[1] (am stands for
1699"apply mailbox"), for importing such an emailed series of patches.
1700Just save all of the patch-containing messages, in order, into a
1701single mailbox file, say "patches.mbox", then run
1702
1703-------------------------------------------------
eb6ae7f4 1704$ git am -3 patches.mbox
d19fbc3c
BF
1705-------------------------------------------------
1706
1707Git will apply each patch in order; if any conflicts are found, it
1708will stop, and you can fix the conflicts as described in
01997b4a
BF
1709"<<resolving-a-merge,Resolving a merge>>". (The "-3" option tells
1710git to perform a merge; if you would prefer it just to abort and
1711leave your tree and index untouched, you may omit that option.)
1712
1713Once the index is updated with the results of the conflict
1714resolution, instead of creating a new commit, just run
d19fbc3c
BF
1715
1716-------------------------------------------------
1717$ git am --resolved
1718-------------------------------------------------
1719
1720and git will create the commit for you and continue applying the
1721remaining patches from the mailbox.
1722
1723The final result will be a series of commits, one for each patch in
1724the original mailbox, with authorship and commit log message each
1725taken from the message containing each patch.
1726
1727[[setting-up-a-public-repository]]
1728Setting up a public repository
b684f830 1729------------------------------
d19fbc3c
BF
1730
1731Another way to submit changes to a project is to simply tell the
1732maintainer of that project to pull from your repository, exactly as
1733you did in the section "<<getting-updates-with-git-pull, Getting
1734updates with git pull>>".
1735
1736If you and maintainer both have accounts on the same machine, then
1737then you can just pull changes from each other's repositories
79c96c57
MC
1738directly; note that all of the commands (gitlink:git-clone[1],
1739git-fetch[1], git-pull[1], etc.) that accept a URL as an argument
21f13ee2 1740will also accept a local directory name; so, for example, you can
d19fbc3c
BF
1741use
1742
1743-------------------------------------------------
1744$ git clone /path/to/repository
1745$ git pull /path/to/other/repository
1746-------------------------------------------------
1747
1748If this sort of setup is inconvenient or impossible, another (more
1749common) option is to set up a public repository on a public server.
1750This also allows you to cleanly separate private work in progress
1751from publicly visible work.
1752
1753You will continue to do your day-to-day work in your personal
1754repository, but periodically "push" changes from your personal
1755repository into your public repository, allowing other developers to
1756pull from that repository. So the flow of changes, in a situation
1757where there is one other developer with a public repository, looks
1758like this:
1759
1760 you push
1761 your personal repo ------------------> your public repo
1762 ^ |
1763 | |
1764 | you pull | they pull
1765 | |
1766 | |
1767 | they push V
1768 their public repo <------------------- their repo
1769
1770Now, assume your personal repository is in the directory ~/proj. We
1771first create a new clone of the repository:
1772
1773-------------------------------------------------
1774$ git clone --bare proj-clone.git
1775-------------------------------------------------
1776
1777The resulting directory proj-clone.git will contains a "bare" git
1778repository--it is just the contents of the ".git" directory, without
1779a checked-out copy of a working directory.
1780
1781Next, copy proj-clone.git to the server where you plan to host the
1782public repository. You can use scp, rsync, or whatever is most
1783convenient.
1784
1785If somebody else maintains the public server, they may already have
1786set up a git service for you, and you may skip to the section
1787"<<pushing-changes-to-a-public-repository,Pushing changes to a public
1788repository>>", below.
1789
1790Otherwise, the following sections explain how to export your newly
1791created public repository:
1792
1793[[exporting-via-http]]
1794Exporting a git repository via http
b684f830 1795-----------------------------------
d19fbc3c
BF
1796
1797The git protocol gives better performance and reliability, but on a
1798host with a web server set up, http exports may be simpler to set up.
1799
1800All you need to do is place the newly created bare git repository in
1801a directory that is exported by the web server, and make some
1802adjustments to give web clients some extra information they need:
1803
1804-------------------------------------------------
1805$ mv proj.git /home/you/public_html/proj.git
1806$ cd proj.git
1807$ git update-server-info
1808$ chmod a+x hooks/post-update
1809-------------------------------------------------
1810
1811(For an explanation of the last two lines, see
1812gitlink:git-update-server-info[1], and the documentation
1813link:hooks.txt[Hooks used by git].)
1814
1815Advertise the url of proj.git. Anybody else should then be able to
1816clone or pull from that url, for example with a commandline like:
1817
1818-------------------------------------------------
1819$ git clone http://yourserver.com/~you/proj.git
1820-------------------------------------------------
1821
1822(See also
1823link:howto/setup-git-server-over-http.txt[setup-git-server-over-http]
1824for a slightly more sophisticated setup using WebDAV which also
1825allows pushing over http.)
1826
1827[[exporting-via-git]]
1828Exporting a git repository via the git protocol
b684f830 1829-----------------------------------------------
d19fbc3c
BF
1830
1831This is the preferred method.
1832
1833For now, we refer you to the gitlink:git-daemon[1] man page for
1834instructions. (See especially the examples section.)
1835
1836[[pushing-changes-to-a-public-repository]]
1837Pushing changes to a public repository
b684f830 1838--------------------------------------
d19fbc3c
BF
1839
1840Note that the two techniques outline above (exporting via
1841<<exporting-via-http,http>> or <<exporting-via-git,git>>) allow other
1842maintainers to fetch your latest changes, but they do not allow write
1843access, which you will need to update the public repository with the
1844latest changes created in your private repository.
1845
1846The simplest way to do this is using gitlink:git-push[1] and ssh; to
1847update the remote branch named "master" with the latest state of your
1848branch named "master", run
1849
1850-------------------------------------------------
1851$ git push ssh://yourserver.com/~you/proj.git master:master
1852-------------------------------------------------
1853
1854or just
1855
1856-------------------------------------------------
1857$ git push ssh://yourserver.com/~you/proj.git master
1858-------------------------------------------------
1859
1860As with git-fetch, git-push will complain if this does not result in
1861a <<fast-forwards,fast forward>>. Normally this is a sign of
1862something wrong. However, if you are sure you know what you're
1863doing, you may force git-push to perform the update anyway by
1864proceeding the branch name by a plus sign:
1865
1866-------------------------------------------------
1867$ git push ssh://yourserver.com/~you/proj.git +master
1868-------------------------------------------------
1869
1870As with git-fetch, you may also set up configuration options to
1871save typing; so, for example, after
1872
1873-------------------------------------------------
1874$ cat >.git/config <<EOF
1875[remote "public-repo"]
1876 url = ssh://yourserver.com/~you/proj.git
1877EOF
1878-------------------------------------------------
1879
1880you should be able to perform the above push with just
1881
1882-------------------------------------------------
1883$ git push public-repo master
1884-------------------------------------------------
1885
1886See the explanations of the remote.<name>.url, branch.<name>.remote,
9d13bda3 1887and remote.<name>.push options in gitlink:git-config[1] for
d19fbc3c
BF
1888details.
1889
1890Setting up a shared repository
b684f830 1891------------------------------
d19fbc3c
BF
1892
1893Another way to collaborate is by using a model similar to that
1894commonly used in CVS, where several developers with special rights
1895all push to and pull from a single shared repository. See
1896link:cvs-migration.txt[git for CVS users] for instructions on how to
1897set this up.
1898
b684f830
BF
1899Allow web browsing of a repository
1900----------------------------------
d19fbc3c 1901
a8cd1402
BF
1902The gitweb cgi script provides users an easy way to browse your
1903project's files and history without having to install git; see the file
04483524 1904gitweb/INSTALL in the git source tree for instructions on setting it up.
d19fbc3c 1905
b684f830
BF
1906Examples
1907--------
d19fbc3c 1908
b684f830 1909TODO: topic branches, typical roles as in everyday.txt, ?
d19fbc3c 1910
d19fbc3c 1911
d19fbc3c 1912[[cleaning-up-history]]
4c63ff45
BF
1913Rewriting history and maintaining patch series
1914==============================================
1915
1916Normally commits are only added to a project, never taken away or
1917replaced. Git is designed with this assumption, and violating it will
1918cause git's merge machinery (for example) to do the wrong thing.
1919
1920However, there is a situation in which it can be useful to violate this
1921assumption.
1922
1923Creating the perfect patch series
1924---------------------------------
1925
1926Suppose you are a contributor to a large project, and you want to add a
1927complicated feature, and to present it to the other developers in a way
1928that makes it easy for them to read your changes, verify that they are
1929correct, and understand why you made each change.
1930
b181d57f 1931If you present all of your changes as a single patch (or commit), they
79c96c57 1932may find that it is too much to digest all at once.
4c63ff45
BF
1933
1934If you present them with the entire history of your work, complete with
1935mistakes, corrections, and dead ends, they may be overwhelmed.
1936
1937So the ideal is usually to produce a series of patches such that:
1938
1939 1. Each patch can be applied in order.
1940
1941 2. Each patch includes a single logical change, together with a
1942 message explaining the change.
1943
1944 3. No patch introduces a regression: after applying any initial
1945 part of the series, the resulting project still compiles and
1946 works, and has no bugs that it didn't have before.
1947
1948 4. The complete series produces the same end result as your own
1949 (probably much messier!) development process did.
1950
b181d57f
BF
1951We will introduce some tools that can help you do this, explain how to
1952use them, and then explain some of the problems that can arise because
1953you are rewriting history.
4c63ff45
BF
1954
1955Keeping a patch series up to date using git-rebase
1956--------------------------------------------------
1957
79c96c57
MC
1958Suppose that you create a branch "mywork" on a remote-tracking branch
1959"origin", and create some commits on top of it:
4c63ff45
BF
1960
1961-------------------------------------------------
1962$ git checkout -b mywork origin
1963$ vi file.txt
1964$ git commit
1965$ vi otherfile.txt
1966$ git commit
1967...
1968-------------------------------------------------
1969
1970You have performed no merges into mywork, so it is just a simple linear
1971sequence of patches on top of "origin":
1972
1dc71a91 1973................................................
4c63ff45
BF
1974 o--o--o <-- origin
1975 \
1976 o--o--o <-- mywork
1dc71a91 1977................................................
4c63ff45
BF
1978
1979Some more interesting work has been done in the upstream project, and
1980"origin" has advanced:
1981
1dc71a91 1982................................................
4c63ff45
BF
1983 o--o--O--o--o--o <-- origin
1984 \
1985 a--b--c <-- mywork
1dc71a91 1986................................................
4c63ff45
BF
1987
1988At this point, you could use "pull" to merge your changes back in;
1989the result would create a new merge commit, like this:
1990
1dc71a91 1991................................................
4c63ff45
BF
1992 o--o--O--o--o--o <-- origin
1993 \ \
1994 a--b--c--m <-- mywork
1dc71a91 1995................................................
4c63ff45
BF
1996
1997However, if you prefer to keep the history in mywork a simple series of
1998commits without any merges, you may instead choose to use
1999gitlink:git-rebase[1]:
2000
2001-------------------------------------------------
2002$ git checkout mywork
2003$ git rebase origin
2004-------------------------------------------------
2005
b181d57f
BF
2006This will remove each of your commits from mywork, temporarily saving
2007them as patches (in a directory named ".dotest"), update mywork to
2008point at the latest version of origin, then apply each of the saved
2009patches to the new mywork. The result will look like:
4c63ff45
BF
2010
2011
1dc71a91 2012................................................
4c63ff45
BF
2013 o--o--O--o--o--o <-- origin
2014 \
2015 a'--b'--c' <-- mywork
1dc71a91 2016................................................
4c63ff45 2017
b181d57f
BF
2018In the process, it may discover conflicts. In that case it will stop
2019and allow you to fix the conflicts; after fixing conflicts, use "git
2020add" to update the index with those contents, and then, instead of
2021running git-commit, just run
4c63ff45
BF
2022
2023-------------------------------------------------
2024$ git rebase --continue
2025-------------------------------------------------
2026
2027and git will continue applying the rest of the patches.
2028
2029At any point you may use the --abort option to abort this process and
2030return mywork to the state it had before you started the rebase:
2031
2032-------------------------------------------------
2033$ git rebase --abort
2034-------------------------------------------------
2035
365aa199
BF
2036Modifying a single commit
2037-------------------------
2038
2039We saw in <<fixing-a-mistake-by-editing-history>> that you can replace the
2040most recent commit using
2041
2042-------------------------------------------------
2043$ git commit --amend
2044-------------------------------------------------
2045
2046which will replace the old commit by a new commit incorporating your
2047changes, giving you a chance to edit the old commit message first.
2048
2049You can also use a combination of this and gitlink:git-rebase[1] to edit
2050commits further back in your history. First, tag the problematic commit with
2051
2052-------------------------------------------------
2053$ git tag bad mywork~5
2054-------------------------------------------------
2055
2056(Either gitk or git-log may be useful for finding the commit.)
2057
2058Then check out a new branch at that commit, edit it, and rebase the rest of
2059the series on top of it:
2060
2061-------------------------------------------------
2062$ git checkout -b TMP bad
2063$ # make changes here and update the index
2064$ git commit --amend
2065$ git rebase --onto TMP bad mywork
2066-------------------------------------------------
2067
2068When you're done, you'll be left with mywork checked out, with the top patches
2069on mywork reapplied on top of the modified commit you created in TMP. You can
2070then clean up with
2071
2072-------------------------------------------------
2073$ git branch -d TMP
2074$ git tag -d bad
2075-------------------------------------------------
2076
2077Note that the immutable nature of git history means that you haven't really
2078"modified" existing commits; instead, you have replaced the old commits with
2079new commits having new object names.
2080
4c63ff45
BF
2081Reordering or selecting from a patch series
2082-------------------------------------------
2083
b181d57f
BF
2084Given one existing commit, the gitlink:git-cherry-pick[1] command
2085allows you to apply the change introduced by that commit and create a
2086new commit that records it. So, for example, if "mywork" points to a
2087series of patches on top of "origin", you might do something like:
2088
2089-------------------------------------------------
2090$ git checkout -b mywork-new origin
2091$ gitk origin..mywork &
2092-------------------------------------------------
2093
2094And browse through the list of patches in the mywork branch using gitk,
2095applying them (possibly in a different order) to mywork-new using
2096cherry-pick, and possibly modifying them as you go using commit
2097--amend.
2098
2099Another technique is to use git-format-patch to create a series of
2100patches, then reset the state to before the patches:
4c63ff45 2101
b181d57f
BF
2102-------------------------------------------------
2103$ git format-patch origin
2104$ git reset --hard origin
2105-------------------------------------------------
4c63ff45 2106
b181d57f
BF
2107Then modify, reorder, or eliminate patches as preferred before applying
2108them again with gitlink:git-am[1].
4c63ff45
BF
2109
2110Other tools
2111-----------
2112
b181d57f 2113There are numerous other tools, such as stgit, which exist for the
79c96c57 2114purpose of maintaining a patch series. These are outside of the scope of
b181d57f 2115this manual.
4c63ff45
BF
2116
2117Problems with rewriting history
2118-------------------------------
2119
b181d57f
BF
2120The primary problem with rewriting the history of a branch has to do
2121with merging. Suppose somebody fetches your branch and merges it into
2122their branch, with a result something like this:
2123
1dc71a91 2124................................................
b181d57f
BF
2125 o--o--O--o--o--o <-- origin
2126 \ \
2127 t--t--t--m <-- their branch:
1dc71a91 2128................................................
b181d57f
BF
2129
2130Then suppose you modify the last three commits:
2131
1dc71a91 2132................................................
b181d57f
BF
2133 o--o--o <-- new head of origin
2134 /
2135 o--o--O--o--o--o <-- old head of origin
1dc71a91 2136................................................
b181d57f
BF
2137
2138If we examined all this history together in one repository, it will
2139look like:
2140
1dc71a91 2141................................................
b181d57f
BF
2142 o--o--o <-- new head of origin
2143 /
2144 o--o--O--o--o--o <-- old head of origin
2145 \ \
2146 t--t--t--m <-- their branch:
1dc71a91 2147................................................
b181d57f
BF
2148
2149Git has no way of knowing that the new head is an updated version of
2150the old head; it treats this situation exactly the same as it would if
2151two developers had independently done the work on the old and new heads
2152in parallel. At this point, if someone attempts to merge the new head
2153in to their branch, git will attempt to merge together the two (old and
2154new) lines of development, instead of trying to replace the old by the
2155new. The results are likely to be unexpected.
2156
2157You may still choose to publish branches whose history is rewritten,
2158and it may be useful for others to be able to fetch those branches in
2159order to examine or test them, but they should not attempt to pull such
2160branches into their own work.
2161
2162For true distributed development that supports proper merging,
2163published branches should never be rewritten.
2164
2165Advanced branch management
2166==========================
4c63ff45 2167
b181d57f
BF
2168Fetching individual branches
2169----------------------------
2170
2171Instead of using gitlink:git-remote[1], you can also choose just
2172to update one branch at a time, and to store it locally under an
2173arbitrary name:
2174
2175-------------------------------------------------
2176$ git fetch origin todo:my-todo-work
2177-------------------------------------------------
2178
2179The first argument, "origin", just tells git to fetch from the
2180repository you originally cloned from. The second argument tells git
2181to fetch the branch named "todo" from the remote repository, and to
2182store it locally under the name refs/heads/my-todo-work.
2183
2184You can also fetch branches from other repositories; so
2185
2186-------------------------------------------------
2187$ git fetch git://example.com/proj.git master:example-master
2188-------------------------------------------------
2189
2190will create a new branch named "example-master" and store in it the
2191branch named "master" from the repository at the given URL. If you
2192already have a branch named example-master, it will attempt to
2193"fast-forward" to the commit given by example.com's master branch. So
2194next we explain what a fast-forward is:
2195
2196[[fast-forwards]]
2197Understanding git history: fast-forwards
2198----------------------------------------
2199
2200In the previous example, when updating an existing branch, "git
2201fetch" checks to make sure that the most recent commit on the remote
2202branch is a descendant of the most recent commit on your copy of the
2203branch before updating your copy of the branch to point at the new
2204commit. Git calls this process a "fast forward".
2205
2206A fast forward looks something like this:
2207
1dc71a91 2208................................................
b181d57f
BF
2209 o--o--o--o <-- old head of the branch
2210 \
2211 o--o--o <-- new head of the branch
1dc71a91 2212................................................
b181d57f
BF
2213
2214
2215In some cases it is possible that the new head will *not* actually be
2216a descendant of the old head. For example, the developer may have
2217realized she made a serious mistake, and decided to backtrack,
2218resulting in a situation like:
2219
1dc71a91 2220................................................
b181d57f
BF
2221 o--o--o--o--a--b <-- old head of the branch
2222 \
2223 o--o--o <-- new head of the branch
1dc71a91 2224................................................
b181d57f
BF
2225
2226In this case, "git fetch" will fail, and print out a warning.
2227
2228In that case, you can still force git to update to the new head, as
2229described in the following section. However, note that in the
2230situation above this may mean losing the commits labeled "a" and "b",
2231unless you've already created a reference of your own pointing to
2232them.
2233
2234Forcing git fetch to do non-fast-forward updates
2235------------------------------------------------
2236
2237If git fetch fails because the new head of a branch is not a
2238descendant of the old head, you may force the update with:
2239
2240-------------------------------------------------
2241$ git fetch git://example.com/proj.git +master:refs/remotes/example/master
2242-------------------------------------------------
2243
79c96c57 2244Note the addition of the "+" sign. Be aware that commits that the
b181d57f
BF
2245old version of example/master pointed at may be lost, as we saw in
2246the previous section.
2247
2248Configuring remote branches
2249---------------------------
2250
2251We saw above that "origin" is just a shortcut to refer to the
79c96c57 2252repository that you originally cloned from. This information is
b181d57f 2253stored in git configuration variables, which you can see using
9d13bda3 2254gitlink:git-config[1]:
b181d57f
BF
2255
2256-------------------------------------------------
9d13bda3 2257$ git config -l
b181d57f
BF
2258core.repositoryformatversion=0
2259core.filemode=true
2260core.logallrefupdates=true
2261remote.origin.url=git://git.kernel.org/pub/scm/git/git.git
2262remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*
2263branch.master.remote=origin
2264branch.master.merge=refs/heads/master
2265-------------------------------------------------
2266
2267If there are other repositories that you also use frequently, you can
2268create similar configuration options to save typing; for example,
2269after
2270
2271-------------------------------------------------
9d13bda3 2272$ git config remote.example.url git://example.com/proj.git
b181d57f
BF
2273-------------------------------------------------
2274
2275then the following two commands will do the same thing:
2276
2277-------------------------------------------------
2278$ git fetch git://example.com/proj.git master:refs/remotes/example/master
2279$ git fetch example master:refs/remotes/example/master
2280-------------------------------------------------
2281
2282Even better, if you add one more option:
2283
2284-------------------------------------------------
9d13bda3 2285$ git config remote.example.fetch master:refs/remotes/example/master
b181d57f
BF
2286-------------------------------------------------
2287
2288then the following commands will all do the same thing:
2289
2290-------------------------------------------------
2291$ git fetch git://example.com/proj.git master:ref/remotes/example/master
2292$ git fetch example master:ref/remotes/example/master
2293$ git fetch example example/master
2294$ git fetch example
2295-------------------------------------------------
2296
2297You can also add a "+" to force the update each time:
2298
2299-------------------------------------------------
9d13bda3 2300$ git config remote.example.fetch +master:ref/remotes/example/master
b181d57f
BF
2301-------------------------------------------------
2302
2303Don't do this unless you're sure you won't mind "git fetch" possibly
2304throwing away commits on mybranch.
2305
2306Also note that all of the above configuration can be performed by
2307directly editing the file .git/config instead of using
9d13bda3 2308gitlink:git-config[1].
b181d57f 2309
9d13bda3 2310See gitlink:git-config[1] for more details on the configuration
b181d57f 2311options mentioned above.
d19fbc3c 2312
d19fbc3c 2313
35121930 2314[[git-internals]]
d19fbc3c
BF
2315Git internals
2316=============
2317
a536b08b
BF
2318Git depends on two fundamental abstractions: the "object database", and
2319the "current directory cache" aka "index".
b181d57f
BF
2320
2321The Object Database
2322-------------------
2323
2324The object database is literally just a content-addressable collection
2325of objects. All objects are named by their content, which is
2326approximated by the SHA1 hash of the object itself. Objects may refer
2327to other objects (by referencing their SHA1 hash), and so you can
2328build up a hierarchy of objects.
2329
2330All objects have a statically determined "type" aka "tag", which is
2331determined at object creation time, and which identifies the format of
2332the object (i.e. how it is used, and how it can refer to other
2333objects). There are currently four different object types: "blob",
a536b08b 2334"tree", "commit", and "tag".
b181d57f 2335
a536b08b
BF
2336A <<def_blob_object,"blob" object>> cannot refer to any other object,
2337and is, as the name implies, a pure storage object containing some
2338user data. It is used to actually store the file data, i.e. a blob
2339object is associated with some particular version of some file.
b181d57f 2340
a536b08b
BF
2341A <<def_tree_object,"tree" object>> is an object that ties one or more
2342"blob" objects into a directory structure. In addition, a tree object
2343can refer to other tree objects, thus creating a directory hierarchy.
b181d57f 2344
a536b08b
BF
2345A <<def_commit_object,"commit" object>> ties such directory hierarchies
2346together into a <<def_DAG,directed acyclic graph>> of revisions - each
2347"commit" is associated with exactly one tree (the directory hierarchy at
2348the time of the commit). In addition, a "commit" refers to one or more
2349"parent" commit objects that describe the history of how we arrived at
2350that directory hierarchy.
b181d57f
BF
2351
2352As a special case, a commit object with no parents is called the "root"
2353object, and is the point of an initial project commit. Each project
2354must have at least one root, and while you can tie several different
2355root objects together into one project by creating a commit object which
2356has two or more separate roots as its ultimate parents, that's probably
2357just going to confuse people. So aim for the notion of "one root object
2358per project", even if git itself does not enforce that.
2359
a536b08b
BF
2360A <<def_tag_object,"tag" object>> symbolically identifies and can be
2361used to sign other objects. It contains the identifier and type of
2362another object, a symbolic name (of course!) and, optionally, a
2363signature.
b181d57f
BF
2364
2365Regardless of object type, all objects share the following
2366characteristics: they are all deflated with zlib, and have a header
2367that not only specifies their type, but also provides size information
2368about the data in the object. It's worth noting that the SHA1 hash
2369that is used to name the object is the hash of the original data
2370plus this header, so `sha1sum` 'file' does not match the object name
2371for 'file'.
2372(Historical note: in the dawn of the age of git the hash
2373was the sha1 of the 'compressed' object.)
2374
2375As a result, the general consistency of an object can always be tested
2376independently of the contents or the type of the object: all objects can
2377be validated by verifying that (a) their hashes match the content of the
2378file and (b) the object successfully inflates to a stream of bytes that
2379forms a sequence of <ascii type without space> + <space> + <ascii decimal
2380size> + <byte\0> + <binary object data>.
2381
2382The structured objects can further have their structure and
2383connectivity to other objects verified. This is generally done with
04e50e94 2384the `git-fsck` program, which generates a full dependency graph
b181d57f
BF
2385of all objects, and verifies their internal consistency (in addition
2386to just verifying their superficial consistency through the hash).
2387
2388The object types in some more detail:
2389
2390Blob Object
2391-----------
2392
2393A "blob" object is nothing but a binary blob of data, and doesn't
2394refer to anything else. There is no signature or any other
2395verification of the data, so while the object is consistent (it 'is'
2396indexed by its sha1 hash, so the data itself is certainly correct), it
2397has absolutely no other attributes. No name associations, no
2398permissions. It is purely a blob of data (i.e. normally "file
2399contents").
2400
2401In particular, since the blob is entirely defined by its data, if two
2402files in a directory tree (or in multiple different versions of the
2403repository) have the same contents, they will share the same blob
2404object. The object is totally independent of its location in the
2405directory tree, and renaming a file does not change the object that
2406file is associated with in any way.
2407
2408A blob is typically created when gitlink:git-update-index[1]
2409is run, and its data can be accessed by gitlink:git-cat-file[1].
2410
2411Tree Object
2412-----------
2413
2414The next hierarchical object type is the "tree" object. A tree object
2415is a list of mode/name/blob data, sorted by name. Alternatively, the
2416mode data may specify a directory mode, in which case instead of
2417naming a blob, that name is associated with another TREE object.
2418
2419Like the "blob" object, a tree object is uniquely determined by the
2420set contents, and so two separate but identical trees will always
2421share the exact same object. This is true at all levels, i.e. it's
2422true for a "leaf" tree (which does not refer to any other trees, only
2423blobs) as well as for a whole subdirectory.
2424
2425For that reason a "tree" object is just a pure data abstraction: it
2426has no history, no signatures, no verification of validity, except
2427that since the contents are again protected by the hash itself, we can
2428trust that the tree is immutable and its contents never change.
2429
2430So you can trust the contents of a tree to be valid, the same way you
2431can trust the contents of a blob, but you don't know where those
2432contents 'came' from.
2433
2434Side note on trees: since a "tree" object is a sorted list of
2435"filename+content", you can create a diff between two trees without
2436actually having to unpack two trees. Just ignore all common parts,
2437and your diff will look right. In other words, you can effectively
2438(and efficiently) tell the difference between any two random trees by
2439O(n) where "n" is the size of the difference, rather than the size of
2440the tree.
2441
2442Side note 2 on trees: since the name of a "blob" depends entirely and
2443exclusively on its contents (i.e. there are no names or permissions
2444involved), you can see trivial renames or permission changes by
2445noticing that the blob stayed the same. However, renames with data
2446changes need a smarter "diff" implementation.
2447
2448A tree is created with gitlink:git-write-tree[1] and
2449its data can be accessed by gitlink:git-ls-tree[1].
2450Two trees can be compared with gitlink:git-diff-tree[1].
2451
2452Commit Object
2453-------------
2454
2455The "commit" object is an object that introduces the notion of
2456history into the picture. In contrast to the other objects, it
2457doesn't just describe the physical state of a tree, it describes how
2458we got there, and why.
2459
2460A "commit" is defined by the tree-object that it results in, the
2461parent commits (zero, one or more) that led up to that point, and a
2462comment on what happened. Again, a commit is not trusted per se:
2463the contents are well-defined and "safe" due to the cryptographically
2464strong signatures at all levels, but there is no reason to believe
2465that the tree is "good" or that the merge information makes sense.
2466The parents do not have to actually have any relationship with the
2467result, for example.
2468
2469Note on commits: unlike real SCM's, commits do not contain
2470rename information or file mode change information. All of that is
2471implicit in the trees involved (the result tree, and the result trees
2472of the parents), and describing that makes no sense in this idiotic
2473file manager.
2474
2475A commit is created with gitlink:git-commit-tree[1] and
2476its data can be accessed by gitlink:git-cat-file[1].
2477
2478Trust
2479-----
2480
2481An aside on the notion of "trust". Trust is really outside the scope
2482of "git", but it's worth noting a few things. First off, since
2483everything is hashed with SHA1, you 'can' trust that an object is
2484intact and has not been messed with by external sources. So the name
2485of an object uniquely identifies a known state - just not a state that
2486you may want to trust.
2487
2488Furthermore, since the SHA1 signature of a commit refers to the
2489SHA1 signatures of the tree it is associated with and the signatures
2490of the parent, a single named commit specifies uniquely a whole set
2491of history, with full contents. You can't later fake any step of the
2492way once you have the name of a commit.
2493
2494So to introduce some real trust in the system, the only thing you need
2495to do is to digitally sign just 'one' special note, which includes the
2496name of a top-level commit. Your digital signature shows others
2497that you trust that commit, and the immutability of the history of
2498commits tells others that they can trust the whole history.
2499
2500In other words, you can easily validate a whole archive by just
2501sending out a single email that tells the people the name (SHA1 hash)
2502of the top commit, and digitally sign that email using something
2503like GPG/PGP.
2504
2505To assist in this, git also provides the tag object...
2506
2507Tag Object
2508----------
2509
2510Git provides the "tag" object to simplify creating, managing and
2511exchanging symbolic and signed tokens. The "tag" object at its
2512simplest simply symbolically identifies another object by containing
2513the sha1, type and symbolic name.
2514
2515However it can optionally contain additional signature information
2516(which git doesn't care about as long as there's less than 8k of
2517it). This can then be verified externally to git.
2518
2519Note that despite the tag features, "git" itself only handles content
2520integrity; the trust framework (and signature provision and
2521verification) has to come from outside.
2522
2523A tag is created with gitlink:git-mktag[1],
2524its data can be accessed by gitlink:git-cat-file[1],
2525and the signature can be verified by
2526gitlink:git-verify-tag[1].
2527
2528
2529The "index" aka "Current Directory Cache"
2530-----------------------------------------
2531
2532The index is a simple binary file, which contains an efficient
2533representation of a virtual directory content at some random time. It
2534does so by a simple array that associates a set of names, dates,
2535permissions and content (aka "blob") objects together. The cache is
2536always kept ordered by name, and names are unique (with a few very
2537specific rules) at any point in time, but the cache has no long-term
2538meaning, and can be partially updated at any time.
2539
2540In particular, the index certainly does not need to be consistent with
2541the current directory contents (in fact, most operations will depend on
2542different ways to make the index 'not' be consistent with the directory
2543hierarchy), but it has three very important attributes:
2544
2545'(a) it can re-generate the full state it caches (not just the
2546directory structure: it contains pointers to the "blob" objects so
2547that it can regenerate the data too)'
2548
2549As a special case, there is a clear and unambiguous one-way mapping
2550from a current directory cache to a "tree object", which can be
2551efficiently created from just the current directory cache without
2552actually looking at any other data. So a directory cache at any one
2553time uniquely specifies one and only one "tree" object (but has
2554additional data to make it easy to match up that tree object with what
2555has happened in the directory)
2556
2557'(b) it has efficient methods for finding inconsistencies between that
2558cached state ("tree object waiting to be instantiated") and the
2559current state.'
2560
2561'(c) it can additionally efficiently represent information about merge
2562conflicts between different tree objects, allowing each pathname to be
2563associated with sufficient information about the trees involved that
2564you can create a three-way merge between them.'
2565
79c96c57 2566Those are the ONLY three things that the directory cache does. It's a
b181d57f
BF
2567cache, and the normal operation is to re-generate it completely from a
2568known tree object, or update/compare it with a live tree that is being
2569developed. If you blow the directory cache away entirely, you generally
2570haven't lost any information as long as you have the name of the tree
2571that it described.
2572
2573At the same time, the index is at the same time also the
2574staging area for creating new trees, and creating a new tree always
2575involves a controlled modification of the index file. In particular,
2576the index file can have the representation of an intermediate tree that
2577has not yet been instantiated. So the index can be thought of as a
2578write-back cache, which can contain dirty information that has not yet
2579been written back to the backing store.
2580
2581
2582
2583The Workflow
2584------------
2585
2586Generally, all "git" operations work on the index file. Some operations
2587work *purely* on the index file (showing the current state of the
2588index), but most operations move data to and from the index file. Either
2589from the database or from the working directory. Thus there are four
2590main combinations:
2591
2592working directory -> index
2593~~~~~~~~~~~~~~~~~~~~~~~~~~
2594
2595You update the index with information from the working directory with
2596the gitlink:git-update-index[1] command. You
2597generally update the index information by just specifying the filename
2598you want to update, like so:
2599
2600-------------------------------------------------
2601$ git-update-index filename
2602-------------------------------------------------
2603
2604but to avoid common mistakes with filename globbing etc, the command
2605will not normally add totally new entries or remove old entries,
2606i.e. it will normally just update existing cache entries.
2607
2608To tell git that yes, you really do realize that certain files no
2609longer exist, or that new files should be added, you
2610should use the `--remove` and `--add` flags respectively.
2611
2612NOTE! A `--remove` flag does 'not' mean that subsequent filenames will
2613necessarily be removed: if the files still exist in your directory
2614structure, the index will be updated with their new status, not
2615removed. The only thing `--remove` means is that update-cache will be
2616considering a removed file to be a valid thing, and if the file really
2617does not exist any more, it will update the index accordingly.
2618
2619As a special case, you can also do `git-update-index --refresh`, which
2620will refresh the "stat" information of each index to match the current
2621stat information. It will 'not' update the object status itself, and
2622it will only update the fields that are used to quickly test whether
2623an object still matches its old backing store object.
2624
2625index -> object database
2626~~~~~~~~~~~~~~~~~~~~~~~~
2627
2628You write your current index file to a "tree" object with the program
2629
2630-------------------------------------------------
2631$ git-write-tree
2632-------------------------------------------------
2633
2634that doesn't come with any options - it will just write out the
2635current index into the set of tree objects that describe that state,
2636and it will return the name of the resulting top-level tree. You can
2637use that tree to re-generate the index at any time by going in the
2638other direction:
2639
2640object database -> index
2641~~~~~~~~~~~~~~~~~~~~~~~~
2642
2643You read a "tree" file from the object database, and use that to
2644populate (and overwrite - don't do this if your index contains any
2645unsaved state that you might want to restore later!) your current
2646index. Normal operation is just
2647
2648-------------------------------------------------
2649$ git-read-tree <sha1 of tree>
2650-------------------------------------------------
2651
2652and your index file will now be equivalent to the tree that you saved
2653earlier. However, that is only your 'index' file: your working
2654directory contents have not been modified.
2655
2656index -> working directory
2657~~~~~~~~~~~~~~~~~~~~~~~~~~
2658
2659You update your working directory from the index by "checking out"
2660files. This is not a very common operation, since normally you'd just
2661keep your files updated, and rather than write to your working
2662directory, you'd tell the index files about the changes in your
2663working directory (i.e. `git-update-index`).
2664
2665However, if you decide to jump to a new version, or check out somebody
2666else's version, or just restore a previous tree, you'd populate your
2667index file with read-tree, and then you need to check out the result
2668with
2669
2670-------------------------------------------------
2671$ git-checkout-index filename
2672-------------------------------------------------
2673
2674or, if you want to check out all of the index, use `-a`.
2675
2676NOTE! git-checkout-index normally refuses to overwrite old files, so
2677if you have an old version of the tree already checked out, you will
2678need to use the "-f" flag ('before' the "-a" flag or the filename) to
2679'force' the checkout.
2680
2681
2682Finally, there are a few odds and ends which are not purely moving
2683from one representation to the other:
2684
2685Tying it all together
2686~~~~~~~~~~~~~~~~~~~~~
2687
2688To commit a tree you have instantiated with "git-write-tree", you'd
2689create a "commit" object that refers to that tree and the history
2690behind it - most notably the "parent" commits that preceded it in
2691history.
2692
2693Normally a "commit" has one parent: the previous state of the tree
2694before a certain change was made. However, sometimes it can have two
2695or more parent commits, in which case we call it a "merge", due to the
2696fact that such a commit brings together ("merges") two or more
2697previous states represented by other commits.
2698
2699In other words, while a "tree" represents a particular directory state
2700of a working directory, a "commit" represents that state in "time",
2701and explains how we got there.
2702
2703You create a commit object by giving it the tree that describes the
2704state at the time of the commit, and a list of parents:
2705
2706-------------------------------------------------
2707$ git-commit-tree <tree> -p <parent> [-p <parent2> ..]
2708-------------------------------------------------
2709
2710and then giving the reason for the commit on stdin (either through
2711redirection from a pipe or file, or by just typing it at the tty).
2712
2713git-commit-tree will return the name of the object that represents
2714that commit, and you should save it away for later use. Normally,
2715you'd commit a new `HEAD` state, and while git doesn't care where you
2716save the note about that state, in practice we tend to just write the
2717result to the file pointed at by `.git/HEAD`, so that we can always see
2718what the last committed state was.
2719
2720Here is an ASCII art by Jon Loeliger that illustrates how
2721various pieces fit together.
2722
2723------------
2724
2725 commit-tree
2726 commit obj
2727 +----+
2728 | |
2729 | |
2730 V V
2731 +-----------+
2732 | Object DB |
2733 | Backing |
2734 | Store |
2735 +-----------+
2736 ^
2737 write-tree | |
2738 tree obj | |
2739 | | read-tree
2740 | | tree obj
2741 V
2742 +-----------+
2743 | Index |
2744 | "cache" |
2745 +-----------+
2746 update-index ^
2747 blob obj | |
2748 | |
2749 checkout-index -u | | checkout-index
2750 stat | | blob obj
2751 V
2752 +-----------+
2753 | Working |
2754 | Directory |
2755 +-----------+
2756
2757------------
2758
2759
2760Examining the data
2761------------------
2762
2763You can examine the data represented in the object database and the
2764index with various helper tools. For every object, you can use
2765gitlink:git-cat-file[1] to examine details about the
2766object:
2767
2768-------------------------------------------------
2769$ git-cat-file -t <objectname>
2770-------------------------------------------------
2771
2772shows the type of the object, and once you have the type (which is
2773usually implicit in where you find the object), you can use
2774
2775-------------------------------------------------
2776$ git-cat-file blob|tree|commit|tag <objectname>
2777-------------------------------------------------
2778
2779to show its contents. NOTE! Trees have binary content, and as a result
2780there is a special helper for showing that content, called
2781`git-ls-tree`, which turns the binary content into a more easily
2782readable form.
2783
2784It's especially instructive to look at "commit" objects, since those
2785tend to be small and fairly self-explanatory. In particular, if you
2786follow the convention of having the top commit name in `.git/HEAD`,
2787you can do
2788
2789-------------------------------------------------
2790$ git-cat-file commit HEAD
2791-------------------------------------------------
2792
2793to see what the top commit was.
2794
2795Merging multiple trees
d19fbc3c
BF
2796----------------------
2797
b181d57f
BF
2798Git helps you do a three-way merge, which you can expand to n-way by
2799repeating the merge procedure arbitrary times until you finally
2800"commit" the state. The normal situation is that you'd only do one
2801three-way merge (two parents), and commit it, but if you like to, you
2802can do multiple parents in one go.
2803
2804To do a three-way merge, you need the two sets of "commit" objects
2805that you want to merge, use those to find the closest common parent (a
2806third "commit" object), and then use those commit objects to find the
2807state of the directory ("tree" object) at these points.
2808
2809To get the "base" for the merge, you first look up the common parent
2810of two commits with
2811
2812-------------------------------------------------
2813$ git-merge-base <commit1> <commit2>
2814-------------------------------------------------
2815
2816which will return you the commit they are both based on. You should
2817now look up the "tree" objects of those commits, which you can easily
2818do with (for example)
2819
2820-------------------------------------------------
2821$ git-cat-file commit <commitname> | head -1
2822-------------------------------------------------
2823
2824since the tree object information is always the first line in a commit
2825object.
2826
1191ee18
BF
2827Once you know the three trees you are going to merge (the one "original"
2828tree, aka the common case, and the two "result" trees, aka the branches
2829you want to merge), you do a "merge" read into the index. This will
2830complain if it has to throw away your old index contents, so you should
b181d57f 2831make sure that you've committed those - in fact you would normally
1191ee18
BF
2832always do a merge against your last commit (which should thus match what
2833you have in your current index anyway).
b181d57f
BF
2834
2835To do the merge, do
2836
2837-------------------------------------------------
2838$ git-read-tree -m -u <origtree> <yourtree> <targettree>
2839-------------------------------------------------
2840
2841which will do all trivial merge operations for you directly in the
2842index file, and you can just write the result out with
2843`git-write-tree`.
2844
2845
2846Merging multiple trees, continued
2847---------------------------------
2848
2849Sadly, many merges aren't trivial. If there are files that have
2850been added.moved or removed, or if both branches have modified the
2851same file, you will be left with an index tree that contains "merge
2852entries" in it. Such an index tree can 'NOT' be written out to a tree
2853object, and you will have to resolve any such merge clashes using
2854other tools before you can write out the result.
2855
2856You can examine such index state with `git-ls-files --unmerged`
2857command. An example:
2858
2859------------------------------------------------
2860$ git-read-tree -m $orig HEAD $target
2861$ git-ls-files --unmerged
2862100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello.c
2863100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello.c
2864100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello.c
2865------------------------------------------------
2866
2867Each line of the `git-ls-files --unmerged` output begins with
2868the blob mode bits, blob SHA1, 'stage number', and the
2869filename. The 'stage number' is git's way to say which tree it
2870came from: stage 1 corresponds to `$orig` tree, stage 2 `HEAD`
2871tree, and stage3 `$target` tree.
2872
2873Earlier we said that trivial merges are done inside
2874`git-read-tree -m`. For example, if the file did not change
2875from `$orig` to `HEAD` nor `$target`, or if the file changed
2876from `$orig` to `HEAD` and `$orig` to `$target` the same way,
2877obviously the final outcome is what is in `HEAD`. What the
2878above example shows is that file `hello.c` was changed from
2879`$orig` to `HEAD` and `$orig` to `$target` in a different way.
2880You could resolve this by running your favorite 3-way merge
2881program, e.g. `diff3` or `merge`, on the blob objects from
2882these three stages yourself, like this:
2883
2884------------------------------------------------
2885$ git-cat-file blob 263414f... >hello.c~1
2886$ git-cat-file blob 06fa6a2... >hello.c~2
2887$ git-cat-file blob cc44c73... >hello.c~3
2888$ merge hello.c~2 hello.c~1 hello.c~3
2889------------------------------------------------
2890
2891This would leave the merge result in `hello.c~2` file, along
2892with conflict markers if there are conflicts. After verifying
2893the merge result makes sense, you can tell git what the final
2894merge result for this file is by:
2895
2896-------------------------------------------------
2897$ mv -f hello.c~2 hello.c
2898$ git-update-index hello.c
2899-------------------------------------------------
2900
2901When a path is in unmerged state, running `git-update-index` for
2902that path tells git to mark the path resolved.
2903
2904The above is the description of a git merge at the lowest level,
2905to help you understand what conceptually happens under the hood.
2906In practice, nobody, not even git itself, uses three `git-cat-file`
2907for this. There is `git-merge-index` program that extracts the
2908stages to temporary files and calls a "merge" script on it:
2909
2910-------------------------------------------------
2911$ git-merge-index git-merge-one-file hello.c
2912-------------------------------------------------
2913
207dfa07 2914and that is what higher level `git merge -s resolve` is implemented with.
b181d57f
BF
2915
2916How git stores objects efficiently: pack files
2917----------------------------------------------
2918
2919We've seen how git stores each object in a file named after the
2920object's SHA1 hash.
2921
2922Unfortunately this system becomes inefficient once a project has a
2923lot of objects. Try this on an old project:
2924
2925------------------------------------------------
2926$ git count-objects
29276930 objects, 47620 kilobytes
2928------------------------------------------------
2929
2930The first number is the number of objects which are kept in
2931individual files. The second is the amount of space taken up by
2932those "loose" objects.
2933
2934You can save space and make git faster by moving these loose objects in
2935to a "pack file", which stores a group of objects in an efficient
2936compressed format; the details of how pack files are formatted can be
2937found in link:technical/pack-format.txt[technical/pack-format.txt].
2938
2939To put the loose objects into a pack, just run git repack:
2940
2941------------------------------------------------
2942$ git repack
2943Generating pack...
2944Done counting 6020 objects.
2945Deltifying 6020 objects.
2946 100% (6020/6020) done
2947Writing 6020 objects.
2948 100% (6020/6020) done
2949Total 6020, written 6020 (delta 4070), reused 0 (delta 0)
2950Pack pack-3e54ad29d5b2e05838c75df582c65257b8d08e1c created.
2951------------------------------------------------
2952
2953You can then run
2954
2955------------------------------------------------
2956$ git prune
2957------------------------------------------------
2958
2959to remove any of the "loose" objects that are now contained in the
2960pack. This will also remove any unreferenced objects (which may be
2961created when, for example, you use "git reset" to remove a commit).
2962You can verify that the loose objects are gone by looking at the
2963.git/objects directory or by running
2964
2965------------------------------------------------
2966$ git count-objects
29670 objects, 0 kilobytes
2968------------------------------------------------
2969
2970Although the object files are gone, any commands that refer to those
2971objects will work exactly as they did before.
2972
2973The gitlink:git-gc[1] command performs packing, pruning, and more for
2974you, so is normally the only high-level command you need.
d19fbc3c 2975
21dcb3b7
BF
2976[[dangling-objects]]
2977Dangling objects
61b41790 2978----------------
21dcb3b7 2979
04e50e94 2980The gitlink:git-fsck[1] command will sometimes complain about dangling
21dcb3b7
BF
2981objects. They are not a problem.
2982
1191ee18
BF
2983The most common cause of dangling objects is that you've rebased a
2984branch, or you have pulled from somebody else who rebased a branch--see
2985<<cleaning-up-history>>. In that case, the old head of the original
2986branch still exists, as does obviously everything it pointed to. The
2987branch pointer itself just doesn't, since you replaced it with another
2988one.
2989
2990There are also other situations too that cause dangling objects. For
2991example, a "dangling blob" may arise because you did a "git add" of a
2992file, but then, before you actually committed it and made it part of the
2993bigger picture, you changed something else in that file and committed
2994that *updated* thing - the old state that you added originally ends up
2995not being pointed to by any commit or tree, so it's now a dangling blob
2996object.
2997
2998Similarly, when the "recursive" merge strategy runs, and finds that
2999there are criss-cross merges and thus more than one merge base (which is
3000fairly unusual, but it does happen), it will generate one temporary
3001midway tree (or possibly even more, if you had lots of criss-crossing
3002merges and more than two merge bases) as a temporary internal merge
3003base, and again, those are real objects, but the end result will not end
3004up pointing to them, so they end up "dangling" in your repository.
3005
3006Generally, dangling objects aren't anything to worry about. They can
3007even be very useful: if you screw something up, the dangling objects can
3008be how you recover your old tree (say, you did a rebase, and realized
3009that you really didn't want to - you can look at what dangling objects
3010you have, and decide to reset your head to some old dangling state).
21dcb3b7 3011
559e4d7a
BF
3012For commits, the most useful thing to do with dangling objects tends to
3013be to do a simple
21dcb3b7
BF
3014
3015------------------------------------------------
3016$ gitk <dangling-commit-sha-goes-here> --not --all
3017------------------------------------------------
3018
1191ee18
BF
3019For blobs and trees, you can't do the same, but you can examine them.
3020You can just do
21dcb3b7
BF
3021
3022------------------------------------------------
3023$ git show <dangling-blob/tree-sha-goes-here>
3024------------------------------------------------
3025
1191ee18
BF
3026to show what the contents of the blob were (or, for a tree, basically
3027what the "ls" for that directory was), and that may give you some idea
3028of what the operation was that left that dangling object.
21dcb3b7 3029
1191ee18
BF
3030Usually, dangling blobs and trees aren't very interesting. They're
3031almost always the result of either being a half-way mergebase (the blob
3032will often even have the conflict markers from a merge in it, if you
3033have had conflicting merges that you fixed up by hand), or simply
3034because you interrupted a "git fetch" with ^C or something like that,
3035leaving _some_ of the new objects in the object database, but just
3036dangling and useless.
21dcb3b7
BF
3037
3038Anyway, once you are sure that you're not interested in any dangling
3039state, you can just prune all unreachable objects:
3040
3041------------------------------------------------
3042$ git prune
3043------------------------------------------------
3044
1191ee18
BF
3045and they'll be gone. But you should only run "git prune" on a quiescent
3046repository - it's kind of like doing a filesystem fsck recovery: you
3047don't want to do that while the filesystem is mounted.
21dcb3b7 3048
04e50e94
BF
3049(The same is true of "git-fsck" itself, btw - but since
3050git-fsck never actually *changes* the repository, it just reports
3051on what it found, git-fsck itself is never "dangerous" to run.
21dcb3b7
BF
3052Running it while somebody is actually changing the repository can cause
3053confusing and scary messages, but it won't actually do anything bad. In
3054contrast, running "git prune" while somebody is actively changing the
3055repository is a *BAD* idea).
3056
d19fbc3c
BF
3057include::glossary.txt[]
3058
6bd9b682
BF
3059Notes and todo list for this manual
3060===================================
3061
3062This is a work in progress.
3063
3064The basic requirements:
2f99710c
BF
3065 - It must be readable in order, from beginning to end, by
3066 someone intelligent with a basic grasp of the unix
3067 commandline, but without any special knowledge of git. If
3068 necessary, any other prerequisites should be specifically
3069 mentioned as they arise.
3070 - Whenever possible, section headings should clearly describe
3071 the task they explain how to do, in language that requires
3072 no more knowledge than necessary: for example, "importing
3073 patches into a project" rather than "the git-am command"
6bd9b682 3074
d5cd5de4
BF
3075Think about how to create a clear chapter dependency graph that will
3076allow people to get to important topics without necessarily reading
3077everything in between.
d19fbc3c 3078
aacd404e
MC
3079Say something about .gitignore.
3080
d19fbc3c
BF
3081Scan Documentation/ for other stuff left out; in particular:
3082 howto's
d19fbc3c
BF
3083 some of technical/?
3084 hooks
0b375ab0 3085 list of commands in gitlink:git[1]
d19fbc3c
BF
3086
3087Scan email archives for other stuff left out
3088
3089Scan man pages to see if any assume more background than this manual
3090provides.
3091
2f99710c 3092Simplify beginning by suggesting disconnected head instead of
b181d57f 3093temporary branch creation?
d19fbc3c 3094
2f99710c
BF
3095Add more good examples. Entire sections of just cookbook examples
3096might be a good idea; maybe make an "advanced examples" section a
3097standard end-of-chapter section?
d19fbc3c
BF
3098
3099Include cross-references to the glossary, where appropriate.
3100
9a241220
BF
3101Document shallow clones? See draft 1.5.0 release notes for some
3102documentation.
3103
3dff5379 3104Add a section on working with other version control systems, including
9a241220
BF
3105CVS, Subversion, and just imports of series of release tarballs.
3106
a8cd1402 3107More details on gitweb?
0b375ab0
BF
3108
3109Write a chapter on using plumbing and writing scripts.