1
Git User's Manual (for version 1.5.3 or newer)
2
______________________________________________
5
Git is a fast distributed revision control system.
7
This manual is designed to be readable by someone with basic UNIX
8
command-line skills, but no previous knowledge of git.
10
<<repositories-and-branches>> and <<exploring-git-history>> explain how
11
to fetch and study a project using git--read these chapters to learn how
12
to build and test a particular version of a software project, search for
13
regressions, and so on.
15
People needing to do actual development will also want to read
16
<<Developing-with-git>> and <<sharing-development>>.
18
Further chapters cover more specialized topics.
20
Comprehensive reference documentation is available through the man
21
pages. For a command such as "git clone", just use
23
------------------------------------------------
25
------------------------------------------------
27
See also <<git-quick-start>> for a brief overview of git commands,
28
without any explanation.
30
Finally, see <<todo>> for ways that you can help make this manual more
34
[[repositories-and-branches]]
35
Repositories and Branches
36
=========================
38
[[how-to-get-a-git-repository]]
39
How to get a git repository
40
---------------------------
42
It will be useful to have a git repository to experiment with as you
45
The best way to get one is by using the gitlink:git-clone[1] command to
46
download a copy of an existing repository. If you don't already have a
47
project in mind, here are some interesting examples:
49
------------------------------------------------
50
# git itself (approx. 10MB download):
51
$ git clone git://git.kernel.org/pub/scm/git/git.git
52
# the linux kernel (approx. 150MB download):
53
$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
54
------------------------------------------------
56
The initial clone may be time-consuming for a large project, but you
57
will only need to clone once.
59
The clone command creates a new directory named after the project
60
("git" or "linux-2.6" in the examples above). After you cd into this
61
directory, you will see that it contains a copy of the project files,
62
together with a special top-level directory named ".git", which
63
contains all the information about the history of the project.
66
How to check out a different version of a project
67
-------------------------------------------------
69
Git is best thought of as a tool for storing the history of a collection
70
of files. It stores the history as a compressed collection of
71
interrelated snapshots of the project's contents. In git each such
72
version is called a <<def_commit,commit>>.
74
A single git repository may contain multiple branches. It keeps track
75
of them by keeping a list of <<def_head,heads>> which reference the
76
latest commit on each branch; the gitlink:git-branch[1] command shows
77
you the list of branch heads:
79
------------------------------------------------
82
------------------------------------------------
84
A freshly cloned repository contains a single branch head, by default
85
named "master", with the working directory initialized to the state of
86
the project referred to by that branch head.
88
Most projects also use <<def_tag,tags>>. Tags, like heads, are
89
references into the project's history, and can be listed using the
90
gitlink:git-tag[1] command:
92
------------------------------------------------
104
------------------------------------------------
106
Tags are expected to always point at the same version of a project,
107
while heads are expected to advance as development progresses.
109
Create a new branch head pointing to one of these versions and check it
110
out using gitlink:git-checkout[1]:
112
------------------------------------------------
113
$ git checkout -b new v2.6.13
114
------------------------------------------------
116
The working directory then reflects the contents that the project had
117
when it was tagged v2.6.13, and gitlink:git-branch[1] shows two
118
branches, with an asterisk marking the currently checked-out branch:
120
------------------------------------------------
124
------------------------------------------------
126
If you decide that you'd rather see version 2.6.17, you can modify
127
the current branch to point at v2.6.17 instead, with
129
------------------------------------------------
130
$ git reset --hard v2.6.17
131
------------------------------------------------
133
Note that if the current branch head was your only reference to a
134
particular point in history, then resetting that branch may leave you
135
with no way to find the history it used to point to; so use this command
138
[[understanding-commits]]
139
Understanding History: Commits
140
------------------------------
142
Every change in the history of a project is represented by a commit.
143
The gitlink:git-show[1] command shows the most recent commit on the
146
------------------------------------------------
148
commit 17cf781661e6d38f737f15f53ab552f1e95960d7
149
Author: Linus Torvalds <torvalds@ppc970.osdl.org.(none)>
150
Date: Tue Apr 19 14:11:06 2005 -0700
152
Remove duplicate getenv(DB_ENVIRONMENT) call
156
diff --git a/init-db.c b/init-db.c
157
index 65898fa..b002dc6 100644
162
int main(int argc, char **argv)
164
- char *sha1_dir = getenv(DB_ENVIRONMENT), *path;
165
+ char *sha1_dir, *path;
168
if (mkdir(".git", 0755) < 0) {
169
------------------------------------------------
171
As you can see, a commit shows who made the latest change, what they
174
Every commit has a 40-hexdigit id, sometimes called the "object name" or the
175
"SHA1 id", shown on the first line of the "git show" output. You can usually
176
refer to a commit by a shorter name, such as a tag or a branch name, but this
177
longer name can also be useful. Most importantly, it is a globally unique
178
name for this commit: so if you tell somebody else the object name (for
179
example in email), then you are guaranteed that name will refer to the same
180
commit in their repository that it does in yours (assuming their repository
181
has that commit at all). Since the object name is computed as a hash over the
182
contents of the commit, you are guaranteed that the commit can never change
183
without its name also changing.
185
In fact, in <<git-concepts>> we shall see that everything stored in git
186
history, including file data and directory contents, is stored in an object
187
with a name that is a hash of its contents.
189
[[understanding-reachability]]
190
Understanding history: commits, parents, and reachability
191
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
193
Every commit (except the very first commit in a project) also has a
194
parent commit which shows what happened before this commit.
195
Following the chain of parents will eventually take you back to the
196
beginning of the project.
198
However, the commits do not form a simple list; git allows lines of
199
development to diverge and then reconverge, and the point where two
200
lines of development reconverge is called a "merge". The commit
201
representing a merge can therefore have more than one parent, with
202
each parent representing the most recent commit on one of the lines
203
of development leading to that point.
205
The best way to see how this works is using the gitlink:gitk[1]
206
command; running gitk now on a git repository and looking for merge
207
commits will help understand how the git organizes history.
209
In the following, we say that commit X is "reachable" from commit Y
210
if commit X is an ancestor of commit Y. Equivalently, you could say
211
that Y is a descendant of X, or that there is a chain of parents
212
leading from commit Y to commit X.
215
Understanding history: History diagrams
216
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
218
We will sometimes represent git history using diagrams like the one
219
below. Commits are shown as "o", and the links between them with
220
lines drawn with - / and \. Time goes left to right:
223
................................................
229
................................................
231
If we need to talk about a particular commit, the character "o" may
232
be replaced with another letter or number.
235
Understanding history: What is a branch?
236
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
238
When we need to be precise, we will use the word "branch" to mean a line
239
of development, and "branch head" (or just "head") to mean a reference
240
to the most recent commit on a branch. In the example above, the branch
241
head named "A" is a pointer to one particular commit, but we refer to
242
the line of three commits leading up to that point as all being part of
245
However, when no confusion will result, we often just use the term
246
"branch" both for branches and for branch heads.
248
[[manipulating-branches]]
249
Manipulating branches
250
---------------------
252
Creating, deleting, and modifying branches is quick and easy; here's
253
a summary of the commands:
257
git branch <branch>::
258
create a new branch named <branch>, referencing the same
259
point in history as the current branch
260
git branch <branch> <start-point>::
261
create a new branch named <branch>, referencing
262
<start-point>, which may be specified any way you like,
263
including using a branch name or a tag name
264
git branch -d <branch>::
265
delete the branch <branch>; if the branch you are deleting
266
points to a commit which is not reachable from the current
267
branch, this command will fail with a warning.
268
git branch -D <branch>::
269
even if the branch points to a commit not reachable
270
from the current branch, you may know that that commit
271
is still reachable from some other branch or tag. In that
272
case it is safe to use this command to force git to delete
274
git checkout <branch>::
275
make the current branch <branch>, updating the working
276
directory to reflect the version referenced by <branch>
277
git checkout -b <new> <start-point>::
278
create a new branch <new> referencing <start-point>, and
281
The special symbol "HEAD" can always be used to refer to the current
282
branch. In fact, git uses a file named "HEAD" in the .git directory to
283
remember which branch is current:
285
------------------------------------------------
287
ref: refs/heads/master
288
------------------------------------------------
291
Examining an old version without creating a new branch
292
------------------------------------------------------
294
The git-checkout command normally expects a branch head, but will also
295
accept an arbitrary commit; for example, you can check out the commit
298
------------------------------------------------
299
$ git checkout v2.6.17
300
Note: moving to "v2.6.17" which isn't a local branch
301
If you want to create a new branch from this checkout, you may do so
302
(now or later) by using -b with the checkout command again. Example:
303
git checkout -b <new_branch_name>
304
HEAD is now at 427abfa... Linux v2.6.17
305
------------------------------------------------
307
The HEAD then refers to the SHA1 of the commit instead of to a branch,
308
and git branch shows that you are no longer on a branch:
310
------------------------------------------------
312
427abfa28afedffadfca9dd8b067eb6d36bac53f
316
------------------------------------------------
318
In this case we say that the HEAD is "detached".
320
This is an easy way to check out a particular version without having to
321
make up a name for the new branch. You can still create a new branch
322
(or tag) for this version later if you decide to.
324
[[examining-remote-branches]]
325
Examining branches from a remote repository
326
-------------------------------------------
328
The "master" branch that was created at the time you cloned is a copy
329
of the HEAD in the repository that you cloned from. That repository
330
may also have had other branches, though, and your local repository
331
keeps branches which track each of those remote branches, which you
332
can view using the "-r" option to gitlink:git-branch[1]:
334
------------------------------------------------
344
------------------------------------------------
346
You cannot check out these remote-tracking branches, but you can
347
examine them on a branch of your own, just as you would a tag:
349
------------------------------------------------
350
$ git checkout -b my-todo-copy origin/todo
351
------------------------------------------------
353
Note that the name "origin" is just the name that git uses by default
354
to refer to the repository that you cloned from.
356
[[how-git-stores-references]]
357
Naming branches, tags, and other references
358
-------------------------------------------
360
Branches, remote-tracking branches, and tags are all references to
361
commits. All references are named with a slash-separated path name
362
starting with "refs"; the names we've been using so far are actually
365
- The branch "test" is short for "refs/heads/test".
366
- The tag "v2.6.18" is short for "refs/tags/v2.6.18".
367
- "origin/master" is short for "refs/remotes/origin/master".
369
The full name is occasionally useful if, for example, there ever
370
exists a tag and a branch with the same name.
372
(Newly created refs are actually stored in the .git/refs directory,
373
under the path given by their name. However, for efficiency reasons
374
they may also be packed together in a single file; see
375
gitlink:git-pack-refs[1]).
377
As another useful shortcut, the "HEAD" of a repository can be referred
378
to just using the name of that repository. So, for example, "origin"
379
is usually a shortcut for the HEAD branch in the repository "origin".
381
For the complete list of paths which git checks for references, and
382
the order it uses to decide which to choose when there are multiple
383
references with the same shorthand name, see the "SPECIFYING
384
REVISIONS" section of gitlink:git-rev-parse[1].
386
[[Updating-a-repository-with-git-fetch]]
387
Updating a repository with git fetch
388
------------------------------------
390
Eventually the developer cloned from will do additional work in her
391
repository, creating new commits and advancing the branches to point
394
The command "git fetch", with no arguments, will update all of the
395
remote-tracking branches to the latest version found in her
396
repository. It will not touch any of your own branches--not even the
397
"master" branch that was created for you on clone.
399
[[fetching-branches]]
400
Fetching branches from other repositories
401
-----------------------------------------
403
You can also track branches from repositories other than the one you
404
cloned from, using gitlink:git-remote[1]:
406
-------------------------------------------------
407
$ git remote add linux-nfs git://linux-nfs.org/pub/nfs-2.6.git
408
$ git fetch linux-nfs
409
* refs/remotes/linux-nfs/master: storing branch 'master' ...
411
-------------------------------------------------
413
New remote-tracking branches will be stored under the shorthand name
414
that you gave "git remote add", in this case linux-nfs:
416
-------------------------------------------------
420
-------------------------------------------------
422
If you run "git fetch <remote>" later, the tracking branches for the
423
named <remote> will be updated.
425
If you examine the file .git/config, you will see that git has added
428
-------------------------------------------------
432
url = git://linux-nfs.org/pub/nfs-2.6.git
433
fetch = +refs/heads/*:refs/remotes/linux-nfs/*
435
-------------------------------------------------
437
This is what causes git to track the remote's branches; you may modify
438
or delete these configuration options by editing .git/config with a
439
text editor. (See the "CONFIGURATION FILE" section of
440
gitlink:git-config[1] for details.)
442
[[exploring-git-history]]
443
Exploring git history
444
=====================
446
Git is best thought of as a tool for storing the history of a
447
collection of files. It does this by storing compressed snapshots of
448
the contents of a file hierarchy, together with "commits" which show
449
the relationships between these snapshots.
451
Git provides extremely flexible and fast tools for exploring the
452
history of a project.
454
We start with one specialized tool that is useful for finding the
455
commit that introduced a bug into a project.
458
How to use bisect to find a regression
459
--------------------------------------
461
Suppose version 2.6.18 of your project worked, but the version at
462
"master" crashes. Sometimes the best way to find the cause of such a
463
regression is to perform a brute-force search through the project's
464
history to find the particular commit that caused the problem. The
465
gitlink:git-bisect[1] command can help you do this:
467
-------------------------------------------------
469
$ git bisect good v2.6.18
470
$ git bisect bad master
471
Bisecting: 3537 revisions left to test after this
472
[65934a9a028b88e83e2b0f8b36618fe503349f8e] BLOCK: Make USB storage depend on SCSI rather than selecting it [try #6]
473
-------------------------------------------------
475
If you run "git branch" at this point, you'll see that git has
476
temporarily moved you to a new branch named "bisect". This branch
477
points to a commit (with commit id 65934...) that is reachable from
478
v2.6.19 but not from v2.6.18. Compile and test it, and see whether
479
it crashes. Assume it does crash. Then:
481
-------------------------------------------------
483
Bisecting: 1769 revisions left to test after this
484
[7eff82c8b1511017ae605f0c99ac275a7e21b867] i2c-core: Drop useless bitmaskings
485
-------------------------------------------------
487
checks out an older version. Continue like this, telling git at each
488
stage whether the version it gives you is good or bad, and notice
489
that the number of revisions left to test is cut approximately in
492
After about 13 tests (in this case), it will output the commit id of
493
the guilty commit. You can then examine the commit with
494
gitlink:git-show[1], find out who wrote it, and mail them your bug
495
report with the commit id. Finally, run
497
-------------------------------------------------
499
-------------------------------------------------
501
to return you to the branch you were on before and delete the
502
temporary "bisect" branch.
504
Note that the version which git-bisect checks out for you at each
505
point is just a suggestion, and you're free to try a different
506
version if you think it would be a good idea. For example,
507
occasionally you may land on a commit that broke something unrelated;
510
-------------------------------------------------
511
$ git bisect visualize
512
-------------------------------------------------
514
which will run gitk and label the commit it chose with a marker that
515
says "bisect". Chose a safe-looking commit nearby, note its commit
516
id, and check it out with:
518
-------------------------------------------------
519
$ git reset --hard fb47ddb2db...
520
-------------------------------------------------
522
then test, run "bisect good" or "bisect bad" as appropriate, and
529
We have seen several ways of naming commits already:
531
- 40-hexdigit object name
532
- branch name: refers to the commit at the head of the given
534
- tag name: refers to the commit pointed to by the given tag
535
(we've seen branches and tags are special cases of
536
<<how-git-stores-references,references>>).
537
- HEAD: refers to the head of the current branch
539
There are many more; see the "SPECIFYING REVISIONS" section of the
540
gitlink:git-rev-parse[1] man page for the complete list of ways to
541
name revisions. Some examples:
543
-------------------------------------------------
544
$ git show fb47ddb2 # the first few characters of the object name
545
# are usually enough to specify it uniquely
546
$ git show HEAD^ # the parent of the HEAD commit
547
$ git show HEAD^^ # the grandparent
548
$ git show HEAD~4 # the great-great-grandparent
549
-------------------------------------------------
551
Recall that merge commits may have more than one parent; by default,
552
^ and ~ follow the first parent listed in the commit, but you can
555
-------------------------------------------------
556
$ git show HEAD^1 # show the first parent of HEAD
557
$ git show HEAD^2 # show the second parent of HEAD
558
-------------------------------------------------
560
In addition to HEAD, there are several other special names for
563
Merges (to be discussed later), as well as operations such as
564
git-reset, which change the currently checked-out commit, generally
565
set ORIG_HEAD to the value HEAD had before the current operation.
567
The git-fetch operation always stores the head of the last fetched
568
branch in FETCH_HEAD. For example, if you run git fetch without
569
specifying a local branch as the target of the operation
571
-------------------------------------------------
572
$ git fetch git://example.com/proj.git theirbranch
573
-------------------------------------------------
575
the fetched commits will still be available from FETCH_HEAD.
577
When we discuss merges we'll also see the special name MERGE_HEAD,
578
which refers to the other branch that we're merging in to the current
581
The gitlink:git-rev-parse[1] command is a low-level command that is
582
occasionally useful for translating some name for a commit to the object
583
name for that commit:
585
-------------------------------------------------
586
$ git rev-parse origin
587
e05db0fd4f31dde7005f075a84f96b360d05984b
588
-------------------------------------------------
594
We can also create a tag to refer to a particular commit; after
597
-------------------------------------------------
598
$ git tag stable-1 1b2e1d63ff
599
-------------------------------------------------
601
You can use stable-1 to refer to the commit 1b2e1d63ff.
603
This creates a "lightweight" tag. If you would also like to include a
604
comment with the tag, and possibly sign it cryptographically, then you
605
should create a tag object instead; see the gitlink:git-tag[1] man page
608
[[browsing-revisions]]
612
The gitlink:git-log[1] command can show lists of commits. On its
613
own, it shows all commits reachable from the parent commit; but you
614
can also make more specific requests:
616
-------------------------------------------------
617
$ git log v2.5.. # commits since (not reachable from) v2.5
618
$ git log test..master # commits reachable from master but not test
619
$ git log master..test # ...reachable from test but not master
620
$ git log master...test # ...reachable from either test or master,
622
$ git log --since="2 weeks ago" # commits from the last 2 weeks
623
$ git log Makefile # commits which modify Makefile
624
$ git log fs/ # ... which modify any file under fs/
625
$ git log -S'foo()' # commits which add or remove any file data
626
# matching the string 'foo()'
627
-------------------------------------------------
629
And of course you can combine all of these; the following finds
630
commits since v2.5 which touch the Makefile or any file under fs:
632
-------------------------------------------------
633
$ git log v2.5.. Makefile fs/
634
-------------------------------------------------
636
You can also ask git log to show patches:
638
-------------------------------------------------
640
-------------------------------------------------
642
See the "--pretty" option in the gitlink:git-log[1] man page for more
645
Note that git log starts with the most recent commit and works
646
backwards through the parents; however, since git history can contain
647
multiple independent lines of development, the particular order that
648
commits are listed in may be somewhat arbitrary.
654
You can generate diffs between any two versions using
657
-------------------------------------------------
658
$ git diff master..test
659
-------------------------------------------------
661
Sometimes what you want instead is a set of patches:
663
-------------------------------------------------
664
$ git format-patch master..test
665
-------------------------------------------------
667
will generate a file with a patch for each commit reachable from test
668
but not from master. Note that if master also has commits which are
669
not reachable from test, then the combined result of these patches
670
will not be the same as the diff produced by the git-diff example.
672
[[viewing-old-file-versions]]
673
Viewing old file versions
674
-------------------------
676
You can always view an old version of a file by just checking out the
677
correct revision first. But sometimes it is more convenient to be
678
able to view an old version of a single file without checking
679
anything out; this command does that:
681
-------------------------------------------------
682
$ git show v2.5:fs/locks.c
683
-------------------------------------------------
685
Before the colon may be anything that names a commit, and after it
686
may be any path to a file tracked by git.
692
[[counting-commits-on-a-branch]]
693
Counting the number of commits on a branch
694
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
696
Suppose you want to know how many commits you've made on "mybranch"
697
since it diverged from "origin":
699
-------------------------------------------------
700
$ git log --pretty=oneline origin..mybranch | wc -l
701
-------------------------------------------------
703
Alternatively, you may often see this sort of thing done with the
704
lower-level command gitlink:git-rev-list[1], which just lists the SHA1's
705
of all the given commits:
707
-------------------------------------------------
708
$ git rev-list origin..mybranch | wc -l
709
-------------------------------------------------
711
[[checking-for-equal-branches]]
712
Check whether two branches point at the same history
713
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
715
Suppose you want to check whether two branches point at the same point
718
-------------------------------------------------
719
$ git diff origin..master
720
-------------------------------------------------
722
will tell you whether the contents of the project are the same at the
723
two branches; in theory, however, it's possible that the same project
724
contents could have been arrived at by two different historical
725
routes. You could compare the object names:
727
-------------------------------------------------
728
$ git rev-list origin
729
e05db0fd4f31dde7005f075a84f96b360d05984b
730
$ git rev-list master
731
e05db0fd4f31dde7005f075a84f96b360d05984b
732
-------------------------------------------------
734
Or you could recall that the ... operator selects all commits
735
contained reachable from either one reference or the other but not
738
-------------------------------------------------
739
$ git log origin...master
740
-------------------------------------------------
742
will return no commits when the two branches are equal.
744
[[finding-tagged-descendants]]
745
Find first tagged version including a given fix
746
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
748
Suppose you know that the commit e05db0fd fixed a certain problem.
749
You'd like to find the earliest tagged release that contains that
752
Of course, there may be more than one answer--if the history branched
753
after commit e05db0fd, then there could be multiple "earliest" tagged
756
You could just visually inspect the commits since e05db0fd:
758
-------------------------------------------------
760
-------------------------------------------------
762
Or you can use gitlink:git-name-rev[1], which will give the commit a
763
name based on any tag it finds pointing to one of the commit's
766
-------------------------------------------------
767
$ git name-rev --tags e05db0fd
768
e05db0fd tags/v1.5.0-rc1^0~23
769
-------------------------------------------------
771
The gitlink:git-describe[1] command does the opposite, naming the
772
revision using a tag on which the given commit is based:
774
-------------------------------------------------
775
$ git describe e05db0fd
776
v1.5.0-rc0-260-ge05db0f
777
-------------------------------------------------
779
but that may sometimes help you guess which tags might come after the
782
If you just want to verify whether a given tagged version contains a
783
given commit, you could use gitlink:git-merge-base[1]:
785
-------------------------------------------------
786
$ git merge-base e05db0fd v1.5.0-rc1
787
e05db0fd4f31dde7005f075a84f96b360d05984b
788
-------------------------------------------------
790
The merge-base command finds a common ancestor of the given commits,
791
and always returns one or the other in the case where one is a
792
descendant of the other; so the above output shows that e05db0fd
793
actually is an ancestor of v1.5.0-rc1.
795
Alternatively, note that
797
-------------------------------------------------
798
$ git log v1.5.0-rc1..e05db0fd
799
-------------------------------------------------
801
will produce empty output if and only if v1.5.0-rc1 includes e05db0fd,
802
because it outputs only commits that are not reachable from v1.5.0-rc1.
804
As yet another alternative, the gitlink:git-show-branch[1] command lists
805
the commits reachable from its arguments with a display on the left-hand
806
side that indicates which arguments that commit is reachable from. So,
807
you can run something like
809
-------------------------------------------------
810
$ git show-branch e05db0fd v1.5.0-rc0 v1.5.0-rc1 v1.5.0-rc2
811
! [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
813
! [v1.5.0-rc0] GIT v1.5.0 preview
814
! [v1.5.0-rc1] GIT v1.5.0-rc1
815
! [v1.5.0-rc2] GIT v1.5.0-rc2
817
-------------------------------------------------
819
then search for a line that looks like
821
-------------------------------------------------
822
+ ++ [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
824
-------------------------------------------------
826
Which shows that e05db0fd is reachable from itself, from v1.5.0-rc1, and
827
from v1.5.0-rc2, but not from v1.5.0-rc0.
829
[[showing-commits-unique-to-a-branch]]
830
Showing commits unique to a given branch
831
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
833
Suppose you would like to see all the commits reachable from the branch
834
head named "master" but not from any other head in your repository.
836
We can list all the heads in this repository with
837
gitlink:git-show-ref[1]:
839
-------------------------------------------------
840
$ git show-ref --heads
841
bf62196b5e363d73353a9dcf094c59595f3153b7 refs/heads/core-tutorial
842
db768d5504c1bb46f63ee9d6e1772bd047e05bf9 refs/heads/maint
843
a07157ac624b2524a059a3414e99f6f44bebc1e7 refs/heads/master
844
24dbc180ea14dc1aebe09f14c8ecf32010690627 refs/heads/tutorial-2
845
1e87486ae06626c2f31eaa63d26fc0fd646c8af2 refs/heads/tutorial-fixes
846
-------------------------------------------------
848
We can get just the branch-head names, and remove "master", with
849
the help of the standard utilities cut and grep:
851
-------------------------------------------------
852
$ git show-ref --heads | cut -d' ' -f2 | grep -v '^refs/heads/master'
853
refs/heads/core-tutorial
855
refs/heads/tutorial-2
856
refs/heads/tutorial-fixes
857
-------------------------------------------------
859
And then we can ask to see all the commits reachable from master
860
but not from these other heads:
862
-------------------------------------------------
863
$ gitk master --not $( git show-ref --heads | cut -d' ' -f2 |
864
grep -v '^refs/heads/master' )
865
-------------------------------------------------
867
Obviously, endless variations are possible; for example, to see all
868
commits reachable from some head but not from any tag in the repository:
870
-------------------------------------------------
871
$ gitk $( git show-ref --heads ) --not $( git show-ref --tags )
872
-------------------------------------------------
874
(See gitlink:git-rev-parse[1] for explanations of commit-selecting
875
syntax such as `--not`.)
878
Creating a changelog and tarball for a software release
879
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
881
The gitlink:git-archive[1] command can create a tar or zip archive from
882
any version of a project; for example:
884
-------------------------------------------------
885
$ git archive --format=tar --prefix=project/ HEAD | gzip >latest.tar.gz
886
-------------------------------------------------
888
will use HEAD to produce a tar archive in which each filename is
889
preceded by "project/".
891
If you're releasing a new version of a software project, you may want
892
to simultaneously make a changelog to include in the release
895
Linus Torvalds, for example, makes new kernel releases by tagging them,
898
-------------------------------------------------
899
$ release-script 2.6.12 2.6.13-rc6 2.6.13-rc7
900
-------------------------------------------------
902
where release-script is a shell script that looks like:
904
-------------------------------------------------
909
echo "# git tag v$new"
910
echo "git archive --prefix=linux-$new/ v$new | gzip -9 > ../linux-$new.tar.gz"
911
echo "git diff v$stable v$new | gzip -9 > ../patch-$new.gz"
912
echo "git log --no-merges v$new ^v$last > ../ChangeLog-$new"
913
echo "git shortlog --no-merges v$new ^v$last > ../ShortLog"
914
echo "git diff --stat --summary -M v$last v$new > ../diffstat-$new"
915
-------------------------------------------------
917
and then he just cut-and-pastes the output commands after verifying that
920
[[Finding-comments-with-given-content]]
921
Finding commits referencing a file with given content
922
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
924
Somebody hands you a copy of a file, and asks which commits modified a
925
file such that it contained the given content either before or after the
926
commit. You can find out with this:
928
-------------------------------------------------
929
$ git log --raw --abbrev=40 --pretty=oneline -- filename |
930
grep -B 1 `git hash-object filename`
931
-------------------------------------------------
933
Figuring out why this works is left as an exercise to the (advanced)
934
student. The gitlink:git-log[1], gitlink:git-diff-tree[1], and
935
gitlink:git-hash-object[1] man pages may prove helpful.
937
[[Developing-with-git]]
941
[[telling-git-your-name]]
942
Telling git your name
943
---------------------
945
Before creating any commits, you should introduce yourself to git. The
946
easiest way to do so is to make sure the following lines appear in a
947
file named .gitconfig in your home directory:
949
------------------------------------------------
951
name = Your Name Comes Here
952
email = you@yourdomain.example.com
953
------------------------------------------------
955
(See the "CONFIGURATION FILE" section of gitlink:git-config[1] for
956
details on the configuration file.)
959
[[creating-a-new-repository]]
960
Creating a new repository
961
-------------------------
963
Creating a new repository from scratch is very easy:
965
-------------------------------------------------
969
-------------------------------------------------
971
If you have some initial content (say, a tarball):
973
-------------------------------------------------
974
$ tar -xzvf project.tar.gz
977
$ git add . # include everything below ./ in the first commit:
979
-------------------------------------------------
981
[[how-to-make-a-commit]]
985
Creating a new commit takes three steps:
987
1. Making some changes to the working directory using your
989
2. Telling git about your changes.
990
3. Creating the commit using the content you told git about
993
In practice, you can interleave and repeat steps 1 and 2 as many
994
times as you want: in order to keep track of what you want committed
995
at step 3, git maintains a snapshot of the tree's contents in a
996
special staging area called "the index."
998
At the beginning, the content of the index will be identical to
999
that of the HEAD. The command "git diff --cached", which shows
1000
the difference between the HEAD and the index, should therefore
1001
produce no output at that point.
1003
Modifying the index is easy:
1005
To update the index with the new contents of a modified file, use
1007
-------------------------------------------------
1008
$ git add path/to/file
1009
-------------------------------------------------
1011
To add the contents of a new file to the index, use
1013
-------------------------------------------------
1014
$ git add path/to/file
1015
-------------------------------------------------
1017
To remove a file from the index and from the working tree,
1019
-------------------------------------------------
1020
$ git rm path/to/file
1021
-------------------------------------------------
1023
After each step you can verify that
1025
-------------------------------------------------
1027
-------------------------------------------------
1029
always shows the difference between the HEAD and the index file--this
1030
is what you'd commit if you created the commit now--and that
1032
-------------------------------------------------
1034
-------------------------------------------------
1036
shows the difference between the working tree and the index file.
1038
Note that "git add" always adds just the current contents of a file
1039
to the index; further changes to the same file will be ignored unless
1040
you run git-add on the file again.
1042
When you're ready, just run
1044
-------------------------------------------------
1046
-------------------------------------------------
1048
and git will prompt you for a commit message and then create the new
1049
commit. Check to make sure it looks like what you expected with
1051
-------------------------------------------------
1053
-------------------------------------------------
1055
As a special shortcut,
1057
-------------------------------------------------
1059
-------------------------------------------------
1061
will update the index with any files that you've modified or removed
1062
and create a commit, all in one step.
1064
A number of commands are useful for keeping track of what you're
1067
-------------------------------------------------
1068
$ git diff --cached # difference between HEAD and the index; what
1069
# would be committed if you ran "commit" now.
1070
$ git diff # difference between the index file and your
1071
# working directory; changes that would not
1072
# be included if you ran "commit" now.
1073
$ git diff HEAD # difference between HEAD and working tree; what
1074
# would be committed if you ran "commit -a" now.
1075
$ git status # a brief per-file summary of the above.
1076
-------------------------------------------------
1078
You can also use gitlink:git-gui[1] to create commits, view changes in
1079
the index and the working tree files, and individually select diff hunks
1080
for inclusion in the index (by right-clicking on the diff hunk and
1081
choosing "Stage Hunk For Commit").
1083
[[creating-good-commit-messages]]
1084
Creating good commit messages
1085
-----------------------------
1087
Though not required, it's a good idea to begin the commit message
1088
with a single short (less than 50 character) line summarizing the
1089
change, followed by a blank line and then a more thorough
1090
description. Tools that turn commits into email, for example, use
1091
the first line on the Subject line and the rest of the commit in the
1098
A project will often generate files that you do 'not' want to track with git.
1099
This typically includes files generated by a build process or temporary
1100
backup files made by your editor. Of course, 'not' tracking files with git
1101
is just a matter of 'not' calling "`git add`" on them. But it quickly becomes
1102
annoying to have these untracked files lying around; e.g. they make
1103
"`git add .`" and "`git commit -a`" practically useless, and they keep
1104
showing up in the output of "`git status`".
1106
You can tell git to ignore certain files by creating a file called .gitignore
1107
in the top level of your working directory, with contents such as:
1109
-------------------------------------------------
1110
# Lines starting with '#' are considered comments.
1111
# Ignore any file named foo.txt.
1113
# Ignore (generated) html files,
1115
# except foo.html which is maintained by hand.
1117
# Ignore objects and archives.
1119
-------------------------------------------------
1121
See gitlink:gitignore[5] for a detailed explanation of the syntax. You can
1122
also place .gitignore files in other directories in your working tree, and they
1123
will apply to those directories and their subdirectories. The `.gitignore`
1124
files can be added to your repository like any other files (just run `git add
1125
.gitignore` and `git commit`, as usual), which is convenient when the exclude
1126
patterns (such as patterns matching build output files) would also make sense
1127
for other users who clone your repository.
1129
If you wish the exclude patterns to affect only certain repositories
1130
(instead of every repository for a given project), you may instead put
1131
them in a file in your repository named .git/info/exclude, or in any file
1132
specified by the `core.excludesfile` configuration variable. Some git
1133
commands can also take exclude patterns directly on the command line.
1134
See gitlink:gitignore[5] for the details.
1140
You can rejoin two diverging branches of development using
1141
gitlink:git-merge[1]:
1143
-------------------------------------------------
1144
$ git merge branchname
1145
-------------------------------------------------
1147
merges the development in the branch "branchname" into the current
1148
branch. If there are conflicts--for example, if the same file is
1149
modified in two different ways in the remote branch and the local
1150
branch--then you are warned; the output may look something like this:
1152
-------------------------------------------------
1155
Auto-merged file.txt
1156
CONFLICT (content): Merge conflict in file.txt
1157
Automatic merge failed; fix conflicts and then commit the result.
1158
-------------------------------------------------
1160
Conflict markers are left in the problematic files, and after
1161
you resolve the conflicts manually, you can update the index
1162
with the contents and run git commit, as you normally would when
1163
creating a new file.
1165
If you examine the resulting commit using gitk, you will see that it
1166
has two parents, one pointing to the top of the current branch, and
1167
one to the top of the other branch.
1169
[[resolving-a-merge]]
1173
When a merge isn't resolved automatically, git leaves the index and
1174
the working tree in a special state that gives you all the
1175
information you need to help resolve the merge.
1177
Files with conflicts are marked specially in the index, so until you
1178
resolve the problem and update the index, gitlink:git-commit[1] will
1181
-------------------------------------------------
1183
file.txt: needs merge
1184
-------------------------------------------------
1186
Also, gitlink:git-status[1] will list those files as "unmerged", and the
1187
files with conflicts will have conflict markers added, like this:
1189
-------------------------------------------------
1190
<<<<<<< HEAD:file.txt
1194
>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1195
-------------------------------------------------
1197
All you need to do is edit the files to resolve the conflicts, and then
1199
-------------------------------------------------
1202
-------------------------------------------------
1204
Note that the commit message will already be filled in for you with
1205
some information about the merge. Normally you can just use this
1206
default message unchanged, but you may add additional commentary of
1207
your own if desired.
1209
The above is all you need to know to resolve a simple merge. But git
1210
also provides more information to help resolve conflicts:
1212
[[conflict-resolution]]
1213
Getting conflict-resolution help during a merge
1214
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1216
All of the changes that git was able to merge automatically are
1217
already added to the index file, so gitlink:git-diff[1] shows only
1218
the conflicts. It uses an unusual syntax:
1220
-------------------------------------------------
1223
index 802992c,2b60207..0000000
1226
@@@ -1,1 -1,1 +1,5 @@@
1227
++<<<<<<< HEAD:file.txt
1231
++>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1232
-------------------------------------------------
1234
Recall that the commit which will be committed after we resolve this
1235
conflict will have two parents instead of the usual one: one parent
1236
will be HEAD, the tip of the current branch; the other will be the
1237
tip of the other branch, which is stored temporarily in MERGE_HEAD.
1239
During the merge, the index holds three versions of each file. Each of
1240
these three "file stages" represents a different version of the file:
1242
-------------------------------------------------
1243
$ git show :1:file.txt # the file in a common ancestor of both branches
1244
$ git show :2:file.txt # the version from HEAD, but including any
1245
# nonconflicting changes from MERGE_HEAD
1246
$ git show :3:file.txt # the version from MERGE_HEAD, but including any
1247
# nonconflicting changes from HEAD.
1248
-------------------------------------------------
1250
Since the stage 2 and stage 3 versions have already been updated with
1251
nonconflicting changes, the only remaining differences between them are
1252
the important ones; thus gitlink:git-diff[1] can use the information in
1253
the index to show only those conflicts.
1255
The diff above shows the differences between the working-tree version of
1256
file.txt and the stage 2 and stage 3 versions. So instead of preceding
1257
each line by a single "+" or "-", it now uses two columns: the first
1258
column is used for differences between the first parent and the working
1259
directory copy, and the second for differences between the second parent
1260
and the working directory copy. (See the "COMBINED DIFF FORMAT" section
1261
of gitlink:git-diff-files[1] for a details of the format.)
1263
After resolving the conflict in the obvious way (but before updating the
1264
index), the diff will look like:
1266
-------------------------------------------------
1269
index 802992c,2b60207..0000000
1272
@@@ -1,1 -1,1 +1,1 @@@
1276
-------------------------------------------------
1278
This shows that our resolved version deleted "Hello world" from the
1279
first parent, deleted "Goodbye" from the second parent, and added
1280
"Goodbye world", which was previously absent from both.
1282
Some special diff options allow diffing the working directory against
1283
any of these stages:
1285
-------------------------------------------------
1286
$ git diff -1 file.txt # diff against stage 1
1287
$ git diff --base file.txt # same as the above
1288
$ git diff -2 file.txt # diff against stage 2
1289
$ git diff --ours file.txt # same as the above
1290
$ git diff -3 file.txt # diff against stage 3
1291
$ git diff --theirs file.txt # same as the above.
1292
-------------------------------------------------
1294
The gitlink:git-log[1] and gitk[1] commands also provide special help
1297
-------------------------------------------------
1300
-------------------------------------------------
1302
These will display all commits which exist only on HEAD or on
1303
MERGE_HEAD, and which touch an unmerged file.
1305
You may also use gitlink:git-mergetool[1], which lets you merge the
1306
unmerged files using external tools such as emacs or kdiff3.
1308
Each time you resolve the conflicts in a file and update the index:
1310
-------------------------------------------------
1312
-------------------------------------------------
1314
the different stages of that file will be "collapsed", after which
1315
git-diff will (by default) no longer show diffs for that file.
1321
If you get stuck and decide to just give up and throw the whole mess
1322
away, you can always return to the pre-merge state with
1324
-------------------------------------------------
1325
$ git reset --hard HEAD
1326
-------------------------------------------------
1328
Or, if you've already committed the merge that you want to throw away,
1330
-------------------------------------------------
1331
$ git reset --hard ORIG_HEAD
1332
-------------------------------------------------
1334
However, this last command can be dangerous in some cases--never
1335
throw away a commit you have already committed if that commit may
1336
itself have been merged into another branch, as doing so may confuse
1343
There is one special case not mentioned above, which is treated
1344
differently. Normally, a merge results in a merge commit, with two
1345
parents, one pointing at each of the two lines of development that
1348
However, if the current branch is a descendant of the other--so every
1349
commit present in the one is already contained in the other--then git
1350
just performs a "fast forward"; the head of the current branch is moved
1351
forward to point at the head of the merged-in branch, without any new
1352
commits being created.
1358
If you've messed up the working tree, but haven't yet committed your
1359
mistake, you can return the entire working tree to the last committed
1362
-------------------------------------------------
1363
$ git reset --hard HEAD
1364
-------------------------------------------------
1366
If you make a commit that you later wish you hadn't, there are two
1367
fundamentally different ways to fix the problem:
1369
1. You can create a new commit that undoes whatever was done
1370
by the previous commit. This is the correct thing if your
1371
mistake has already been made public.
1373
2. You can go back and modify the old commit. You should
1374
never do this if you have already made the history public;
1375
git does not normally expect the "history" of a project to
1376
change, and cannot correctly perform repeated merges from
1377
a branch that has had its history changed.
1379
[[reverting-a-commit]]
1380
Fixing a mistake with a new commit
1381
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1383
Creating a new commit that reverts an earlier change is very easy;
1384
just pass the gitlink:git-revert[1] command a reference to the bad
1385
commit; for example, to revert the most recent commit:
1387
-------------------------------------------------
1389
-------------------------------------------------
1391
This will create a new commit which undoes the change in HEAD. You
1392
will be given a chance to edit the commit message for the new commit.
1394
You can also revert an earlier change, for example, the next-to-last:
1396
-------------------------------------------------
1398
-------------------------------------------------
1400
In this case git will attempt to undo the old change while leaving
1401
intact any changes made since then. If more recent changes overlap
1402
with the changes to be reverted, then you will be asked to fix
1403
conflicts manually, just as in the case of <<resolving-a-merge,
1404
resolving a merge>>.
1406
[[fixing-a-mistake-by-editing-history]]
1407
Fixing a mistake by editing history
1408
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1410
If the problematic commit is the most recent commit, and you have not
1411
yet made that commit public, then you may just
1412
<<undoing-a-merge,destroy it using git-reset>>.
1415
can edit the working directory and update the index to fix your
1416
mistake, just as if you were going to <<how-to-make-a-commit,create a
1417
new commit>>, then run
1419
-------------------------------------------------
1420
$ git commit --amend
1421
-------------------------------------------------
1423
which will replace the old commit by a new commit incorporating your
1424
changes, giving you a chance to edit the old commit message first.
1426
Again, you should never do this to a commit that may already have
1427
been merged into another branch; use gitlink:git-revert[1] instead in
1430
It is also possible to edit commits further back in the history, but
1431
this is an advanced topic to be left for
1432
<<cleaning-up-history,another chapter>>.
1434
[[checkout-of-path]]
1435
Checking out an old version of a file
1436
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1438
In the process of undoing a previous bad change, you may find it
1439
useful to check out an older version of a particular file using
1440
gitlink:git-checkout[1]. We've used git checkout before to switch
1441
branches, but it has quite different behavior if it is given a path
1444
-------------------------------------------------
1445
$ git checkout HEAD^ path/to/file
1446
-------------------------------------------------
1448
replaces path/to/file by the contents it had in the commit HEAD^, and
1449
also updates the index to match. It does not change branches.
1451
If you just want to look at an old version of the file, without
1452
modifying the working directory, you can do that with
1453
gitlink:git-show[1]:
1455
-------------------------------------------------
1456
$ git show HEAD^:path/to/file
1457
-------------------------------------------------
1459
which will display the given version of the file.
1461
[[interrupted-work]]
1462
Temporarily setting aside work in progress
1463
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1465
While you are in the middle of working on something complicated, you
1466
find an unrelated but obvious and trivial bug. You would like to fix it
1467
before continuing. You can use gitlink:git-stash[1] to save the current
1468
state of your work, and after fixing the bug (or, optionally after doing
1469
so on a different branch and then coming back), unstash the
1470
work-in-progress changes.
1472
------------------------------------------------
1473
$ git stash "work in progress for foo feature"
1474
------------------------------------------------
1476
This command will save your changes away to the `stash`, and
1477
reset your working tree and the index to match the tip of your
1478
current branch. Then you can make your fix as usual.
1480
------------------------------------------------
1481
... edit and test ...
1482
$ git commit -a -m "blorpl: typofix"
1483
------------------------------------------------
1485
After that, you can go back to what you were working on with
1488
------------------------------------------------
1490
------------------------------------------------
1493
[[ensuring-good-performance]]
1494
Ensuring good performance
1495
-------------------------
1497
On large repositories, git depends on compression to keep the history
1498
information from taking up to much space on disk or in memory.
1500
This compression is not performed automatically. Therefore you
1501
should occasionally run gitlink:git-gc[1]:
1503
-------------------------------------------------
1505
-------------------------------------------------
1507
to recompress the archive. This can be very time-consuming, so
1508
you may prefer to run git-gc when you are not doing other work.
1511
[[ensuring-reliability]]
1512
Ensuring reliability
1513
--------------------
1515
[[checking-for-corruption]]
1516
Checking the repository for corruption
1517
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1519
The gitlink:git-fsck[1] command runs a number of self-consistency checks
1520
on the repository, and reports on any problems. This may take some
1521
time. The most common warning by far is about "dangling" objects:
1523
-------------------------------------------------
1525
dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1526
dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1527
dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1528
dangling blob 218761f9d90712d37a9c5e36f406f92202db07eb
1529
dangling commit bf093535a34a4d35731aa2bd90fe6b176302f14f
1530
dangling commit 8e4bec7f2ddaa268bef999853c25755452100f8e
1531
dangling tree d50bb86186bf27b681d25af89d3b5b68382e4085
1532
dangling tree b24c2473f1fd3d91352a624795be026d64c8841f
1534
-------------------------------------------------
1536
Dangling objects are not a problem. At worst they may take up a little
1537
extra disk space. They can sometimes provide a last-resort method for
1538
recovering lost work--see <<dangling-objects>> for details. However, if
1539
you wish, you can remove them with gitlink:git-prune[1] or the --prune
1540
option to gitlink:git-gc[1]:
1542
-------------------------------------------------
1544
-------------------------------------------------
1546
This may be time-consuming. Unlike most other git operations (including
1547
git-gc when run without any options), it is not safe to prune while
1548
other git operations are in progress in the same repository.
1550
[[recovering-lost-changes]]
1551
Recovering lost changes
1552
~~~~~~~~~~~~~~~~~~~~~~~
1558
Say you modify a branch with gitlink:git-reset[1] --hard, and then
1559
realize that the branch was the only reference you had to that point in
1562
Fortunately, git also keeps a log, called a "reflog", of all the
1563
previous values of each branch. So in this case you can still find the
1564
old history using, for example,
1566
-------------------------------------------------
1567
$ git log master@{1}
1568
-------------------------------------------------
1570
This lists the commits reachable from the previous version of the head.
1571
This syntax can be used to with any git command that accepts a commit,
1572
not just with git log. Some other examples:
1574
-------------------------------------------------
1575
$ git show master@{2} # See where the branch pointed 2,
1576
$ git show master@{3} # 3, ... changes ago.
1577
$ gitk master@{yesterday} # See where it pointed yesterday,
1578
$ gitk master@{"1 week ago"} # ... or last week
1579
$ git log --walk-reflogs master # show reflog entries for master
1580
-------------------------------------------------
1582
A separate reflog is kept for the HEAD, so
1584
-------------------------------------------------
1585
$ git show HEAD@{"1 week ago"}
1586
-------------------------------------------------
1588
will show what HEAD pointed to one week ago, not what the current branch
1589
pointed to one week ago. This allows you to see the history of what
1592
The reflogs are kept by default for 30 days, after which they may be
1593
pruned. See gitlink:git-reflog[1] and gitlink:git-gc[1] to learn
1594
how to control this pruning, and see the "SPECIFYING REVISIONS"
1595
section of gitlink:git-rev-parse[1] for details.
1597
Note that the reflog history is very different from normal git history.
1598
While normal history is shared by every repository that works on the
1599
same project, the reflog history is not shared: it tells you only about
1600
how the branches in your local repository have changed over time.
1602
[[dangling-object-recovery]]
1603
Examining dangling objects
1604
^^^^^^^^^^^^^^^^^^^^^^^^^^
1606
In some situations the reflog may not be able to save you. For example,
1607
suppose you delete a branch, then realize you need the history it
1608
contained. The reflog is also deleted; however, if you have not yet
1609
pruned the repository, then you may still be able to find the lost
1610
commits in the dangling objects that git-fsck reports. See
1611
<<dangling-objects>> for the details.
1613
-------------------------------------------------
1615
dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1616
dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1617
dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1619
-------------------------------------------------
1622
one of those dangling commits with, for example,
1624
------------------------------------------------
1625
$ gitk 7281251ddd --not --all
1626
------------------------------------------------
1628
which does what it sounds like: it says that you want to see the commit
1629
history that is described by the dangling commit(s), but not the
1630
history that is described by all your existing branches and tags. Thus
1631
you get exactly the history reachable from that commit that is lost.
1632
(And notice that it might not be just one commit: we only report the
1633
"tip of the line" as being dangling, but there might be a whole deep
1634
and complex commit history that was dropped.)
1636
If you decide you want the history back, you can always create a new
1637
reference pointing to it, for example, a new branch:
1639
------------------------------------------------
1640
$ git branch recovered-branch 7281251ddd
1641
------------------------------------------------
1643
Other types of dangling objects (blobs and trees) are also possible, and
1644
dangling objects can arise in other situations.
1647
[[sharing-development]]
1648
Sharing development with others
1649
===============================
1651
[[getting-updates-with-git-pull]]
1652
Getting updates with git pull
1653
-----------------------------
1655
After you clone a repository and make a few changes of your own, you
1656
may wish to check the original repository for updates and merge them
1659
We have already seen <<Updating-a-repository-with-git-fetch,how to
1660
keep remote tracking branches up to date>> with gitlink:git-fetch[1],
1661
and how to merge two branches. So you can merge in changes from the
1662
original repository's master branch with:
1664
-------------------------------------------------
1666
$ git merge origin/master
1667
-------------------------------------------------
1669
However, the gitlink:git-pull[1] command provides a way to do this in
1672
-------------------------------------------------
1673
$ git pull origin master
1674
-------------------------------------------------
1676
In fact, if you have "master" checked out, then by default "git pull"
1677
merges from the HEAD branch of the origin repository. So often you can
1678
accomplish the above with just a simple
1680
-------------------------------------------------
1682
-------------------------------------------------
1684
More generally, a branch that is created from a remote branch will pull
1685
by default from that branch. See the descriptions of the
1686
branch.<name>.remote and branch.<name>.merge options in
1687
gitlink:git-config[1], and the discussion of the --track option in
1688
gitlink:git-checkout[1], to learn how to control these defaults.
1690
In addition to saving you keystrokes, "git pull" also helps you by
1691
producing a default commit message documenting the branch and
1692
repository that you pulled from.
1694
(But note that no such commit will be created in the case of a
1695
<<fast-forwards,fast forward>>; instead, your branch will just be
1696
updated to point to the latest commit from the upstream branch.)
1698
The git-pull command can also be given "." as the "remote" repository,
1699
in which case it just merges in a branch from the current repository; so
1702
-------------------------------------------------
1705
-------------------------------------------------
1707
are roughly equivalent. The former is actually very commonly used.
1709
[[submitting-patches]]
1710
Submitting patches to a project
1711
-------------------------------
1713
If you just have a few changes, the simplest way to submit them may
1714
just be to send them as patches in email:
1716
First, use gitlink:git-format-patch[1]; for example:
1718
-------------------------------------------------
1719
$ git format-patch origin
1720
-------------------------------------------------
1722
will produce a numbered series of files in the current directory, one
1723
for each patch in the current branch but not in origin/HEAD.
1725
You can then import these into your mail client and send them by
1726
hand. However, if you have a lot to send at once, you may prefer to
1727
use the gitlink:git-send-email[1] script to automate the process.
1728
Consult the mailing list for your project first to determine how they
1729
prefer such patches be handled.
1731
[[importing-patches]]
1732
Importing patches to a project
1733
------------------------------
1735
Git also provides a tool called gitlink:git-am[1] (am stands for
1736
"apply mailbox"), for importing such an emailed series of patches.
1737
Just save all of the patch-containing messages, in order, into a
1738
single mailbox file, say "patches.mbox", then run
1740
-------------------------------------------------
1741
$ git am -3 patches.mbox
1742
-------------------------------------------------
1744
Git will apply each patch in order; if any conflicts are found, it
1745
will stop, and you can fix the conflicts as described in
1746
"<<resolving-a-merge,Resolving a merge>>". (The "-3" option tells
1747
git to perform a merge; if you would prefer it just to abort and
1748
leave your tree and index untouched, you may omit that option.)
1750
Once the index is updated with the results of the conflict
1751
resolution, instead of creating a new commit, just run
1753
-------------------------------------------------
1755
-------------------------------------------------
1757
and git will create the commit for you and continue applying the
1758
remaining patches from the mailbox.
1760
The final result will be a series of commits, one for each patch in
1761
the original mailbox, with authorship and commit log message each
1762
taken from the message containing each patch.
1764
[[public-repositories]]
1765
Public git repositories
1766
-----------------------
1768
Another way to submit changes to a project is to tell the maintainer
1769
of that project to pull the changes from your repository using
1770
gitlink:git-pull[1]. In the section "<<getting-updates-with-git-pull,
1771
Getting updates with git pull>>" we described this as a way to get
1772
updates from the "main" repository, but it works just as well in the
1775
If you and the maintainer both have accounts on the same machine, then
1776
you can just pull changes from each other's repositories directly;
1777
commands that accept repository URLs as arguments will also accept a
1778
local directory name:
1780
-------------------------------------------------
1781
$ git clone /path/to/repository
1782
$ git pull /path/to/other/repository
1783
-------------------------------------------------
1787
-------------------------------------------------
1788
$ git clone ssh://yourhost/~you/repository
1789
-------------------------------------------------
1791
For projects with few developers, or for synchronizing a few private
1792
repositories, this may be all you need.
1794
However, the more common way to do this is to maintain a separate public
1795
repository (usually on a different host) for others to pull changes
1796
from. This is usually more convenient, and allows you to cleanly
1797
separate private work in progress from publicly visible work.
1799
You will continue to do your day-to-day work in your personal
1800
repository, but periodically "push" changes from your personal
1801
repository into your public repository, allowing other developers to
1802
pull from that repository. So the flow of changes, in a situation
1803
where there is one other developer with a public repository, looks
1807
your personal repo ------------------> your public repo
1810
| you pull | they pull
1814
their public repo <------------------- their repo
1816
We explain how to do this in the following sections.
1818
[[setting-up-a-public-repository]]
1819
Setting up a public repository
1820
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1822
Assume your personal repository is in the directory ~/proj. We
1823
first create a new clone of the repository and tell git-daemon that it
1824
is meant to be public:
1826
-------------------------------------------------
1827
$ git clone --bare ~/proj proj.git
1828
$ touch proj.git/git-daemon-export-ok
1829
-------------------------------------------------
1831
The resulting directory proj.git contains a "bare" git repository--it is
1832
just the contents of the ".git" directory, without any files checked out
1835
Next, copy proj.git to the server where you plan to host the
1836
public repository. You can use scp, rsync, or whatever is most
1839
[[exporting-via-git]]
1840
Exporting a git repository via the git protocol
1841
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1843
This is the preferred method.
1845
If someone else administers the server, they should tell you what
1846
directory to put the repository in, and what git:// url it will appear
1847
at. You can then skip to the section
1848
"<<pushing-changes-to-a-public-repository,Pushing changes to a public
1849
repository>>", below.
1851
Otherwise, all you need to do is start gitlink:git-daemon[1]; it will
1852
listen on port 9418. By default, it will allow access to any directory
1853
that looks like a git directory and contains the magic file
1854
git-daemon-export-ok. Passing some directory paths as git-daemon
1855
arguments will further restrict the exports to those paths.
1857
You can also run git-daemon as an inetd service; see the
1858
gitlink:git-daemon[1] man page for details. (See especially the
1861
[[exporting-via-http]]
1862
Exporting a git repository via http
1863
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1865
The git protocol gives better performance and reliability, but on a
1866
host with a web server set up, http exports may be simpler to set up.
1868
All you need to do is place the newly created bare git repository in
1869
a directory that is exported by the web server, and make some
1870
adjustments to give web clients some extra information they need:
1872
-------------------------------------------------
1873
$ mv proj.git /home/you/public_html/proj.git
1875
$ git --bare update-server-info
1876
$ chmod a+x hooks/post-update
1877
-------------------------------------------------
1879
(For an explanation of the last two lines, see
1880
gitlink:git-update-server-info[1], and the documentation
1881
link:hooks.html[Hooks used by git].)
1883
Advertise the url of proj.git. Anybody else should then be able to
1884
clone or pull from that url, for example with a command line like:
1886
-------------------------------------------------
1887
$ git clone http://yourserver.com/~you/proj.git
1888
-------------------------------------------------
1891
link:howto/setup-git-server-over-http.txt[setup-git-server-over-http]
1892
for a slightly more sophisticated setup using WebDAV which also
1893
allows pushing over http.)
1895
[[pushing-changes-to-a-public-repository]]
1896
Pushing changes to a public repository
1897
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1899
Note that the two techniques outlined above (exporting via
1900
<<exporting-via-http,http>> or <<exporting-via-git,git>>) allow other
1901
maintainers to fetch your latest changes, but they do not allow write
1902
access, which you will need to update the public repository with the
1903
latest changes created in your private repository.
1905
The simplest way to do this is using gitlink:git-push[1] and ssh; to
1906
update the remote branch named "master" with the latest state of your
1907
branch named "master", run
1909
-------------------------------------------------
1910
$ git push ssh://yourserver.com/~you/proj.git master:master
1911
-------------------------------------------------
1915
-------------------------------------------------
1916
$ git push ssh://yourserver.com/~you/proj.git master
1917
-------------------------------------------------
1919
As with git-fetch, git-push will complain if this does not result in
1920
a <<fast-forwards,fast forward>>. Normally this is a sign of
1921
something wrong. However, if you are sure you know what you're
1922
doing, you may force git-push to perform the update anyway by
1923
proceeding the branch name by a plus sign:
1925
-------------------------------------------------
1926
$ git push ssh://yourserver.com/~you/proj.git +master
1927
-------------------------------------------------
1929
Note that the target of a "push" is normally a
1930
<<def_bare_repository,bare>> repository. You can also push to a
1931
repository that has a checked-out working tree, but the working tree
1932
will not be updated by the push. This may lead to unexpected results if
1933
the branch you push to is the currently checked-out branch!
1935
As with git-fetch, you may also set up configuration options to
1936
save typing; so, for example, after
1938
-------------------------------------------------
1939
$ cat >>.git/config <<EOF
1940
[remote "public-repo"]
1941
url = ssh://yourserver.com/~you/proj.git
1943
-------------------------------------------------
1945
you should be able to perform the above push with just
1947
-------------------------------------------------
1948
$ git push public-repo master
1949
-------------------------------------------------
1951
See the explanations of the remote.<name>.url, branch.<name>.remote,
1952
and remote.<name>.push options in gitlink:git-config[1] for
1955
[[setting-up-a-shared-repository]]
1956
Setting up a shared repository
1957
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1959
Another way to collaborate is by using a model similar to that
1960
commonly used in CVS, where several developers with special rights
1961
all push to and pull from a single shared repository. See
1962
link:cvs-migration.html[git for CVS users] for instructions on how to
1965
However, while there is nothing wrong with git's support for shared
1966
repositories, this mode of operation is not generally recommended,
1967
simply because the mode of collaboration that git supports--by
1968
exchanging patches and pulling from public repositories--has so many
1969
advantages over the central shared repository:
1971
- Git's ability to quickly import and merge patches allows a
1972
single maintainer to process incoming changes even at very
1973
high rates. And when that becomes too much, git-pull provides
1974
an easy way for that maintainer to delegate this job to other
1975
maintainers while still allowing optional review of incoming
1977
- Since every developer's repository has the same complete copy
1978
of the project history, no repository is special, and it is
1979
trivial for another developer to take over maintenance of a
1980
project, either by mutual agreement, or because a maintainer
1981
becomes unresponsive or difficult to work with.
1982
- The lack of a central group of "committers" means there is
1983
less need for formal decisions about who is "in" and who is
1986
[[setting-up-gitweb]]
1987
Allowing web browsing of a repository
1988
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1990
The gitweb cgi script provides users an easy way to browse your
1991
project's files and history without having to install git; see the file
1992
gitweb/INSTALL in the git source tree for instructions on setting it up.
1994
[[sharing-development-examples]]
1998
[[maintaining-topic-branches]]
1999
Maintaining topic branches for a Linux subsystem maintainer
2000
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2002
This describes how Tony Luck uses git in his role as maintainer of the
2003
IA64 architecture for the Linux kernel.
2005
He uses two public branches:
2007
- A "test" tree into which patches are initially placed so that they
2008
can get some exposure when integrated with other ongoing development.
2009
This tree is available to Andrew for pulling into -mm whenever he
2012
- A "release" tree into which tested patches are moved for final sanity
2013
checking, and as a vehicle to send them upstream to Linus (by sending
2014
him a "please pull" request.)
2016
He also uses a set of temporary branches ("topic branches"), each
2017
containing a logical grouping of patches.
2019
To set this up, first create your work tree by cloning Linus's public
2022
-------------------------------------------------
2023
$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git work
2025
-------------------------------------------------
2027
Linus's tree will be stored in the remote branch named origin/master,
2028
and can be updated using gitlink:git-fetch[1]; you can track other
2029
public trees using gitlink:git-remote[1] to set up a "remote" and
2030
gitlink:git-fetch[1] to keep them up-to-date; see
2031
<<repositories-and-branches>>.
2033
Now create the branches in which you are going to work; these start out
2034
at the current tip of origin/master branch, and should be set up (using
2035
the --track option to gitlink:git-branch[1]) to merge changes in from
2038
-------------------------------------------------
2039
$ git branch --track test origin/master
2040
$ git branch --track release origin/master
2041
-------------------------------------------------
2043
These can be easily kept up to date using gitlink:git-pull[1]
2045
-------------------------------------------------
2046
$ git checkout test && git pull
2047
$ git checkout release && git pull
2048
-------------------------------------------------
2050
Important note! If you have any local changes in these branches, then
2051
this merge will create a commit object in the history (with no local
2052
changes git will simply do a "Fast forward" merge). Many people dislike
2053
the "noise" that this creates in the Linux history, so you should avoid
2054
doing this capriciously in the "release" branch, as these noisy commits
2055
will become part of the permanent history when you ask Linus to pull
2056
from the release branch.
2058
A few configuration variables (see gitlink:git-config[1]) can
2059
make it easy to push both branches to your public tree. (See
2060
<<setting-up-a-public-repository>>.)
2062
-------------------------------------------------
2063
$ cat >> .git/config <<EOF
2065
url = master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6.git
2069
-------------------------------------------------
2071
Then you can push both the test and release trees using
2072
gitlink:git-push[1]:
2074
-------------------------------------------------
2076
-------------------------------------------------
2078
or push just one of the test and release branches using:
2080
-------------------------------------------------
2081
$ git push mytree test
2082
-------------------------------------------------
2086
-------------------------------------------------
2087
$ git push mytree release
2088
-------------------------------------------------
2090
Now to apply some patches from the community. Think of a short
2091
snappy name for a branch to hold this patch (or related group of
2092
patches), and create a new branch from the current tip of Linus's
2095
-------------------------------------------------
2096
$ git checkout -b speed-up-spinlocks origin
2097
-------------------------------------------------
2099
Now you apply the patch(es), run some tests, and commit the change(s). If
2100
the patch is a multi-part series, then you should apply each as a separate
2101
commit to this branch.
2103
-------------------------------------------------
2104
$ ... patch ... test ... commit [ ... patch ... test ... commit ]*
2105
-------------------------------------------------
2107
When you are happy with the state of this change, you can pull it into the
2108
"test" branch in preparation to make it public:
2110
-------------------------------------------------
2111
$ git checkout test && git pull . speed-up-spinlocks
2112
-------------------------------------------------
2114
It is unlikely that you would have any conflicts here ... but you might if you
2115
spent a while on this step and had also pulled new versions from upstream.
2117
Some time later when enough time has passed and testing done, you can pull the
2118
same branch into the "release" tree ready to go upstream. This is where you
2119
see the value of keeping each patch (or patch series) in its own branch. It
2120
means that the patches can be moved into the "release" tree in any order.
2122
-------------------------------------------------
2123
$ git checkout release && git pull . speed-up-spinlocks
2124
-------------------------------------------------
2126
After a while, you will have a number of branches, and despite the
2127
well chosen names you picked for each of them, you may forget what
2128
they are for, or what status they are in. To get a reminder of what
2129
changes are in a specific branch, use:
2131
-------------------------------------------------
2132
$ git log linux..branchname | git-shortlog
2133
-------------------------------------------------
2135
To see whether it has already been merged into the test or release branches
2138
-------------------------------------------------
2139
$ git log test..branchname
2140
-------------------------------------------------
2144
-------------------------------------------------
2145
$ git log release..branchname
2146
-------------------------------------------------
2148
(If this branch has not yet been merged you will see some log entries.
2149
If it has been merged, then there will be no output.)
2151
Once a patch completes the great cycle (moving from test to release,
2152
then pulled by Linus, and finally coming back into your local
2153
"origin/master" branch) the branch for this change is no longer needed.
2154
You detect this when the output from:
2156
-------------------------------------------------
2157
$ git log origin..branchname
2158
-------------------------------------------------
2160
is empty. At this point the branch can be deleted:
2162
-------------------------------------------------
2163
$ git branch -d branchname
2164
-------------------------------------------------
2166
Some changes are so trivial that it is not necessary to create a separate
2167
branch and then merge into each of the test and release branches. For
2168
these changes, just apply directly to the "release" branch, and then
2169
merge that into the "test" branch.
2171
To create diffstat and shortlog summaries of changes to include in a "please
2172
pull" request to Linus you can use:
2174
-------------------------------------------------
2175
$ git diff --stat origin..release
2176
-------------------------------------------------
2180
-------------------------------------------------
2181
$ git log -p origin..release | git shortlog
2182
-------------------------------------------------
2184
Here are some of the scripts that simplify all this even further.
2186
-------------------------------------------------
2187
==== update script ====
2188
# Update a branch in my GIT tree. If the branch to be updated
2189
# is origin, then pull from kernel.org. Otherwise merge
2190
# origin/master branch into test|release branch
2194
git checkout $1 && git pull . origin
2197
before=$(git rev-parse refs/remotes/origin/master)
2199
after=$(git rev-parse refs/remotes/origin/master)
2200
if [ $before != $after ]
2202
git log $before..$after | git shortlog
2206
echo "Usage: $0 origin|test|release" 1>&2
2210
-------------------------------------------------
2212
-------------------------------------------------
2213
==== merge script ====
2214
# Merge a branch into either the test or release branch
2220
echo "Usage: $pname branch test|release" 1>&2
2224
git show-ref -q --verify -- refs/heads/"$1" || {
2225
echo "Can't see branch <$1>" 1>&2
2231
if [ $(git log $2..$1 | wc -c) -eq 0 ]
2233
echo $1 already merged into $2 1>&2
2236
git checkout $2 && git pull . $1
2242
-------------------------------------------------
2244
-------------------------------------------------
2245
==== status script ====
2246
# report on status of my ia64 GIT tree
2250
restore=$(tput setab 9)
2252
if [ `git rev-list test..release | wc -c` -gt 0 ]
2254
echo $rb Warning: commits in release that are not in test $restore
2255
git log test..release
2258
for branch in `git show-ref --heads | sed 's|^.*/||'`
2260
if [ $branch = test -o $branch = release ]
2265
echo -n $gb ======= $branch ====== $restore " "
2267
for ref in test release origin/master
2269
if [ `git rev-list $ref..$branch | wc -c` -gt 0 ]
2271
status=$status${ref:0:1}
2276
echo $rb Need to pull into test $restore
2282
echo "Waiting for linus"
2285
echo $rb All done $restore
2288
echo $rb "<$status>" $restore
2291
git log origin/master..$branch | git shortlog
2293
-------------------------------------------------
2296
[[cleaning-up-history]]
2297
Rewriting history and maintaining patch series
2298
==============================================
2300
Normally commits are only added to a project, never taken away or
2301
replaced. Git is designed with this assumption, and violating it will
2302
cause git's merge machinery (for example) to do the wrong thing.
2304
However, there is a situation in which it can be useful to violate this
2308
Creating the perfect patch series
2309
---------------------------------
2311
Suppose you are a contributor to a large project, and you want to add a
2312
complicated feature, and to present it to the other developers in a way
2313
that makes it easy for them to read your changes, verify that they are
2314
correct, and understand why you made each change.
2316
If you present all of your changes as a single patch (or commit), they
2317
may find that it is too much to digest all at once.
2319
If you present them with the entire history of your work, complete with
2320
mistakes, corrections, and dead ends, they may be overwhelmed.
2322
So the ideal is usually to produce a series of patches such that:
2324
1. Each patch can be applied in order.
2326
2. Each patch includes a single logical change, together with a
2327
message explaining the change.
2329
3. No patch introduces a regression: after applying any initial
2330
part of the series, the resulting project still compiles and
2331
works, and has no bugs that it didn't have before.
2333
4. The complete series produces the same end result as your own
2334
(probably much messier!) development process did.
2336
We will introduce some tools that can help you do this, explain how to
2337
use them, and then explain some of the problems that can arise because
2338
you are rewriting history.
2340
[[using-git-rebase]]
2341
Keeping a patch series up to date using git-rebase
2342
--------------------------------------------------
2344
Suppose that you create a branch "mywork" on a remote-tracking branch
2345
"origin", and create some commits on top of it:
2347
-------------------------------------------------
2348
$ git checkout -b mywork origin
2354
-------------------------------------------------
2356
You have performed no merges into mywork, so it is just a simple linear
2357
sequence of patches on top of "origin":
2359
................................................
2363
................................................
2365
Some more interesting work has been done in the upstream project, and
2366
"origin" has advanced:
2368
................................................
2369
o--o--O--o--o--o <-- origin
2372
................................................
2374
At this point, you could use "pull" to merge your changes back in;
2375
the result would create a new merge commit, like this:
2377
................................................
2378
o--o--O--o--o--o <-- origin
2380
a--b--c--m <-- mywork
2381
................................................
2383
However, if you prefer to keep the history in mywork a simple series of
2384
commits without any merges, you may instead choose to use
2385
gitlink:git-rebase[1]:
2387
-------------------------------------------------
2388
$ git checkout mywork
2390
-------------------------------------------------
2392
This will remove each of your commits from mywork, temporarily saving
2393
them as patches (in a directory named ".dotest"), update mywork to
2394
point at the latest version of origin, then apply each of the saved
2395
patches to the new mywork. The result will look like:
2398
................................................
2399
o--o--O--o--o--o <-- origin
2401
a'--b'--c' <-- mywork
2402
................................................
2404
In the process, it may discover conflicts. In that case it will stop
2405
and allow you to fix the conflicts; after fixing conflicts, use "git
2406
add" to update the index with those contents, and then, instead of
2407
running git-commit, just run
2409
-------------------------------------------------
2410
$ git rebase --continue
2411
-------------------------------------------------
2413
and git will continue applying the rest of the patches.
2415
At any point you may use the --abort option to abort this process and
2416
return mywork to the state it had before you started the rebase:
2418
-------------------------------------------------
2419
$ git rebase --abort
2420
-------------------------------------------------
2422
[[modifying-one-commit]]
2423
Modifying a single commit
2424
-------------------------
2426
We saw in <<fixing-a-mistake-by-editing-history>> that you can replace the
2427
most recent commit using
2429
-------------------------------------------------
2430
$ git commit --amend
2431
-------------------------------------------------
2433
which will replace the old commit by a new commit incorporating your
2434
changes, giving you a chance to edit the old commit message first.
2436
You can also use a combination of this and gitlink:git-rebase[1] to edit
2437
commits further back in your history. First, tag the problematic commit with
2439
-------------------------------------------------
2440
$ git tag bad mywork~5
2441
-------------------------------------------------
2443
(Either gitk or git-log may be useful for finding the commit.)
2445
Then check out that commit, edit it, and rebase the rest of the series
2446
on top of it (note that we could check out the commit on a temporary
2447
branch, but instead we're using a <<detached-head,detached head>>):
2449
-------------------------------------------------
2451
$ # make changes here and update the index
2452
$ git commit --amend
2453
$ git rebase --onto HEAD bad mywork
2454
-------------------------------------------------
2456
When you're done, you'll be left with mywork checked out, with the top
2457
patches on mywork reapplied on top of your modified commit. You can
2460
-------------------------------------------------
2462
-------------------------------------------------
2464
Note that the immutable nature of git history means that you haven't really
2465
"modified" existing commits; instead, you have replaced the old commits with
2466
new commits having new object names.
2468
[[reordering-patch-series]]
2469
Reordering or selecting from a patch series
2470
-------------------------------------------
2472
Given one existing commit, the gitlink:git-cherry-pick[1] command
2473
allows you to apply the change introduced by that commit and create a
2474
new commit that records it. So, for example, if "mywork" points to a
2475
series of patches on top of "origin", you might do something like:
2477
-------------------------------------------------
2478
$ git checkout -b mywork-new origin
2479
$ gitk origin..mywork &
2480
-------------------------------------------------
2482
And browse through the list of patches in the mywork branch using gitk,
2483
applying them (possibly in a different order) to mywork-new using
2484
cherry-pick, and possibly modifying them as you go using commit --amend.
2485
The gitlink:git-gui[1] command may also help as it allows you to
2486
individually select diff hunks for inclusion in the index (by
2487
right-clicking on the diff hunk and choosing "Stage Hunk for Commit").
2489
Another technique is to use git-format-patch to create a series of
2490
patches, then reset the state to before the patches:
2492
-------------------------------------------------
2493
$ git format-patch origin
2494
$ git reset --hard origin
2495
-------------------------------------------------
2497
Then modify, reorder, or eliminate patches as preferred before applying
2498
them again with gitlink:git-am[1].
2500
[[patch-series-tools]]
2504
There are numerous other tools, such as StGIT, which exist for the
2505
purpose of maintaining a patch series. These are outside of the scope of
2508
[[problems-with-rewriting-history]]
2509
Problems with rewriting history
2510
-------------------------------
2512
The primary problem with rewriting the history of a branch has to do
2513
with merging. Suppose somebody fetches your branch and merges it into
2514
their branch, with a result something like this:
2516
................................................
2517
o--o--O--o--o--o <-- origin
2519
t--t--t--m <-- their branch:
2520
................................................
2522
Then suppose you modify the last three commits:
2524
................................................
2525
o--o--o <-- new head of origin
2527
o--o--O--o--o--o <-- old head of origin
2528
................................................
2530
If we examined all this history together in one repository, it will
2533
................................................
2534
o--o--o <-- new head of origin
2536
o--o--O--o--o--o <-- old head of origin
2538
t--t--t--m <-- their branch:
2539
................................................
2541
Git has no way of knowing that the new head is an updated version of
2542
the old head; it treats this situation exactly the same as it would if
2543
two developers had independently done the work on the old and new heads
2544
in parallel. At this point, if someone attempts to merge the new head
2545
in to their branch, git will attempt to merge together the two (old and
2546
new) lines of development, instead of trying to replace the old by the
2547
new. The results are likely to be unexpected.
2549
You may still choose to publish branches whose history is rewritten,
2550
and it may be useful for others to be able to fetch those branches in
2551
order to examine or test them, but they should not attempt to pull such
2552
branches into their own work.
2554
For true distributed development that supports proper merging,
2555
published branches should never be rewritten.
2557
[[advanced-branch-management]]
2558
Advanced branch management
2559
==========================
2561
[[fetching-individual-branches]]
2562
Fetching individual branches
2563
----------------------------
2565
Instead of using gitlink:git-remote[1], you can also choose just
2566
to update one branch at a time, and to store it locally under an
2569
-------------------------------------------------
2570
$ git fetch origin todo:my-todo-work
2571
-------------------------------------------------
2573
The first argument, "origin", just tells git to fetch from the
2574
repository you originally cloned from. The second argument tells git
2575
to fetch the branch named "todo" from the remote repository, and to
2576
store it locally under the name refs/heads/my-todo-work.
2578
You can also fetch branches from other repositories; so
2580
-------------------------------------------------
2581
$ git fetch git://example.com/proj.git master:example-master
2582
-------------------------------------------------
2584
will create a new branch named "example-master" and store in it the
2585
branch named "master" from the repository at the given URL. If you
2586
already have a branch named example-master, it will attempt to
2587
<<fast-forwards,fast-forward>> to the commit given by example.com's
2588
master branch. In more detail:
2590
[[fetch-fast-forwards]]
2591
git fetch and fast-forwards
2592
---------------------------
2594
In the previous example, when updating an existing branch, "git
2595
fetch" checks to make sure that the most recent commit on the remote
2596
branch is a descendant of the most recent commit on your copy of the
2597
branch before updating your copy of the branch to point at the new
2598
commit. Git calls this process a <<fast-forwards,fast forward>>.
2600
A fast forward looks something like this:
2602
................................................
2603
o--o--o--o <-- old head of the branch
2605
o--o--o <-- new head of the branch
2606
................................................
2609
In some cases it is possible that the new head will *not* actually be
2610
a descendant of the old head. For example, the developer may have
2611
realized she made a serious mistake, and decided to backtrack,
2612
resulting in a situation like:
2614
................................................
2615
o--o--o--o--a--b <-- old head of the branch
2617
o--o--o <-- new head of the branch
2618
................................................
2620
In this case, "git fetch" will fail, and print out a warning.
2622
In that case, you can still force git to update to the new head, as
2623
described in the following section. However, note that in the
2624
situation above this may mean losing the commits labeled "a" and "b",
2625
unless you've already created a reference of your own pointing to
2629
Forcing git fetch to do non-fast-forward updates
2630
------------------------------------------------
2632
If git fetch fails because the new head of a branch is not a
2633
descendant of the old head, you may force the update with:
2635
-------------------------------------------------
2636
$ git fetch git://example.com/proj.git +master:refs/remotes/example/master
2637
-------------------------------------------------
2639
Note the addition of the "+" sign. Alternatively, you can use the "-f"
2640
flag to force updates of all the fetched branches, as in:
2642
-------------------------------------------------
2643
$ git fetch -f origin
2644
-------------------------------------------------
2646
Be aware that commits that the old version of example/master pointed at
2647
may be lost, as we saw in the previous section.
2649
[[remote-branch-configuration]]
2650
Configuring remote branches
2651
---------------------------
2653
We saw above that "origin" is just a shortcut to refer to the
2654
repository that you originally cloned from. This information is
2655
stored in git configuration variables, which you can see using
2656
gitlink:git-config[1]:
2658
-------------------------------------------------
2660
core.repositoryformatversion=0
2662
core.logallrefupdates=true
2663
remote.origin.url=git://git.kernel.org/pub/scm/git/git.git
2664
remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*
2665
branch.master.remote=origin
2666
branch.master.merge=refs/heads/master
2667
-------------------------------------------------
2669
If there are other repositories that you also use frequently, you can
2670
create similar configuration options to save typing; for example,
2673
-------------------------------------------------
2674
$ git config remote.example.url git://example.com/proj.git
2675
-------------------------------------------------
2677
then the following two commands will do the same thing:
2679
-------------------------------------------------
2680
$ git fetch git://example.com/proj.git master:refs/remotes/example/master
2681
$ git fetch example master:refs/remotes/example/master
2682
-------------------------------------------------
2684
Even better, if you add one more option:
2686
-------------------------------------------------
2687
$ git config remote.example.fetch master:refs/remotes/example/master
2688
-------------------------------------------------
2690
then the following commands will all do the same thing:
2692
-------------------------------------------------
2693
$ git fetch git://example.com/proj.git master:refs/remotes/example/master
2694
$ git fetch example master:refs/remotes/example/master
2696
-------------------------------------------------
2698
You can also add a "+" to force the update each time:
2700
-------------------------------------------------
2701
$ git config remote.example.fetch +master:ref/remotes/example/master
2702
-------------------------------------------------
2704
Don't do this unless you're sure you won't mind "git fetch" possibly
2705
throwing away commits on mybranch.
2707
Also note that all of the above configuration can be performed by
2708
directly editing the file .git/config instead of using
2709
gitlink:git-config[1].
2711
See gitlink:git-config[1] for more details on the configuration
2712
options mentioned above.
2719
Git is built on a small number of simple but powerful ideas. While it
2720
is possible to get things done without understanding them, you will find
2721
git much more intuitive if you do.
2723
We start with the most important, the <<def_object_database,object
2724
database>> and the <<def_index,index>>.
2726
[[the-object-database]]
2731
We already saw in <<understanding-commits>> that all commits are stored
2732
under a 40-digit "object name". In fact, all the information needed to
2733
represent the history of a project is stored in objects with such names.
2734
In each case the name is calculated by taking the SHA1 hash of the
2735
contents of the object. The SHA1 hash is a cryptographic hash function.
2736
What that means to us is that it is impossible to find two different
2737
objects with the same name. This has a number of advantages; among
2740
- Git can quickly determine whether two objects are identical or not,
2741
just by comparing names.
2742
- Since object names are computed the same way in ever repository, the
2743
same content stored in two repositories will always be stored under
2745
- Git can detect errors when it reads an object, by checking that the
2746
object's name is still the SHA1 hash of its contents.
2748
(See <<object-details>> for the details of the object formatting and
2751
There are four different types of objects: "blob", "tree", "commit", and
2754
- A <<def_blob_object,"blob" object>> is used to store file data.
2755
- A <<def_tree_object,"tree" object>> is an object that ties one or more
2756
"blob" objects into a directory structure. In addition, a tree object
2757
can refer to other tree objects, thus creating a directory hierarchy.
2758
- A <<def_commit_object,"commit" object>> ties such directory hierarchies
2759
together into a <<def_DAG,directed acyclic graph>> of revisions - each
2760
commit contains the object name of exactly one tree designating the
2761
directory hierarchy at the time of the commit. In addition, a commit
2762
refers to "parent" commit objects that describe the history of how we
2763
arrived at that directory hierarchy.
2764
- A <<def_tag_object,"tag" object>> symbolically identifies and can be
2765
used to sign other objects. It contains the object name and type of
2766
another object, a symbolic name (of course!) and, optionally, a
2769
The object types in some more detail:
2775
The "commit" object links a physical state of a tree with a description
2776
of how we got there and why. Use the --pretty=raw option to
2777
gitlink:git-show[1] or gitlink:git-log[1] to examine your favorite
2780
------------------------------------------------
2781
$ git show -s --pretty=raw 2be7fcb476
2782
commit 2be7fcb4764f2dbcee52635b91fedb1b3dcf7ab4
2783
tree fb3a8bdd0ceddd019615af4d57a53f43d8cee2bf
2784
parent 257a84d9d02e90447b149af58b271c19405edb6a
2785
author Dave Watson <dwatson@mimvista.com> 1187576872 -0400
2786
committer Junio C Hamano <gitster@pobox.com> 1187591163 -0700
2788
Fix misspelling of 'suppress' in docs
2790
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2791
------------------------------------------------
2793
As you can see, a commit is defined by:
2795
- a tree: The SHA1 name of a tree object (as defined below), representing
2796
the contents of a directory at a certain point in time.
2797
- parent(s): The SHA1 name of some number of commits which represent the
2798
immediately prevoius step(s) in the history of the project. The
2799
example above has one parent; merge commits may have more than
2800
one. A commit with no parents is called a "root" commit, and
2801
represents the initial revision of a project. Each project must have
2802
at least one root. A project can also have multiple roots, though
2803
that isn't common (or necessarily a good idea).
2804
- an author: The name of the person responsible for this change, together
2806
- a committer: The name of the person who actually created the commit,
2807
with the date it was done. This may be different from the author, for
2808
example, if the author was someone who wrote a patch and emailed it
2809
to the person who used it to create the commit.
2810
- a comment describing this commit.
2812
Note that a commit does not itself contain any information about what
2813
actually changed; all changes are calculated by comparing the contents
2814
of the tree referred to by this commit with the trees associated with
2815
its parents. In particular, git does not attempt to record file renames
2816
explicitly, though it can identify cases where the existence of the same
2817
file data at changing paths suggests a rename. (See, for example, the
2818
-M option to gitlink:git-diff[1]).
2820
A commit is usually created by gitlink:git-commit[1], which creates a
2821
commit whose parent is normally the current HEAD, and whose tree is
2822
taken from the content currently stored in the index.
2828
The ever-versatile gitlink:git-show[1] command can also be used to
2829
examine tree objects, but gitlink:git-ls-tree[1] will give you more
2832
------------------------------------------------
2833
$ git ls-tree fb3a8bdd0ce
2834
100644 blob 63c918c667fa005ff12ad89437f2fdc80926e21c .gitignore
2835
100644 blob 5529b198e8d14decbe4ad99db3f7fb632de0439d .mailmap
2836
100644 blob 6ff87c4664981e4397625791c8ea3bbb5f2279a3 COPYING
2837
040000 tree 2fb783e477100ce076f6bf57e4a6f026013dc745 Documentation
2838
100755 blob 3c0032cec592a765692234f1cba47dfdcc3a9200 GIT-VERSION-GEN
2839
100644 blob 289b046a443c0647624607d471289b2c7dcd470b INSTALL
2840
100644 blob 4eb463797adc693dc168b926b6932ff53f17d0b1 Makefile
2841
100644 blob 548142c327a6790ff8821d67c2ee1eff7a656b52 README
2843
------------------------------------------------
2845
As you can see, a tree object contains a list of entries, each with a
2846
mode, object type, SHA1 name, and name, sorted by name. It represents
2847
the contents of a single directory tree.
2849
The object type may be a blob, representing the contents of a file, or
2850
another tree, representing the contents of a subdirectory. Since trees
2851
and blobs, like all other objects, are named by the SHA1 hash of their
2852
contents, two trees have the same SHA1 name if and only if their
2853
contents (including, recursively, the contents of all subdirectories)
2854
are identical. This allows git to quickly determine the differences
2855
between two related tree objects, since it can ignore any entries with
2856
identical object names.
2858
(Note: in the presence of submodules, trees may also have commits as
2859
entries. See <<submodules>> for documentation.)
2861
Note that the files all have mode 644 or 755: git actually only pays
2862
attention to the executable bit.
2868
You can use gitlink:git-show[1] to examine the contents of a blob; take,
2869
for example, the blob in the entry for "COPYING" from the tree above:
2871
------------------------------------------------
2872
$ git show 6ff87c4664
2874
Note that the only valid version of the GPL as far as this project
2875
is concerned is _this_ particular version of the license (ie v2, not
2876
v2.2 or v3.x or whatever), unless explicitly otherwise stated.
2878
------------------------------------------------
2880
A "blob" object is nothing but a binary blob of data. It doesn't refer
2881
to anything else or have attributes of any kind.
2883
Since the blob is entirely defined by its data, if two files in a
2884
directory tree (or in multiple different versions of the repository)
2885
have the same contents, they will share the same blob object. The object
2886
is totally independent of its location in the directory tree, and
2887
renaming a file does not change the object that file is associated with.
2889
Note that any tree or blob object can be examined using
2890
gitlink:git-show[1] with the <revision>:<path> syntax. This can
2891
sometimes be useful for browsing the contents of a tree that is not
2892
currently checked out.
2898
If you receive the SHA1 name of a blob from one source, and its contents
2899
from another (possibly untrusted) source, you can still trust that those
2900
contents are correct as long as the SHA1 name agrees. This is because
2901
the SHA1 is designed so that it is infeasible to find different contents
2902
that produce the same hash.
2904
Similarly, you need only trust the SHA1 name of a top-level tree object
2905
to trust the contents of the entire directory that it refers to, and if
2906
you receive the SHA1 name of a commit from a trusted source, then you
2907
can easily verify the entire history of commits reachable through
2908
parents of that commit, and all of those contents of the trees referred
2909
to by those commits.
2911
So to introduce some real trust in the system, the only thing you need
2912
to do is to digitally sign just 'one' special note, which includes the
2913
name of a top-level commit. Your digital signature shows others
2914
that you trust that commit, and the immutability of the history of
2915
commits tells others that they can trust the whole history.
2917
In other words, you can easily validate a whole archive by just
2918
sending out a single email that tells the people the name (SHA1 hash)
2919
of the top commit, and digitally sign that email using something
2922
To assist in this, git also provides the tag object...
2928
A tag object contains an object, object type, tag name, the name of the
2929
person ("tagger") who created the tag, and a message, which may contain
2930
a signature, as can be seen using the gitlink:git-cat-file[1]:
2932
------------------------------------------------
2933
$ git cat-file tag v1.5.0
2934
object 437b1b20df4b356c9342dac8d38849f24ef44f27
2937
tagger Junio C Hamano <junkio@cox.net> 1171411200 +0000
2940
-----BEGIN PGP SIGNATURE-----
2941
Version: GnuPG v1.4.6 (GNU/Linux)
2943
iD8DBQBF0lGqwMbZpPMRm5oRAuRiAJ9ohBLd7s2kqjkKlq1qqC57SbnmzQCdG4ui
2944
nLE/L9aUXdWeTFPron96DLA=
2946
-----END PGP SIGNATURE-----
2947
------------------------------------------------
2949
See the gitlink:git-tag[1] command to learn how to create and verify tag
2950
objects. (Note that gitlink:git-tag[1] can also be used to create
2951
"lightweight tags", which are not tag objects at all, but just simple
2952
references whose names begin with "refs/tags/").
2955
How git stores objects efficiently: pack files
2956
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2958
Newly created objects are initially created in a file named after the
2959
object's SHA1 hash (stored in .git/objects).
2961
Unfortunately this system becomes inefficient once a project has a
2962
lot of objects. Try this on an old project:
2964
------------------------------------------------
2966
6930 objects, 47620 kilobytes
2967
------------------------------------------------
2969
The first number is the number of objects which are kept in
2970
individual files. The second is the amount of space taken up by
2971
those "loose" objects.
2973
You can save space and make git faster by moving these loose objects in
2974
to a "pack file", which stores a group of objects in an efficient
2975
compressed format; the details of how pack files are formatted can be
2976
found in link:technical/pack-format.txt[technical/pack-format.txt].
2978
To put the loose objects into a pack, just run git repack:
2980
------------------------------------------------
2983
Done counting 6020 objects.
2984
Deltifying 6020 objects.
2985
100% (6020/6020) done
2986
Writing 6020 objects.
2987
100% (6020/6020) done
2988
Total 6020, written 6020 (delta 4070), reused 0 (delta 0)
2989
Pack pack-3e54ad29d5b2e05838c75df582c65257b8d08e1c created.
2990
------------------------------------------------
2994
------------------------------------------------
2996
------------------------------------------------
2998
to remove any of the "loose" objects that are now contained in the
2999
pack. This will also remove any unreferenced objects (which may be
3000
created when, for example, you use "git reset" to remove a commit).
3001
You can verify that the loose objects are gone by looking at the
3002
.git/objects directory or by running
3004
------------------------------------------------
3006
0 objects, 0 kilobytes
3007
------------------------------------------------
3009
Although the object files are gone, any commands that refer to those
3010
objects will work exactly as they did before.
3012
The gitlink:git-gc[1] command performs packing, pruning, and more for
3013
you, so is normally the only high-level command you need.
3015
[[dangling-objects]]
3019
The gitlink:git-fsck[1] command will sometimes complain about dangling
3020
objects. They are not a problem.
3022
The most common cause of dangling objects is that you've rebased a
3023
branch, or you have pulled from somebody else who rebased a branch--see
3024
<<cleaning-up-history>>. In that case, the old head of the original
3025
branch still exists, as does everything it pointed to. The branch
3026
pointer itself just doesn't, since you replaced it with another one.
3028
There are also other situations that cause dangling objects. For
3029
example, a "dangling blob" may arise because you did a "git add" of a
3030
file, but then, before you actually committed it and made it part of the
3031
bigger picture, you changed something else in that file and committed
3032
that *updated* thing - the old state that you added originally ends up
3033
not being pointed to by any commit or tree, so it's now a dangling blob
3036
Similarly, when the "recursive" merge strategy runs, and finds that
3037
there are criss-cross merges and thus more than one merge base (which is
3038
fairly unusual, but it does happen), it will generate one temporary
3039
midway tree (or possibly even more, if you had lots of criss-crossing
3040
merges and more than two merge bases) as a temporary internal merge
3041
base, and again, those are real objects, but the end result will not end
3042
up pointing to them, so they end up "dangling" in your repository.
3044
Generally, dangling objects aren't anything to worry about. They can
3045
even be very useful: if you screw something up, the dangling objects can
3046
be how you recover your old tree (say, you did a rebase, and realized
3047
that you really didn't want to - you can look at what dangling objects
3048
you have, and decide to reset your head to some old dangling state).
3050
For commits, you can just use:
3052
------------------------------------------------
3053
$ gitk <dangling-commit-sha-goes-here> --not --all
3054
------------------------------------------------
3056
This asks for all the history reachable from the given commit but not
3057
from any branch, tag, or other reference. If you decide it's something
3058
you want, you can always create a new reference to it, e.g.,
3060
------------------------------------------------
3061
$ git branch recovered-branch <dangling-commit-sha-goes-here>
3062
------------------------------------------------
3064
For blobs and trees, you can't do the same, but you can still examine
3065
them. You can just do
3067
------------------------------------------------
3068
$ git show <dangling-blob/tree-sha-goes-here>
3069
------------------------------------------------
3071
to show what the contents of the blob were (or, for a tree, basically
3072
what the "ls" for that directory was), and that may give you some idea
3073
of what the operation was that left that dangling object.
3075
Usually, dangling blobs and trees aren't very interesting. They're
3076
almost always the result of either being a half-way mergebase (the blob
3077
will often even have the conflict markers from a merge in it, if you
3078
have had conflicting merges that you fixed up by hand), or simply
3079
because you interrupted a "git fetch" with ^C or something like that,
3080
leaving _some_ of the new objects in the object database, but just
3081
dangling and useless.
3083
Anyway, once you are sure that you're not interested in any dangling
3084
state, you can just prune all unreachable objects:
3086
------------------------------------------------
3088
------------------------------------------------
3090
and they'll be gone. But you should only run "git prune" on a quiescent
3091
repository - it's kind of like doing a filesystem fsck recovery: you
3092
don't want to do that while the filesystem is mounted.
3094
(The same is true of "git-fsck" itself, btw - but since
3095
git-fsck never actually *changes* the repository, it just reports
3096
on what it found, git-fsck itself is never "dangerous" to run.
3097
Running it while somebody is actually changing the repository can cause
3098
confusing and scary messages, but it won't actually do anything bad. In
3099
contrast, running "git prune" while somebody is actively changing the
3100
repository is a *BAD* idea).
3106
The index is a binary file (generally kept in .git/index) containing a
3107
sorted list of path names, each with permissions and the SHA1 of a blob
3108
object; gitlink:git-ls-files[1] can show you the contents of the index:
3110
-------------------------------------------------
3111
$ git ls-files --stage
3112
100644 63c918c667fa005ff12ad89437f2fdc80926e21c 0 .gitignore
3113
100644 5529b198e8d14decbe4ad99db3f7fb632de0439d 0 .mailmap
3114
100644 6ff87c4664981e4397625791c8ea3bbb5f2279a3 0 COPYING
3115
100644 a37b2152bd26be2c2289e1f57a292534a51a93c7 0 Documentation/.gitignore
3116
100644 fbefe9a45b00a54b58d94d06eca48b03d40a50e0 0 Documentation/Makefile
3118
100644 2511aef8d89ab52be5ec6a5e46236b4b6bcd07ea 0 xdiff/xtypes.h
3119
100644 2ade97b2574a9f77e7ae4002a4e07a6a38e46d07 0 xdiff/xutils.c
3120
100644 d5de8292e05e7c36c4b68857c1cf9855e3d2f70a 0 xdiff/xutils.h
3121
-------------------------------------------------
3123
Note that in older documentation you may see the index called the
3124
"current directory cache" or just the "cache". It has three important
3127
1. The index contains all the information necessary to generate a single
3128
(uniquely determined) tree object.
3130
For example, running gitlink:git-commit[1] generates this tree object
3131
from the index, stores it in the object database, and uses it as the
3132
tree object associated with the new commit.
3134
2. The index enables fast comparisons between the tree object it defines
3135
and the working tree.
3137
It does this by storing some additional data for each entry (such as
3138
the last modified time). This data is not displayed above, and is not
3139
stored in the created tree object, but it can be used to determine
3140
quickly which files in the working directory differ from what was
3141
stored in the index, and thus save git from having to read all of the
3142
data from such files to look for changes.
3144
3. It can efficiently represent information about merge conflicts
3145
between different tree objects, allowing each pathname to be
3146
associated with sufficient information about the trees involved that
3147
you can create a three-way merge between them.
3149
We saw in <<conflict-resolution>> that during a merge the index can
3150
store multiple versions of a single file (called "stages"). The third
3151
column in the gitlink:git-ls-files[1] output above is the stage
3152
number, and will take on values other than 0 for files with merge
3155
The index is thus a sort of temporary staging area, which is filled with
3156
a tree which you are in the process of working on.
3158
If you blow the index away entirely, you generally haven't lost any
3159
information as long as you have the name of the tree that it described.
3165
Large projects are often composed of smaller, self-contained modules. For
3166
example, an embedded Linux distribution's source tree would include every
3167
piece of software in the distribution with some local modifications; a movie
3168
player might need to build against a specific, known-working version of a
3169
decompression library; several independent programs might all share the same
3172
With centralized revision control systems this is often accomplished by
3173
including every module in one single repository. Developers can check out
3174
all modules or only the modules they need to work with. They can even modify
3175
files across several modules in a single commit while moving things around
3176
or updating APIs and translations.
3178
Git does not allow partial checkouts, so duplicating this approach in Git
3179
would force developers to keep a local copy of modules they are not
3180
interested in touching. Commits in an enormous checkout would be slower
3181
than you'd expect as Git would have to scan every directory for changes.
3182
If modules have a lot of local history, clones would take forever.
3184
On the plus side, distributed revision control systems can much better
3185
integrate with external sources. In a centralized model, a single arbitrary
3186
snapshot of the external project is exported from its own revision control
3187
and then imported into the local revision control on a vendor branch. All
3188
the history is hidden. With distributed revision control you can clone the
3189
entire external history and much more easily follow development and re-merge
3192
Git's submodule support allows a repository to contain, as a subdirectory, a
3193
checkout of an external project. Submodules maintain their own identity;
3194
the submodule support just stores the submodule repository location and
3195
commit ID, so other developers who clone the containing project
3196
("superproject") can easily clone all the submodules at the same revision.
3197
Partial checkouts of the superproject are possible: you can tell Git to
3198
clone none, some or all of the submodules.
3200
The gitlink:git-submodule[1] command is available since Git 1.5.3. Users
3201
with Git 1.5.2 can look up the submodule commits in the repository and
3202
manually check them out; earlier versions won't recognize the submodules at
3205
To see how submodule support works, create (for example) four example
3206
repositories that can be used later as a submodule:
3208
-------------------------------------------------
3216
echo "module $i" > $i.txt
3218
git commit -m "Initial commit, submodule $i"
3221
-------------------------------------------------
3223
Now create the superproject and add all the submodules:
3225
-------------------------------------------------
3231
git submodule add ~/git/$i
3233
-------------------------------------------------
3235
NOTE: Do not use local URLs here if you plan to publish your superproject!
3237
See what files `git submodule` created:
3239
-------------------------------------------------
3241
. .. .git .gitmodules a b c d
3242
-------------------------------------------------
3244
The `git submodule add` command does a couple of things:
3246
- It clones the submodule under the current directory and by default checks out
3248
- It adds the submodule's clone path to the gitlink:gitmodules[5] file and
3249
adds this file to the index, ready to be committed.
3250
- It adds the submodule's current commit ID to the index, ready to be
3253
Commit the superproject:
3255
-------------------------------------------------
3256
$ git commit -m "Add submodules a, b, c and d."
3257
-------------------------------------------------
3259
Now clone the superproject:
3261
-------------------------------------------------
3263
$ git clone super cloned
3265
-------------------------------------------------
3267
The submodule directories are there, but they're empty:
3269
-------------------------------------------------
3272
$ git submodule status
3273
-d266b9873ad50488163457f025db7cdd9683d88b a
3274
-e81d457da15309b4fef4249aba9b50187999670d b
3275
-c1536a972b9affea0f16e0680ba87332dc059146 c
3276
-d96249ff5d57de5de093e6baff9e0aafa5276a74 d
3277
-------------------------------------------------
3279
NOTE: The commit object names shown above would be different for you, but they
3280
should match the HEAD commit object names of your repositories. You can check
3281
it by running `git ls-remote ../a`.
3283
Pulling down the submodules is a two-step process. First run `git submodule
3284
init` to add the submodule repository URLs to `.git/config`:
3286
-------------------------------------------------
3287
$ git submodule init
3288
-------------------------------------------------
3290
Now use `git submodule update` to clone the repositories and check out the
3291
commits specified in the superproject:
3293
-------------------------------------------------
3294
$ git submodule update
3298
-------------------------------------------------
3300
One major difference between `git submodule update` and `git submodule add` is
3301
that `git submodule update` checks out a specific commit, rather than the tip
3302
of a branch. It's like checking out a tag: the head is detached, so you're not
3303
working on a branch.
3305
-------------------------------------------------
3309
-------------------------------------------------
3311
If you want to make a change within a submodule and you have a detached head,
3312
then you should create or checkout a branch, make your changes, publish the
3313
change within the submodule, and then update the superproject to reference the
3316
-------------------------------------------------
3317
$ git checkout master
3318
-------------------------------------------------
3322
-------------------------------------------------
3323
$ git checkout -b fix-up
3324
-------------------------------------------------
3328
-------------------------------------------------
3329
$ echo "adding a line again" >> a.txt
3330
$ git commit -a -m "Updated the submodule from within the superproject."
3335
index d266b98..261dfac 160000
3339
-Subproject commit d266b9873ad50488163457f025db7cdd9683d88b
3340
+Subproject commit 261dfac35cb99d380eb966e102c1197139f7fa24
3342
$ git commit -m "Updated submodule a."
3344
-------------------------------------------------
3346
You have to run `git submodule update` after `git pull` if you want to update
3349
Pitfalls with submodules
3350
------------------------
3352
Always publish the submodule change before publishing the change to the
3353
superproject that references it. If you forget to publish the submodule change,
3354
others won't be able to clone the repository:
3356
-------------------------------------------------
3358
$ echo i added another line to this file >> a.txt
3359
$ git commit -a -m "doing it wrong this time"
3362
$ git commit -m "Updated submodule a again."
3366
$ git submodule update
3367
error: pathspec '261dfac35cb99d380eb966e102c1197139f7fa24' did not match any file(s) known to git.
3368
Did you forget to 'git add'?
3369
Unable to checkout '261dfac35cb99d380eb966e102c1197139f7fa24' in submodule path 'a'
3370
-------------------------------------------------
3372
You also should not rewind branches in a submodule beyond commits that were
3373
ever recorded in any superproject.
3375
It's not safe to run `git submodule update` if you've made and committed
3376
changes within a submodule without checking out a branch first. They will be
3377
silently overwritten:
3379
-------------------------------------------------
3382
$ echo line added from private2 >> a.txt
3383
$ git commit -a -m "line added inside private2"
3385
$ git submodule update
3386
Submodule path 'a': checked out 'd266b9873ad50488163457f025db7cdd9683d88b'
3390
-------------------------------------------------
3392
NOTE: The changes are still visible in the submodule's reflog.
3394
This is not the case if you did not commit your changes.
3396
[[low-level-operations]]
3397
Low-level git operations
3398
========================
3400
Many of the higher-level commands were originally implemented as shell
3401
scripts using a smaller core of low-level git commands. These can still
3402
be useful when doing unusual things with git, or just as a way to
3403
understand its inner workings.
3405
[[object-manipulation]]
3406
Object access and manipulation
3407
------------------------------
3409
The gitlink:git-cat-file[1] command can show the contents of any object,
3410
though the higher-level gitlink:git-show[1] is usually more useful.
3412
The gitlink:git-commit-tree[1] command allows constructing commits with
3413
arbitrary parents and trees.
3415
A tree can be created with gitlink:git-write-tree[1] and its data can be
3416
accessed by gitlink:git-ls-tree[1]. Two trees can be compared with
3417
gitlink:git-diff-tree[1].
3419
A tag is created with gitlink:git-mktag[1], and the signature can be
3420
verified by gitlink:git-verify-tag[1], though it is normally simpler to
3421
use gitlink:git-tag[1] for both.
3427
High-level operations such as gitlink:git-commit[1],
3428
gitlink:git-checkout[1] and git-reset[1] work by moving data between the
3429
working tree, the index, and the object database. Git provides
3430
low-level operations which perform each of these steps individually.
3432
Generally, all "git" operations work on the index file. Some operations
3433
work *purely* on the index file (showing the current state of the
3434
index), but most operations move data between the index file and either
3435
the database or the working directory. Thus there are four main
3438
[[working-directory-to-index]]
3439
working directory -> index
3440
~~~~~~~~~~~~~~~~~~~~~~~~~~
3442
The gitlink:git-update-index[1] command updates the index with
3443
information from the working directory. You generally update the
3444
index information by just specifying the filename you want to update,
3447
-------------------------------------------------
3448
$ git update-index filename
3449
-------------------------------------------------
3451
but to avoid common mistakes with filename globbing etc, the command
3452
will not normally add totally new entries or remove old entries,
3453
i.e. it will normally just update existing cache entries.
3455
To tell git that yes, you really do realize that certain files no
3456
longer exist, or that new files should be added, you
3457
should use the `--remove` and `--add` flags respectively.
3459
NOTE! A `--remove` flag does 'not' mean that subsequent filenames will
3460
necessarily be removed: if the files still exist in your directory
3461
structure, the index will be updated with their new status, not
3462
removed. The only thing `--remove` means is that update-cache will be
3463
considering a removed file to be a valid thing, and if the file really
3464
does not exist any more, it will update the index accordingly.
3466
As a special case, you can also do `git-update-index --refresh`, which
3467
will refresh the "stat" information of each index to match the current
3468
stat information. It will 'not' update the object status itself, and
3469
it will only update the fields that are used to quickly test whether
3470
an object still matches its old backing store object.
3472
The previously introduced gitlink:git-add[1] is just a wrapper for
3473
gitlink:git-update-index[1].
3475
[[index-to-object-database]]
3476
index -> object database
3477
~~~~~~~~~~~~~~~~~~~~~~~~
3479
You write your current index file to a "tree" object with the program
3481
-------------------------------------------------
3483
-------------------------------------------------
3485
that doesn't come with any options - it will just write out the
3486
current index into the set of tree objects that describe that state,
3487
and it will return the name of the resulting top-level tree. You can
3488
use that tree to re-generate the index at any time by going in the
3491
[[object-database-to-index]]
3492
object database -> index
3493
~~~~~~~~~~~~~~~~~~~~~~~~
3495
You read a "tree" file from the object database, and use that to
3496
populate (and overwrite - don't do this if your index contains any
3497
unsaved state that you might want to restore later!) your current
3498
index. Normal operation is just
3500
-------------------------------------------------
3501
$ git-read-tree <sha1 of tree>
3502
-------------------------------------------------
3504
and your index file will now be equivalent to the tree that you saved
3505
earlier. However, that is only your 'index' file: your working
3506
directory contents have not been modified.
3508
[[index-to-working-directory]]
3509
index -> working directory
3510
~~~~~~~~~~~~~~~~~~~~~~~~~~
3512
You update your working directory from the index by "checking out"
3513
files. This is not a very common operation, since normally you'd just
3514
keep your files updated, and rather than write to your working
3515
directory, you'd tell the index files about the changes in your
3516
working directory (i.e. `git-update-index`).
3518
However, if you decide to jump to a new version, or check out somebody
3519
else's version, or just restore a previous tree, you'd populate your
3520
index file with read-tree, and then you need to check out the result
3523
-------------------------------------------------
3524
$ git-checkout-index filename
3525
-------------------------------------------------
3527
or, if you want to check out all of the index, use `-a`.
3529
NOTE! git-checkout-index normally refuses to overwrite old files, so
3530
if you have an old version of the tree already checked out, you will
3531
need to use the "-f" flag ('before' the "-a" flag or the filename) to
3532
'force' the checkout.
3535
Finally, there are a few odds and ends which are not purely moving
3536
from one representation to the other:
3538
[[tying-it-all-together]]
3539
Tying it all together
3540
~~~~~~~~~~~~~~~~~~~~~
3542
To commit a tree you have instantiated with "git-write-tree", you'd
3543
create a "commit" object that refers to that tree and the history
3544
behind it - most notably the "parent" commits that preceded it in
3547
Normally a "commit" has one parent: the previous state of the tree
3548
before a certain change was made. However, sometimes it can have two
3549
or more parent commits, in which case we call it a "merge", due to the
3550
fact that such a commit brings together ("merges") two or more
3551
previous states represented by other commits.
3553
In other words, while a "tree" represents a particular directory state
3554
of a working directory, a "commit" represents that state in "time",
3555
and explains how we got there.
3557
You create a commit object by giving it the tree that describes the
3558
state at the time of the commit, and a list of parents:
3560
-------------------------------------------------
3561
$ git-commit-tree <tree> -p <parent> [-p <parent2> ..]
3562
-------------------------------------------------
3564
and then giving the reason for the commit on stdin (either through
3565
redirection from a pipe or file, or by just typing it at the tty).
3567
git-commit-tree will return the name of the object that represents
3568
that commit, and you should save it away for later use. Normally,
3569
you'd commit a new `HEAD` state, and while git doesn't care where you
3570
save the note about that state, in practice we tend to just write the
3571
result to the file pointed at by `.git/HEAD`, so that we can always see
3572
what the last committed state was.
3574
Here is an ASCII art by Jon Loeliger that illustrates how
3575
various pieces fit together.
3603
checkout-index -u | | checkout-index
3614
[[examining-the-data]]
3618
You can examine the data represented in the object database and the
3619
index with various helper tools. For every object, you can use
3620
gitlink:git-cat-file[1] to examine details about the
3623
-------------------------------------------------
3624
$ git-cat-file -t <objectname>
3625
-------------------------------------------------
3627
shows the type of the object, and once you have the type (which is
3628
usually implicit in where you find the object), you can use
3630
-------------------------------------------------
3631
$ git-cat-file blob|tree|commit|tag <objectname>
3632
-------------------------------------------------
3634
to show its contents. NOTE! Trees have binary content, and as a result
3635
there is a special helper for showing that content, called
3636
`git-ls-tree`, which turns the binary content into a more easily
3639
It's especially instructive to look at "commit" objects, since those
3640
tend to be small and fairly self-explanatory. In particular, if you
3641
follow the convention of having the top commit name in `.git/HEAD`,
3644
-------------------------------------------------
3645
$ git-cat-file commit HEAD
3646
-------------------------------------------------
3648
to see what the top commit was.
3650
[[merging-multiple-trees]]
3651
Merging multiple trees
3652
----------------------
3654
Git helps you do a three-way merge, which you can expand to n-way by
3655
repeating the merge procedure arbitrary times until you finally
3656
"commit" the state. The normal situation is that you'd only do one
3657
three-way merge (two parents), and commit it, but if you like to, you
3658
can do multiple parents in one go.
3660
To do a three-way merge, you need the two sets of "commit" objects
3661
that you want to merge, use those to find the closest common parent (a
3662
third "commit" object), and then use those commit objects to find the
3663
state of the directory ("tree" object) at these points.
3665
To get the "base" for the merge, you first look up the common parent
3668
-------------------------------------------------
3669
$ git-merge-base <commit1> <commit2>
3670
-------------------------------------------------
3672
which will return you the commit they are both based on. You should
3673
now look up the "tree" objects of those commits, which you can easily
3674
do with (for example)
3676
-------------------------------------------------
3677
$ git-cat-file commit <commitname> | head -1
3678
-------------------------------------------------
3680
since the tree object information is always the first line in a commit
3683
Once you know the three trees you are going to merge (the one "original"
3684
tree, aka the common tree, and the two "result" trees, aka the branches
3685
you want to merge), you do a "merge" read into the index. This will
3686
complain if it has to throw away your old index contents, so you should
3687
make sure that you've committed those - in fact you would normally
3688
always do a merge against your last commit (which should thus match what
3689
you have in your current index anyway).
3693
-------------------------------------------------
3694
$ git-read-tree -m -u <origtree> <yourtree> <targettree>
3695
-------------------------------------------------
3697
which will do all trivial merge operations for you directly in the
3698
index file, and you can just write the result out with
3702
[[merging-multiple-trees-2]]
3703
Merging multiple trees, continued
3704
---------------------------------
3706
Sadly, many merges aren't trivial. If there are files that have
3707
been added.moved or removed, or if both branches have modified the
3708
same file, you will be left with an index tree that contains "merge
3709
entries" in it. Such an index tree can 'NOT' be written out to a tree
3710
object, and you will have to resolve any such merge clashes using
3711
other tools before you can write out the result.
3713
You can examine such index state with `git-ls-files --unmerged`
3714
command. An example:
3716
------------------------------------------------
3717
$ git-read-tree -m $orig HEAD $target
3718
$ git-ls-files --unmerged
3719
100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello.c
3720
100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello.c
3721
100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello.c
3722
------------------------------------------------
3724
Each line of the `git-ls-files --unmerged` output begins with
3725
the blob mode bits, blob SHA1, 'stage number', and the
3726
filename. The 'stage number' is git's way to say which tree it
3727
came from: stage 1 corresponds to `$orig` tree, stage 2 `HEAD`
3728
tree, and stage3 `$target` tree.
3730
Earlier we said that trivial merges are done inside
3731
`git-read-tree -m`. For example, if the file did not change
3732
from `$orig` to `HEAD` nor `$target`, or if the file changed
3733
from `$orig` to `HEAD` and `$orig` to `$target` the same way,
3734
obviously the final outcome is what is in `HEAD`. What the
3735
above example shows is that file `hello.c` was changed from
3736
`$orig` to `HEAD` and `$orig` to `$target` in a different way.
3737
You could resolve this by running your favorite 3-way merge
3738
program, e.g. `diff3`, `merge`, or git's own merge-file, on
3739
the blob objects from these three stages yourself, like this:
3741
------------------------------------------------
3742
$ git-cat-file blob 263414f... >hello.c~1
3743
$ git-cat-file blob 06fa6a2... >hello.c~2
3744
$ git-cat-file blob cc44c73... >hello.c~3
3745
$ git merge-file hello.c~2 hello.c~1 hello.c~3
3746
------------------------------------------------
3748
This would leave the merge result in `hello.c~2` file, along
3749
with conflict markers if there are conflicts. After verifying
3750
the merge result makes sense, you can tell git what the final
3751
merge result for this file is by:
3753
-------------------------------------------------
3754
$ mv -f hello.c~2 hello.c
3755
$ git-update-index hello.c
3756
-------------------------------------------------
3758
When a path is in unmerged state, running `git-update-index` for
3759
that path tells git to mark the path resolved.
3761
The above is the description of a git merge at the lowest level,
3762
to help you understand what conceptually happens under the hood.
3763
In practice, nobody, not even git itself, uses three `git-cat-file`
3764
for this. There is `git-merge-index` program that extracts the
3765
stages to temporary files and calls a "merge" script on it:
3767
-------------------------------------------------
3768
$ git-merge-index git-merge-one-file hello.c
3769
-------------------------------------------------
3771
and that is what higher level `git merge -s resolve` is implemented with.
3777
This chapter covers internal details of the git implementation which
3778
probably only git developers need to understand.
3781
Object storage format
3782
---------------------
3784
All objects have a statically determined "type" which identifies the
3785
format of the object (i.e. how it is used, and how it can refer to other
3786
objects). There are currently four different object types: "blob",
3787
"tree", "commit", and "tag".
3789
Regardless of object type, all objects share the following
3790
characteristics: they are all deflated with zlib, and have a header
3791
that not only specifies their type, but also provides size information
3792
about the data in the object. It's worth noting that the SHA1 hash
3793
that is used to name the object is the hash of the original data
3794
plus this header, so `sha1sum` 'file' does not match the object name
3796
(Historical note: in the dawn of the age of git the hash
3797
was the sha1 of the 'compressed' object.)
3799
As a result, the general consistency of an object can always be tested
3800
independently of the contents or the type of the object: all objects can
3801
be validated by verifying that (a) their hashes match the content of the
3802
file and (b) the object successfully inflates to a stream of bytes that
3803
forms a sequence of <ascii type without space> {plus} <space> {plus} <ascii decimal
3804
size> {plus} <byte\0> {plus} <binary object data>.
3806
The structured objects can further have their structure and
3807
connectivity to other objects verified. This is generally done with
3808
the `git-fsck` program, which generates a full dependency graph
3809
of all objects, and verifies their internal consistency (in addition
3810
to just verifying their superficial consistency through the hash).
3812
[[birdview-on-the-source-code]]
3813
A birds-eye view of Git's source code
3814
-------------------------------------
3816
It is not always easy for new developers to find their way through Git's
3817
source code. This section gives you a little guidance to show where to
3820
A good place to start is with the contents of the initial commit, with:
3822
----------------------------------------------------
3823
$ git checkout e83c5163
3824
----------------------------------------------------
3826
The initial revision lays the foundation for almost everything git has
3827
today, but is small enough to read in one sitting.
3829
Note that terminology has changed since that revision. For example, the
3830
README in that revision uses the word "changeset" to describe what we
3831
now call a <<def_commit_object,commit>>.
3833
Also, we do not call it "cache" any more, but "index", however, the
3834
file is still called `cache.h`. Remark: Not much reason to change it now,
3835
especially since there is no good single name for it anyway, because it is
3836
basically _the_ header file which is included by _all_ of Git's C sources.
3838
If you grasp the ideas in that initial commit, you should check out a
3839
more recent version and skim `cache.h`, `object.h` and `commit.h`.
3841
In the early days, Git (in the tradition of UNIX) was a bunch of programs
3842
which were extremely simple, and which you used in scripts, piping the
3843
output of one into another. This turned out to be good for initial
3844
development, since it was easier to test new things. However, recently
3845
many of these parts have become builtins, and some of the core has been
3846
"libified", i.e. put into libgit.a for performance, portability reasons,
3847
and to avoid code duplication.
3849
By now, you know what the index is (and find the corresponding data
3850
structures in `cache.h`), and that there are just a couple of object types
3851
(blobs, trees, commits and tags) which inherit their common structure from
3852
`struct object`, which is their first member (and thus, you can cast e.g.
3853
`(struct object *)commit` to achieve the _same_ as `&commit->object`, i.e.
3854
get at the object name and flags).
3856
Now is a good point to take a break to let this information sink in.
3858
Next step: get familiar with the object naming. Read <<naming-commits>>.
3859
There are quite a few ways to name an object (and not only revisions!).
3860
All of these are handled in `sha1_name.c`. Just have a quick look at
3861
the function `get_sha1()`. A lot of the special handling is done by
3862
functions like `get_sha1_basic()` or the likes.
3864
This is just to get you into the groove for the most libified part of Git:
3865
the revision walker.
3867
Basically, the initial version of `git log` was a shell script:
3869
----------------------------------------------------------------
3870
$ git-rev-list --pretty $(git-rev-parse --default HEAD "$@") | \
3871
LESS=-S ${PAGER:-less}
3872
----------------------------------------------------------------
3874
What does this mean?
3876
`git-rev-list` is the original version of the revision walker, which
3877
_always_ printed a list of revisions to stdout. It is still functional,
3878
and needs to, since most new Git programs start out as scripts using
3881
`git-rev-parse` is not as important any more; it was only used to filter out
3882
options that were relevant for the different plumbing commands that were
3883
called by the script.
3885
Most of what `git-rev-list` did is contained in `revision.c` and
3886
`revision.h`. It wraps the options in a struct named `rev_info`, which
3887
controls how and what revisions are walked, and more.
3889
The original job of `git-rev-parse` is now taken by the function
3890
`setup_revisions()`, which parses the revisions and the common command line
3891
options for the revision walker. This information is stored in the struct
3892
`rev_info` for later consumption. You can do your own command line option
3893
parsing after calling `setup_revisions()`. After that, you have to call
3894
`prepare_revision_walk()` for initialization, and then you can get the
3895
commits one by one with the function `get_revision()`.
3897
If you are interested in more details of the revision walking process,
3898
just have a look at the first implementation of `cmd_log()`; call
3899
`git-show v1.3.0~155^2~4` and scroll down to that function (note that you
3900
no longer need to call `setup_pager()` directly).
3902
Nowadays, `git log` is a builtin, which means that it is _contained_ in the
3903
command `git`. The source side of a builtin is
3905
- a function called `cmd_<bla>`, typically defined in `builtin-<bla>.c`,
3906
and declared in `builtin.h`,
3908
- an entry in the `commands[]` array in `git.c`, and
3910
- an entry in `BUILTIN_OBJECTS` in the `Makefile`.
3912
Sometimes, more than one builtin is contained in one source file. For
3913
example, `cmd_whatchanged()` and `cmd_log()` both reside in `builtin-log.c`,
3914
since they share quite a bit of code. In that case, the commands which are
3915
_not_ named like the `.c` file in which they live have to be listed in
3916
`BUILT_INS` in the `Makefile`.
3918
`git log` looks more complicated in C than it does in the original script,
3919
but that allows for a much greater flexibility and performance.
3921
Here again it is a good point to take a pause.
3923
Lesson three is: study the code. Really, it is the best way to learn about
3924
the organization of Git (after you know the basic concepts).
3926
So, think about something which you are interested in, say, "how can I
3927
access a blob just knowing the object name of it?". The first step is to
3928
find a Git command with which you can do it. In this example, it is either
3929
`git show` or `git cat-file`.
3931
For the sake of clarity, let's stay with `git cat-file`, because it
3935
- was around even in the initial commit (it literally went only through
3936
some 20 revisions as `cat-file.c`, was renamed to `builtin-cat-file.c`
3937
when made a builtin, and then saw less than 10 versions).
3939
So, look into `builtin-cat-file.c`, search for `cmd_cat_file()` and look what
3942
------------------------------------------------------------------
3943
git_config(git_default_config);
3945
usage("git-cat-file [-t|-s|-e|-p|<type>] <sha1>");
3946
if (get_sha1(argv[2], sha1))
3947
die("Not a valid object name %s", argv[2]);
3948
------------------------------------------------------------------
3950
Let's skip over the obvious details; the only really interesting part
3951
here is the call to `get_sha1()`. It tries to interpret `argv[2]` as an
3952
object name, and if it refers to an object which is present in the current
3953
repository, it writes the resulting SHA-1 into the variable `sha1`.
3955
Two things are interesting here:
3957
- `get_sha1()` returns 0 on _success_. This might surprise some new
3958
Git hackers, but there is a long tradition in UNIX to return different
3959
negative numbers in case of different errors -- and 0 on success.
3961
- the variable `sha1` in the function signature of `get_sha1()` is `unsigned
3962
char \*`, but is actually expected to be a pointer to `unsigned
3963
char[20]`. This variable will contain the 160-bit SHA-1 of the given
3964
commit. Note that whenever a SHA-1 is passed as `unsigned char \*`, it
3965
is the binary representation, as opposed to the ASCII representation in
3966
hex characters, which is passed as `char *`.
3968
You will see both of these things throughout the code.
3972
-----------------------------------------------------------------------------
3974
buf = read_object_with_reference(sha1, argv[1], &size, NULL);
3975
-----------------------------------------------------------------------------
3977
This is how you read a blob (actually, not only a blob, but any type of
3978
object). To know how the function `read_object_with_reference()` actually
3979
works, find the source code for it (something like `git grep
3980
read_object_with | grep ":[a-z]"` in the git repository), and read
3983
To find out how the result can be used, just read on in `cmd_cat_file()`:
3985
-----------------------------------
3986
write_or_die(1, buf, size);
3987
-----------------------------------
3989
Sometimes, you do not know where to look for a feature. In many such cases,
3990
it helps to search through the output of `git log`, and then `git show` the
3991
corresponding commit.
3993
Example: If you know that there was some test case for `git bundle`, but
3994
do not remember where it was (yes, you _could_ `git grep bundle t/`, but that
3995
does not illustrate the point!):
3997
------------------------
3998
$ git log --no-merges t/
3999
------------------------
4001
In the pager (`less`), just search for "bundle", go a few lines back,
4002
and see that it is in commit 18449ab0... Now just copy this object name,
4003
and paste it into the command line
4011
Another example: Find out what to do in order to make some script a
4014
-------------------------------------------------
4015
$ git log --no-merges --diff-filter=A builtin-*.c
4016
-------------------------------------------------
4018
You see, Git is actually the best tool to find out about the source of Git
4022
include::glossary.txt[]
4025
Appendix A: Git Quick Reference
4026
===============================
4028
This is a quick summary of the major commands; the previous chapters
4029
explain how these work in more detail.
4031
[[quick-creating-a-new-repository]]
4032
Creating a new repository
4033
-------------------------
4037
-----------------------------------------------
4038
$ tar xzf project.tar.gz
4041
Initialized empty Git repository in .git/
4044
-----------------------------------------------
4046
From a remote repository:
4048
-----------------------------------------------
4049
$ git clone git://example.com/pub/project.git
4051
-----------------------------------------------
4053
[[managing-branches]]
4057
-----------------------------------------------
4058
$ git branch # list all local branches in this repo
4059
$ git checkout test # switch working directory to branch "test"
4060
$ git branch new # create branch "new" starting at current HEAD
4061
$ git branch -d new # delete branch "new"
4062
-----------------------------------------------
4064
Instead of basing new branch on current HEAD (the default), use:
4066
-----------------------------------------------
4067
$ git branch new test # branch named "test"
4068
$ git branch new v2.6.15 # tag named v2.6.15
4069
$ git branch new HEAD^ # commit before the most recent
4070
$ git branch new HEAD^^ # commit before that
4071
$ git branch new test~10 # ten commits before tip of branch "test"
4072
-----------------------------------------------
4074
Create and switch to a new branch at the same time:
4076
-----------------------------------------------
4077
$ git checkout -b new v2.6.15
4078
-----------------------------------------------
4080
Update and examine branches from the repository you cloned from:
4082
-----------------------------------------------
4083
$ git fetch # update
4084
$ git branch -r # list
4088
$ git checkout -b masterwork origin/master
4089
-----------------------------------------------
4091
Fetch a branch from a different repository, and give it a new
4092
name in your repository:
4094
-----------------------------------------------
4095
$ git fetch git://example.com/project.git theirbranch:mybranch
4096
$ git fetch git://example.com/project.git v2.6.15:mybranch
4097
-----------------------------------------------
4099
Keep a list of repositories you work with regularly:
4101
-----------------------------------------------
4102
$ git remote add example git://example.com/project.git
4103
$ git remote # list remote repositories
4106
$ git remote show example # get details
4108
URL: git://example.com/project.git
4109
Tracked remote branches
4111
$ git fetch example # update branches from example
4112
$ git branch -r # list all remote branches
4113
-----------------------------------------------
4116
[[exploring-history]]
4120
-----------------------------------------------
4121
$ gitk # visualize and browse history
4122
$ git log # list all commits
4123
$ git log src/ # ...modifying src/
4124
$ git log v2.6.15..v2.6.16 # ...in v2.6.16, not in v2.6.15
4125
$ git log master..test # ...in branch test, not in branch master
4126
$ git log test..master # ...in branch master, but not in test
4127
$ git log test...master # ...in one branch, not in both
4128
$ git log -S'foo()' # ...where difference contain "foo()"
4129
$ git log --since="2 weeks ago"
4130
$ git log -p # show patches as well
4131
$ git show # most recent commit
4132
$ git diff v2.6.15..v2.6.16 # diff between two tagged versions
4133
$ git diff v2.6.15..HEAD # diff with current head
4134
$ git grep "foo()" # search working directory for "foo()"
4135
$ git grep v2.6.15 "foo()" # search old tree for "foo()"
4136
$ git show v2.6.15:a.txt # look at old version of a.txt
4137
-----------------------------------------------
4139
Search for regressions:
4141
-----------------------------------------------
4143
$ git bisect bad # current version is bad
4144
$ git bisect good v2.6.13-rc2 # last known good revision
4145
Bisecting: 675 revisions left to test after this
4147
$ git bisect good # if this revision is good, or
4148
$ git bisect bad # if this revision is bad.
4149
# repeat until done.
4150
-----------------------------------------------
4156
Make sure git knows who to blame:
4158
------------------------------------------------
4159
$ cat >>~/.gitconfig <<\EOF
4161
name = Your Name Comes Here
4162
email = you@yourdomain.example.com
4164
------------------------------------------------
4166
Select file contents to include in the next commit, then make the
4169
-----------------------------------------------
4170
$ git add a.txt # updated file
4171
$ git add b.txt # new file
4172
$ git rm c.txt # old file
4174
-----------------------------------------------
4176
Or, prepare and create the commit in one step:
4178
-----------------------------------------------
4179
$ git commit d.txt # use latest content only of d.txt
4180
$ git commit -a # use latest content of all tracked files
4181
-----------------------------------------------
4187
-----------------------------------------------
4188
$ git merge test # merge branch "test" into the current branch
4189
$ git pull git://example.com/project.git master
4190
# fetch and merge in remote branch
4191
$ git pull . test # equivalent to git merge test
4192
-----------------------------------------------
4194
[[sharing-your-changes]]
4195
Sharing your changes
4196
--------------------
4198
Importing or exporting patches:
4200
-----------------------------------------------
4201
$ git format-patch origin..HEAD # format a patch for each commit
4202
# in HEAD but not in origin
4203
$ git am mbox # import patches from the mailbox "mbox"
4204
-----------------------------------------------
4206
Fetch a branch in a different git repository, then merge into the
4209
-----------------------------------------------
4210
$ git pull git://example.com/project.git theirbranch
4211
-----------------------------------------------
4213
Store the fetched branch into a local branch before merging into the
4216
-----------------------------------------------
4217
$ git pull git://example.com/project.git theirbranch:mybranch
4218
-----------------------------------------------
4220
After creating commits on a local branch, update the remote
4221
branch with your commits:
4223
-----------------------------------------------
4224
$ git push ssh://example.com/project.git mybranch:theirbranch
4225
-----------------------------------------------
4227
When remote and local branch are both named "test":
4229
-----------------------------------------------
4230
$ git push ssh://example.com/project.git test
4231
-----------------------------------------------
4233
Shortcut version for a frequently used remote repository:
4235
-----------------------------------------------
4236
$ git remote add example ssh://example.com/project.git
4237
$ git push example test
4238
-----------------------------------------------
4240
[[repository-maintenance]]
4241
Repository maintenance
4242
----------------------
4244
Check for corruption:
4246
-----------------------------------------------
4248
-----------------------------------------------
4250
Recompress, remove unused cruft:
4252
-----------------------------------------------
4254
-----------------------------------------------
4258
Appendix B: Notes and todo list for this manual
4259
===============================================
4261
This is a work in progress.
4263
The basic requirements:
4265
- It must be readable in order, from beginning to end, by someone
4266
intelligent with a basic grasp of the UNIX command line, but without
4267
any special knowledge of git. If necessary, any other prerequisites
4268
should be specifically mentioned as they arise.
4269
- Whenever possible, section headings should clearly describe the task
4270
they explain how to do, in language that requires no more knowledge
4271
than necessary: for example, "importing patches into a project" rather
4272
than "the git-am command"
4274
Think about how to create a clear chapter dependency graph that will
4275
allow people to get to important topics without necessarily reading
4276
everything in between.
4278
Scan Documentation/ for other stuff left out; in particular:
4281
- some of technical/?
4283
- list of commands in gitlink:git[1]
4285
Scan email archives for other stuff left out
4287
Scan man pages to see if any assume more background than this manual
4290
Simplify beginning by suggesting disconnected head instead of
4291
temporary branch creation?
4293
Add more good examples. Entire sections of just cookbook examples
4294
might be a good idea; maybe make an "advanced examples" section a
4295
standard end-of-chapter section?
4297
Include cross-references to the glossary, where appropriate.
4299
Document shallow clones? See draft 1.5.0 release notes for some
4302
Add a section on working with other version control systems, including
4303
CVS, Subversion, and just imports of series of release tarballs.
4305
More details on gitweb?
4307
Write a chapter on using plumbing and writing scripts.
4309
Alternates, clone -reference, etc.
4311
git unpack-objects -r for recovery