Merge branch 'maint' of git://linux-nfs.org/~bfields/git into maint
[git/git.git] / Documentation / user-manual.txt
1 Git User's Manual (for version 1.5.3 or newer)
2 ______________________________________________
3
4
5 Git is a fast distributed revision control system.
6
7 This manual is designed to be readable by someone with basic UNIX
8 command-line skills, but no previous knowledge of git.
9
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.
14
15 People needing to do actual development will also want to read
16 <<Developing-with-git>> and <<sharing-development>>.
17
18 Further chapters cover more specialized topics.
19
20 Comprehensive reference documentation is available through the man
21 pages. For a command such as "git clone", just use
22
23 ------------------------------------------------
24 $ man git-clone
25 ------------------------------------------------
26
27 See also <<git-quick-start>> for a brief overview of git commands,
28 without any explanation.
29
30 Finally, see <<todo>> for ways that you can help make this manual more
31 complete.
32
33
34 [[repositories-and-branches]]
35 Repositories and Branches
36 =========================
37
38 [[how-to-get-a-git-repository]]
39 How to get a git repository
40 ---------------------------
41
42 It will be useful to have a git repository to experiment with as you
43 read this manual.
44
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:
48
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 ------------------------------------------------
55
56 The initial clone may be time-consuming for a large project, but you
57 will only need to clone once.
58
59 The clone command creates a new directory named after the project ("git"
60 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 called the <<def_working_tree,working tree>>, together with a special
63 top-level directory named ".git", which contains all the information
64 about the history of the project.
65
66 [[how-to-check-out]]
67 How to check out a different version of a project
68 -------------------------------------------------
69
70 Git is best thought of as a tool for storing the history of a collection
71 of files. It stores the history as a compressed collection of
72 interrelated snapshots of the project's contents. In git each such
73 version is called a <<def_commit,commit>>.
74
75 Those snapshots aren't necessarily all arranged in a single line from
76 oldest to newest; instead, work may simultaneously proceed along
77 parallel lines of development, called <def_branch,branches>>, which may
78 merge and diverge.
79
80 A single git repository can track development on multiple branches. It
81 does this by keeping a list of <<def_head,heads>> which reference the
82 latest commit on each branch; the gitlink:git-branch[1] command shows
83 you the list of branch heads:
84
85 ------------------------------------------------
86 $ git branch
87 * master
88 ------------------------------------------------
89
90 A freshly cloned repository contains a single branch head, by default
91 named "master", with the working directory initialized to the state of
92 the project referred to by that branch head.
93
94 Most projects also use <<def_tag,tags>>. Tags, like heads, are
95 references into the project's history, and can be listed using the
96 gitlink:git-tag[1] command:
97
98 ------------------------------------------------
99 $ git tag -l
100 v2.6.11
101 v2.6.11-tree
102 v2.6.12
103 v2.6.12-rc2
104 v2.6.12-rc3
105 v2.6.12-rc4
106 v2.6.12-rc5
107 v2.6.12-rc6
108 v2.6.13
109 ...
110 ------------------------------------------------
111
112 Tags are expected to always point at the same version of a project,
113 while heads are expected to advance as development progresses.
114
115 Create a new branch head pointing to one of these versions and check it
116 out using gitlink:git-checkout[1]:
117
118 ------------------------------------------------
119 $ git checkout -b new v2.6.13
120 ------------------------------------------------
121
122 The working directory then reflects the contents that the project had
123 when it was tagged v2.6.13, and gitlink:git-branch[1] shows two
124 branches, with an asterisk marking the currently checked-out branch:
125
126 ------------------------------------------------
127 $ git branch
128 master
129 * new
130 ------------------------------------------------
131
132 If you decide that you'd rather see version 2.6.17, you can modify
133 the current branch to point at v2.6.17 instead, with
134
135 ------------------------------------------------
136 $ git reset --hard v2.6.17
137 ------------------------------------------------
138
139 Note that if the current branch head was your only reference to a
140 particular point in history, then resetting that branch may leave you
141 with no way to find the history it used to point to; so use this command
142 carefully.
143
144 [[understanding-commits]]
145 Understanding History: Commits
146 ------------------------------
147
148 Every change in the history of a project is represented by a commit.
149 The gitlink:git-show[1] command shows the most recent commit on the
150 current branch:
151
152 ------------------------------------------------
153 $ git show
154 commit 17cf781661e6d38f737f15f53ab552f1e95960d7
155 Author: Linus Torvalds <torvalds@ppc970.osdl.org.(none)>
156 Date: Tue Apr 19 14:11:06 2005 -0700
157
158 Remove duplicate getenv(DB_ENVIRONMENT) call
159
160 Noted by Tony Luck.
161
162 diff --git a/init-db.c b/init-db.c
163 index 65898fa..b002dc6 100644
164 --- a/init-db.c
165 +++ b/init-db.c
166 @@ -7,7 +7,7 @@
167
168 int main(int argc, char **argv)
169 {
170 - char *sha1_dir = getenv(DB_ENVIRONMENT), *path;
171 + char *sha1_dir, *path;
172 int len, i;
173
174 if (mkdir(".git", 0755) < 0) {
175 ------------------------------------------------
176
177 As you can see, a commit shows who made the latest change, what they
178 did, and why.
179
180 Every commit has a 40-hexdigit id, sometimes called the "object name" or the
181 "SHA1 id", shown on the first line of the "git show" output. You can usually
182 refer to a commit by a shorter name, such as a tag or a branch name, but this
183 longer name can also be useful. Most importantly, it is a globally unique
184 name for this commit: so if you tell somebody else the object name (for
185 example in email), then you are guaranteed that name will refer to the same
186 commit in their repository that it does in yours (assuming their repository
187 has that commit at all). Since the object name is computed as a hash over the
188 contents of the commit, you are guaranteed that the commit can never change
189 without its name also changing.
190
191 In fact, in <<git-concepts>> we shall see that everything stored in git
192 history, including file data and directory contents, is stored in an object
193 with a name that is a hash of its contents.
194
195 [[understanding-reachability]]
196 Understanding history: commits, parents, and reachability
197 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
198
199 Every commit (except the very first commit in a project) also has a
200 parent commit which shows what happened before this commit.
201 Following the chain of parents will eventually take you back to the
202 beginning of the project.
203
204 However, the commits do not form a simple list; git allows lines of
205 development to diverge and then reconverge, and the point where two
206 lines of development reconverge is called a "merge". The commit
207 representing a merge can therefore have more than one parent, with
208 each parent representing the most recent commit on one of the lines
209 of development leading to that point.
210
211 The best way to see how this works is using the gitlink:gitk[1]
212 command; running gitk now on a git repository and looking for merge
213 commits will help understand how the git organizes history.
214
215 In the following, we say that commit X is "reachable" from commit Y
216 if commit X is an ancestor of commit Y. Equivalently, you could say
217 that Y is a descendant of X, or that there is a chain of parents
218 leading from commit Y to commit X.
219
220 [[history-diagrams]]
221 Understanding history: History diagrams
222 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
223
224 We will sometimes represent git history using diagrams like the one
225 below. Commits are shown as "o", and the links between them with
226 lines drawn with - / and \. Time goes left to right:
227
228
229 ................................................
230 o--o--o <-- Branch A
231 /
232 o--o--o <-- master
233 \
234 o--o--o <-- Branch B
235 ................................................
236
237 If we need to talk about a particular commit, the character "o" may
238 be replaced with another letter or number.
239
240 [[what-is-a-branch]]
241 Understanding history: What is a branch?
242 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
243
244 When we need to be precise, we will use the word "branch" to mean a line
245 of development, and "branch head" (or just "head") to mean a reference
246 to the most recent commit on a branch. In the example above, the branch
247 head named "A" is a pointer to one particular commit, but we refer to
248 the line of three commits leading up to that point as all being part of
249 "branch A".
250
251 However, when no confusion will result, we often just use the term
252 "branch" both for branches and for branch heads.
253
254 [[manipulating-branches]]
255 Manipulating branches
256 ---------------------
257
258 Creating, deleting, and modifying branches is quick and easy; here's
259 a summary of the commands:
260
261 git branch::
262 list all branches
263 git branch <branch>::
264 create a new branch named <branch>, referencing the same
265 point in history as the current branch
266 git branch <branch> <start-point>::
267 create a new branch named <branch>, referencing
268 <start-point>, which may be specified any way you like,
269 including using a branch name or a tag name
270 git branch -d <branch>::
271 delete the branch <branch>; if the branch you are deleting
272 points to a commit which is not reachable from the current
273 branch, this command will fail with a warning.
274 git branch -D <branch>::
275 even if the branch points to a commit not reachable
276 from the current branch, you may know that that commit
277 is still reachable from some other branch or tag. In that
278 case it is safe to use this command to force git to delete
279 the branch.
280 git checkout <branch>::
281 make the current branch <branch>, updating the working
282 directory to reflect the version referenced by <branch>
283 git checkout -b <new> <start-point>::
284 create a new branch <new> referencing <start-point>, and
285 check it out.
286
287 The special symbol "HEAD" can always be used to refer to the current
288 branch. In fact, git uses a file named "HEAD" in the .git directory to
289 remember which branch is current:
290
291 ------------------------------------------------
292 $ cat .git/HEAD
293 ref: refs/heads/master
294 ------------------------------------------------
295
296 [[detached-head]]
297 Examining an old version without creating a new branch
298 ------------------------------------------------------
299
300 The git-checkout command normally expects a branch head, but will also
301 accept an arbitrary commit; for example, you can check out the commit
302 referenced by a tag:
303
304 ------------------------------------------------
305 $ git checkout v2.6.17
306 Note: moving to "v2.6.17" which isn't a local branch
307 If you want to create a new branch from this checkout, you may do so
308 (now or later) by using -b with the checkout command again. Example:
309 git checkout -b <new_branch_name>
310 HEAD is now at 427abfa... Linux v2.6.17
311 ------------------------------------------------
312
313 The HEAD then refers to the SHA1 of the commit instead of to a branch,
314 and git branch shows that you are no longer on a branch:
315
316 ------------------------------------------------
317 $ cat .git/HEAD
318 427abfa28afedffadfca9dd8b067eb6d36bac53f
319 $ git branch
320 * (no branch)
321 master
322 ------------------------------------------------
323
324 In this case we say that the HEAD is "detached".
325
326 This is an easy way to check out a particular version without having to
327 make up a name for the new branch. You can still create a new branch
328 (or tag) for this version later if you decide to.
329
330 [[examining-remote-branches]]
331 Examining branches from a remote repository
332 -------------------------------------------
333
334 The "master" branch that was created at the time you cloned is a copy
335 of the HEAD in the repository that you cloned from. That repository
336 may also have had other branches, though, and your local repository
337 keeps branches which track each of those remote branches, which you
338 can view using the "-r" option to gitlink:git-branch[1]:
339
340 ------------------------------------------------
341 $ git branch -r
342 origin/HEAD
343 origin/html
344 origin/maint
345 origin/man
346 origin/master
347 origin/next
348 origin/pu
349 origin/todo
350 ------------------------------------------------
351
352 You cannot check out these remote-tracking branches, but you can
353 examine them on a branch of your own, just as you would a tag:
354
355 ------------------------------------------------
356 $ git checkout -b my-todo-copy origin/todo
357 ------------------------------------------------
358
359 Note that the name "origin" is just the name that git uses by default
360 to refer to the repository that you cloned from.
361
362 [[how-git-stores-references]]
363 Naming branches, tags, and other references
364 -------------------------------------------
365
366 Branches, remote-tracking branches, and tags are all references to
367 commits. All references are named with a slash-separated path name
368 starting with "refs"; the names we've been using so far are actually
369 shorthand:
370
371 - The branch "test" is short for "refs/heads/test".
372 - The tag "v2.6.18" is short for "refs/tags/v2.6.18".
373 - "origin/master" is short for "refs/remotes/origin/master".
374
375 The full name is occasionally useful if, for example, there ever
376 exists a tag and a branch with the same name.
377
378 (Newly created refs are actually stored in the .git/refs directory,
379 under the path given by their name. However, for efficiency reasons
380 they may also be packed together in a single file; see
381 gitlink:git-pack-refs[1]).
382
383 As another useful shortcut, the "HEAD" of a repository can be referred
384 to just using the name of that repository. So, for example, "origin"
385 is usually a shortcut for the HEAD branch in the repository "origin".
386
387 For the complete list of paths which git checks for references, and
388 the order it uses to decide which to choose when there are multiple
389 references with the same shorthand name, see the "SPECIFYING
390 REVISIONS" section of gitlink:git-rev-parse[1].
391
392 [[Updating-a-repository-with-git-fetch]]
393 Updating a repository with git fetch
394 ------------------------------------
395
396 Eventually the developer cloned from will do additional work in her
397 repository, creating new commits and advancing the branches to point
398 at the new commits.
399
400 The command "git fetch", with no arguments, will update all of the
401 remote-tracking branches to the latest version found in her
402 repository. It will not touch any of your own branches--not even the
403 "master" branch that was created for you on clone.
404
405 [[fetching-branches]]
406 Fetching branches from other repositories
407 -----------------------------------------
408
409 You can also track branches from repositories other than the one you
410 cloned from, using gitlink:git-remote[1]:
411
412 -------------------------------------------------
413 $ git remote add linux-nfs git://linux-nfs.org/pub/nfs-2.6.git
414 $ git fetch linux-nfs
415 * refs/remotes/linux-nfs/master: storing branch 'master' ...
416 commit: bf81b46
417 -------------------------------------------------
418
419 New remote-tracking branches will be stored under the shorthand name
420 that you gave "git remote add", in this case linux-nfs:
421
422 -------------------------------------------------
423 $ git branch -r
424 linux-nfs/master
425 origin/master
426 -------------------------------------------------
427
428 If you run "git fetch <remote>" later, the tracking branches for the
429 named <remote> will be updated.
430
431 If you examine the file .git/config, you will see that git has added
432 a new stanza:
433
434 -------------------------------------------------
435 $ cat .git/config
436 ...
437 [remote "linux-nfs"]
438 url = git://linux-nfs.org/pub/nfs-2.6.git
439 fetch = +refs/heads/*:refs/remotes/linux-nfs/*
440 ...
441 -------------------------------------------------
442
443 This is what causes git to track the remote's branches; you may modify
444 or delete these configuration options by editing .git/config with a
445 text editor. (See the "CONFIGURATION FILE" section of
446 gitlink:git-config[1] for details.)
447
448 [[exploring-git-history]]
449 Exploring git history
450 =====================
451
452 Git is best thought of as a tool for storing the history of a
453 collection of files. It does this by storing compressed snapshots of
454 the contents of a file hierarchy, together with "commits" which show
455 the relationships between these snapshots.
456
457 Git provides extremely flexible and fast tools for exploring the
458 history of a project.
459
460 We start with one specialized tool that is useful for finding the
461 commit that introduced a bug into a project.
462
463 [[using-bisect]]
464 How to use bisect to find a regression
465 --------------------------------------
466
467 Suppose version 2.6.18 of your project worked, but the version at
468 "master" crashes. Sometimes the best way to find the cause of such a
469 regression is to perform a brute-force search through the project's
470 history to find the particular commit that caused the problem. The
471 gitlink:git-bisect[1] command can help you do this:
472
473 -------------------------------------------------
474 $ git bisect start
475 $ git bisect good v2.6.18
476 $ git bisect bad master
477 Bisecting: 3537 revisions left to test after this
478 [65934a9a028b88e83e2b0f8b36618fe503349f8e] BLOCK: Make USB storage depend on SCSI rather than selecting it [try #6]
479 -------------------------------------------------
480
481 If you run "git branch" at this point, you'll see that git has
482 temporarily moved you to a new branch named "bisect". This branch
483 points to a commit (with commit id 65934...) that is reachable from
484 v2.6.19 but not from v2.6.18. Compile and test it, and see whether
485 it crashes. Assume it does crash. Then:
486
487 -------------------------------------------------
488 $ git bisect bad
489 Bisecting: 1769 revisions left to test after this
490 [7eff82c8b1511017ae605f0c99ac275a7e21b867] i2c-core: Drop useless bitmaskings
491 -------------------------------------------------
492
493 checks out an older version. Continue like this, telling git at each
494 stage whether the version it gives you is good or bad, and notice
495 that the number of revisions left to test is cut approximately in
496 half each time.
497
498 After about 13 tests (in this case), it will output the commit id of
499 the guilty commit. You can then examine the commit with
500 gitlink:git-show[1], find out who wrote it, and mail them your bug
501 report with the commit id. Finally, run
502
503 -------------------------------------------------
504 $ git bisect reset
505 -------------------------------------------------
506
507 to return you to the branch you were on before and delete the
508 temporary "bisect" branch.
509
510 Note that the version which git-bisect checks out for you at each
511 point is just a suggestion, and you're free to try a different
512 version if you think it would be a good idea. For example,
513 occasionally you may land on a commit that broke something unrelated;
514 run
515
516 -------------------------------------------------
517 $ git bisect visualize
518 -------------------------------------------------
519
520 which will run gitk and label the commit it chose with a marker that
521 says "bisect". Chose a safe-looking commit nearby, note its commit
522 id, and check it out with:
523
524 -------------------------------------------------
525 $ git reset --hard fb47ddb2db...
526 -------------------------------------------------
527
528 then test, run "bisect good" or "bisect bad" as appropriate, and
529 continue.
530
531 [[naming-commits]]
532 Naming commits
533 --------------
534
535 We have seen several ways of naming commits already:
536
537 - 40-hexdigit object name
538 - branch name: refers to the commit at the head of the given
539 branch
540 - tag name: refers to the commit pointed to by the given tag
541 (we've seen branches and tags are special cases of
542 <<how-git-stores-references,references>>).
543 - HEAD: refers to the head of the current branch
544
545 There are many more; see the "SPECIFYING REVISIONS" section of the
546 gitlink:git-rev-parse[1] man page for the complete list of ways to
547 name revisions. Some examples:
548
549 -------------------------------------------------
550 $ git show fb47ddb2 # the first few characters of the object name
551 # are usually enough to specify it uniquely
552 $ git show HEAD^ # the parent of the HEAD commit
553 $ git show HEAD^^ # the grandparent
554 $ git show HEAD~4 # the great-great-grandparent
555 -------------------------------------------------
556
557 Recall that merge commits may have more than one parent; by default,
558 ^ and ~ follow the first parent listed in the commit, but you can
559 also choose:
560
561 -------------------------------------------------
562 $ git show HEAD^1 # show the first parent of HEAD
563 $ git show HEAD^2 # show the second parent of HEAD
564 -------------------------------------------------
565
566 In addition to HEAD, there are several other special names for
567 commits:
568
569 Merges (to be discussed later), as well as operations such as
570 git-reset, which change the currently checked-out commit, generally
571 set ORIG_HEAD to the value HEAD had before the current operation.
572
573 The git-fetch operation always stores the head of the last fetched
574 branch in FETCH_HEAD. For example, if you run git fetch without
575 specifying a local branch as the target of the operation
576
577 -------------------------------------------------
578 $ git fetch git://example.com/proj.git theirbranch
579 -------------------------------------------------
580
581 the fetched commits will still be available from FETCH_HEAD.
582
583 When we discuss merges we'll also see the special name MERGE_HEAD,
584 which refers to the other branch that we're merging in to the current
585 branch.
586
587 The gitlink:git-rev-parse[1] command is a low-level command that is
588 occasionally useful for translating some name for a commit to the object
589 name for that commit:
590
591 -------------------------------------------------
592 $ git rev-parse origin
593 e05db0fd4f31dde7005f075a84f96b360d05984b
594 -------------------------------------------------
595
596 [[creating-tags]]
597 Creating tags
598 -------------
599
600 We can also create a tag to refer to a particular commit; after
601 running
602
603 -------------------------------------------------
604 $ git tag stable-1 1b2e1d63ff
605 -------------------------------------------------
606
607 You can use stable-1 to refer to the commit 1b2e1d63ff.
608
609 This creates a "lightweight" tag. If you would also like to include a
610 comment with the tag, and possibly sign it cryptographically, then you
611 should create a tag object instead; see the gitlink:git-tag[1] man page
612 for details.
613
614 [[browsing-revisions]]
615 Browsing revisions
616 ------------------
617
618 The gitlink:git-log[1] command can show lists of commits. On its
619 own, it shows all commits reachable from the parent commit; but you
620 can also make more specific requests:
621
622 -------------------------------------------------
623 $ git log v2.5.. # commits since (not reachable from) v2.5
624 $ git log test..master # commits reachable from master but not test
625 $ git log master..test # ...reachable from test but not master
626 $ git log master...test # ...reachable from either test or master,
627 # but not both
628 $ git log --since="2 weeks ago" # commits from the last 2 weeks
629 $ git log Makefile # commits which modify Makefile
630 $ git log fs/ # ... which modify any file under fs/
631 $ git log -S'foo()' # commits which add or remove any file data
632 # matching the string 'foo()'
633 -------------------------------------------------
634
635 And of course you can combine all of these; the following finds
636 commits since v2.5 which touch the Makefile or any file under fs:
637
638 -------------------------------------------------
639 $ git log v2.5.. Makefile fs/
640 -------------------------------------------------
641
642 You can also ask git log to show patches:
643
644 -------------------------------------------------
645 $ git log -p
646 -------------------------------------------------
647
648 See the "--pretty" option in the gitlink:git-log[1] man page for more
649 display options.
650
651 Note that git log starts with the most recent commit and works
652 backwards through the parents; however, since git history can contain
653 multiple independent lines of development, the particular order that
654 commits are listed in may be somewhat arbitrary.
655
656 [[generating-diffs]]
657 Generating diffs
658 ----------------
659
660 You can generate diffs between any two versions using
661 gitlink:git-diff[1]:
662
663 -------------------------------------------------
664 $ git diff master..test
665 -------------------------------------------------
666
667 That will produce the diff between the tips of the two branches. If
668 you'd prefer to find the diff from their common ancestor to test, you
669 can use three dots instead of two:
670
671 -------------------------------------------------
672 $ git diff master...test
673 -------------------------------------------------
674
675 Sometimes what you want instead is a set of patches; for this you can
676 use gitlink:git-format-patch[1]:
677
678 -------------------------------------------------
679 $ git format-patch master..test
680 -------------------------------------------------
681
682 will generate a file with a patch for each commit reachable from test
683 but not from master.
684
685 [[viewing-old-file-versions]]
686 Viewing old file versions
687 -------------------------
688
689 You can always view an old version of a file by just checking out the
690 correct revision first. But sometimes it is more convenient to be
691 able to view an old version of a single file without checking
692 anything out; this command does that:
693
694 -------------------------------------------------
695 $ git show v2.5:fs/locks.c
696 -------------------------------------------------
697
698 Before the colon may be anything that names a commit, and after it
699 may be any path to a file tracked by git.
700
701 [[history-examples]]
702 Examples
703 --------
704
705 [[counting-commits-on-a-branch]]
706 Counting the number of commits on a branch
707 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
708
709 Suppose you want to know how many commits you've made on "mybranch"
710 since it diverged from "origin":
711
712 -------------------------------------------------
713 $ git log --pretty=oneline origin..mybranch | wc -l
714 -------------------------------------------------
715
716 Alternatively, you may often see this sort of thing done with the
717 lower-level command gitlink:git-rev-list[1], which just lists the SHA1's
718 of all the given commits:
719
720 -------------------------------------------------
721 $ git rev-list origin..mybranch | wc -l
722 -------------------------------------------------
723
724 [[checking-for-equal-branches]]
725 Check whether two branches point at the same history
726 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
727
728 Suppose you want to check whether two branches point at the same point
729 in history.
730
731 -------------------------------------------------
732 $ git diff origin..master
733 -------------------------------------------------
734
735 will tell you whether the contents of the project are the same at the
736 two branches; in theory, however, it's possible that the same project
737 contents could have been arrived at by two different historical
738 routes. You could compare the object names:
739
740 -------------------------------------------------
741 $ git rev-list origin
742 e05db0fd4f31dde7005f075a84f96b360d05984b
743 $ git rev-list master
744 e05db0fd4f31dde7005f075a84f96b360d05984b
745 -------------------------------------------------
746
747 Or you could recall that the ... operator selects all commits
748 contained reachable from either one reference or the other but not
749 both: so
750
751 -------------------------------------------------
752 $ git log origin...master
753 -------------------------------------------------
754
755 will return no commits when the two branches are equal.
756
757 [[finding-tagged-descendants]]
758 Find first tagged version including a given fix
759 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
760
761 Suppose you know that the commit e05db0fd fixed a certain problem.
762 You'd like to find the earliest tagged release that contains that
763 fix.
764
765 Of course, there may be more than one answer--if the history branched
766 after commit e05db0fd, then there could be multiple "earliest" tagged
767 releases.
768
769 You could just visually inspect the commits since e05db0fd:
770
771 -------------------------------------------------
772 $ gitk e05db0fd..
773 -------------------------------------------------
774
775 Or you can use gitlink:git-name-rev[1], which will give the commit a
776 name based on any tag it finds pointing to one of the commit's
777 descendants:
778
779 -------------------------------------------------
780 $ git name-rev --tags e05db0fd
781 e05db0fd tags/v1.5.0-rc1^0~23
782 -------------------------------------------------
783
784 The gitlink:git-describe[1] command does the opposite, naming the
785 revision using a tag on which the given commit is based:
786
787 -------------------------------------------------
788 $ git describe e05db0fd
789 v1.5.0-rc0-260-ge05db0f
790 -------------------------------------------------
791
792 but that may sometimes help you guess which tags might come after the
793 given commit.
794
795 If you just want to verify whether a given tagged version contains a
796 given commit, you could use gitlink:git-merge-base[1]:
797
798 -------------------------------------------------
799 $ git merge-base e05db0fd v1.5.0-rc1
800 e05db0fd4f31dde7005f075a84f96b360d05984b
801 -------------------------------------------------
802
803 The merge-base command finds a common ancestor of the given commits,
804 and always returns one or the other in the case where one is a
805 descendant of the other; so the above output shows that e05db0fd
806 actually is an ancestor of v1.5.0-rc1.
807
808 Alternatively, note that
809
810 -------------------------------------------------
811 $ git log v1.5.0-rc1..e05db0fd
812 -------------------------------------------------
813
814 will produce empty output if and only if v1.5.0-rc1 includes e05db0fd,
815 because it outputs only commits that are not reachable from v1.5.0-rc1.
816
817 As yet another alternative, the gitlink:git-show-branch[1] command lists
818 the commits reachable from its arguments with a display on the left-hand
819 side that indicates which arguments that commit is reachable from. So,
820 you can run something like
821
822 -------------------------------------------------
823 $ git show-branch e05db0fd v1.5.0-rc0 v1.5.0-rc1 v1.5.0-rc2
824 ! [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
825 available
826 ! [v1.5.0-rc0] GIT v1.5.0 preview
827 ! [v1.5.0-rc1] GIT v1.5.0-rc1
828 ! [v1.5.0-rc2] GIT v1.5.0-rc2
829 ...
830 -------------------------------------------------
831
832 then search for a line that looks like
833
834 -------------------------------------------------
835 + ++ [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
836 available
837 -------------------------------------------------
838
839 Which shows that e05db0fd is reachable from itself, from v1.5.0-rc1, and
840 from v1.5.0-rc2, but not from v1.5.0-rc0.
841
842 [[showing-commits-unique-to-a-branch]]
843 Showing commits unique to a given branch
844 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
845
846 Suppose you would like to see all the commits reachable from the branch
847 head named "master" but not from any other head in your repository.
848
849 We can list all the heads in this repository with
850 gitlink:git-show-ref[1]:
851
852 -------------------------------------------------
853 $ git show-ref --heads
854 bf62196b5e363d73353a9dcf094c59595f3153b7 refs/heads/core-tutorial
855 db768d5504c1bb46f63ee9d6e1772bd047e05bf9 refs/heads/maint
856 a07157ac624b2524a059a3414e99f6f44bebc1e7 refs/heads/master
857 24dbc180ea14dc1aebe09f14c8ecf32010690627 refs/heads/tutorial-2
858 1e87486ae06626c2f31eaa63d26fc0fd646c8af2 refs/heads/tutorial-fixes
859 -------------------------------------------------
860
861 We can get just the branch-head names, and remove "master", with
862 the help of the standard utilities cut and grep:
863
864 -------------------------------------------------
865 $ git show-ref --heads | cut -d' ' -f2 | grep -v '^refs/heads/master'
866 refs/heads/core-tutorial
867 refs/heads/maint
868 refs/heads/tutorial-2
869 refs/heads/tutorial-fixes
870 -------------------------------------------------
871
872 And then we can ask to see all the commits reachable from master
873 but not from these other heads:
874
875 -------------------------------------------------
876 $ gitk master --not $( git show-ref --heads | cut -d' ' -f2 |
877 grep -v '^refs/heads/master' )
878 -------------------------------------------------
879
880 Obviously, endless variations are possible; for example, to see all
881 commits reachable from some head but not from any tag in the repository:
882
883 -------------------------------------------------
884 $ gitk $( git show-ref --heads ) --not $( git show-ref --tags )
885 -------------------------------------------------
886
887 (See gitlink:git-rev-parse[1] for explanations of commit-selecting
888 syntax such as `--not`.)
889
890 [[making-a-release]]
891 Creating a changelog and tarball for a software release
892 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
893
894 The gitlink:git-archive[1] command can create a tar or zip archive from
895 any version of a project; for example:
896
897 -------------------------------------------------
898 $ git archive --format=tar --prefix=project/ HEAD | gzip >latest.tar.gz
899 -------------------------------------------------
900
901 will use HEAD to produce a tar archive in which each filename is
902 preceded by "project/".
903
904 If you're releasing a new version of a software project, you may want
905 to simultaneously make a changelog to include in the release
906 announcement.
907
908 Linus Torvalds, for example, makes new kernel releases by tagging them,
909 then running:
910
911 -------------------------------------------------
912 $ release-script 2.6.12 2.6.13-rc6 2.6.13-rc7
913 -------------------------------------------------
914
915 where release-script is a shell script that looks like:
916
917 -------------------------------------------------
918 #!/bin/sh
919 stable="$1"
920 last="$2"
921 new="$3"
922 echo "# git tag v$new"
923 echo "git archive --prefix=linux-$new/ v$new | gzip -9 > ../linux-$new.tar.gz"
924 echo "git diff v$stable v$new | gzip -9 > ../patch-$new.gz"
925 echo "git log --no-merges v$new ^v$last > ../ChangeLog-$new"
926 echo "git shortlog --no-merges v$new ^v$last > ../ShortLog"
927 echo "git diff --stat --summary -M v$last v$new > ../diffstat-$new"
928 -------------------------------------------------
929
930 and then he just cut-and-pastes the output commands after verifying that
931 they look OK.
932
933 [[Finding-comments-with-given-content]]
934 Finding commits referencing a file with given content
935 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
936
937 Somebody hands you a copy of a file, and asks which commits modified a
938 file such that it contained the given content either before or after the
939 commit. You can find out with this:
940
941 -------------------------------------------------
942 $ git log --raw --abbrev=40 --pretty=oneline -- filename |
943 grep -B 1 `git hash-object filename`
944 -------------------------------------------------
945
946 Figuring out why this works is left as an exercise to the (advanced)
947 student. The gitlink:git-log[1], gitlink:git-diff-tree[1], and
948 gitlink:git-hash-object[1] man pages may prove helpful.
949
950 [[Developing-with-git]]
951 Developing with git
952 ===================
953
954 [[telling-git-your-name]]
955 Telling git your name
956 ---------------------
957
958 Before creating any commits, you should introduce yourself to git. The
959 easiest way to do so is to make sure the following lines appear in a
960 file named .gitconfig in your home directory:
961
962 ------------------------------------------------
963 [user]
964 name = Your Name Comes Here
965 email = you@yourdomain.example.com
966 ------------------------------------------------
967
968 (See the "CONFIGURATION FILE" section of gitlink:git-config[1] for
969 details on the configuration file.)
970
971
972 [[creating-a-new-repository]]
973 Creating a new repository
974 -------------------------
975
976 Creating a new repository from scratch is very easy:
977
978 -------------------------------------------------
979 $ mkdir project
980 $ cd project
981 $ git init
982 -------------------------------------------------
983
984 If you have some initial content (say, a tarball):
985
986 -------------------------------------------------
987 $ tar -xzvf project.tar.gz
988 $ cd project
989 $ git init
990 $ git add . # include everything below ./ in the first commit:
991 $ git commit
992 -------------------------------------------------
993
994 [[how-to-make-a-commit]]
995 How to make a commit
996 --------------------
997
998 Creating a new commit takes three steps:
999
1000 1. Making some changes to the working directory using your
1001 favorite editor.
1002 2. Telling git about your changes.
1003 3. Creating the commit using the content you told git about
1004 in step 2.
1005
1006 In practice, you can interleave and repeat steps 1 and 2 as many
1007 times as you want: in order to keep track of what you want committed
1008 at step 3, git maintains a snapshot of the tree's contents in a
1009 special staging area called "the index."
1010
1011 At the beginning, the content of the index will be identical to
1012 that of the HEAD. The command "git diff --cached", which shows
1013 the difference between the HEAD and the index, should therefore
1014 produce no output at that point.
1015
1016 Modifying the index is easy:
1017
1018 To update the index with the new contents of a modified file, use
1019
1020 -------------------------------------------------
1021 $ git add path/to/file
1022 -------------------------------------------------
1023
1024 To add the contents of a new file to the index, use
1025
1026 -------------------------------------------------
1027 $ git add path/to/file
1028 -------------------------------------------------
1029
1030 To remove a file from the index and from the working tree,
1031
1032 -------------------------------------------------
1033 $ git rm path/to/file
1034 -------------------------------------------------
1035
1036 After each step you can verify that
1037
1038 -------------------------------------------------
1039 $ git diff --cached
1040 -------------------------------------------------
1041
1042 always shows the difference between the HEAD and the index file--this
1043 is what you'd commit if you created the commit now--and that
1044
1045 -------------------------------------------------
1046 $ git diff
1047 -------------------------------------------------
1048
1049 shows the difference between the working tree and the index file.
1050
1051 Note that "git add" always adds just the current contents of a file
1052 to the index; further changes to the same file will be ignored unless
1053 you run git-add on the file again.
1054
1055 When you're ready, just run
1056
1057 -------------------------------------------------
1058 $ git commit
1059 -------------------------------------------------
1060
1061 and git will prompt you for a commit message and then create the new
1062 commit. Check to make sure it looks like what you expected with
1063
1064 -------------------------------------------------
1065 $ git show
1066 -------------------------------------------------
1067
1068 As a special shortcut,
1069
1070 -------------------------------------------------
1071 $ git commit -a
1072 -------------------------------------------------
1073
1074 will update the index with any files that you've modified or removed
1075 and create a commit, all in one step.
1076
1077 A number of commands are useful for keeping track of what you're
1078 about to commit:
1079
1080 -------------------------------------------------
1081 $ git diff --cached # difference between HEAD and the index; what
1082 # would be committed if you ran "commit" now.
1083 $ git diff # difference between the index file and your
1084 # working directory; changes that would not
1085 # be included if you ran "commit" now.
1086 $ git diff HEAD # difference between HEAD and working tree; what
1087 # would be committed if you ran "commit -a" now.
1088 $ git status # a brief per-file summary of the above.
1089 -------------------------------------------------
1090
1091 You can also use gitlink:git-gui[1] to create commits, view changes in
1092 the index and the working tree files, and individually select diff hunks
1093 for inclusion in the index (by right-clicking on the diff hunk and
1094 choosing "Stage Hunk For Commit").
1095
1096 [[creating-good-commit-messages]]
1097 Creating good commit messages
1098 -----------------------------
1099
1100 Though not required, it's a good idea to begin the commit message
1101 with a single short (less than 50 character) line summarizing the
1102 change, followed by a blank line and then a more thorough
1103 description. Tools that turn commits into email, for example, use
1104 the first line on the Subject line and the rest of the commit in the
1105 body.
1106
1107 [[ignoring-files]]
1108 Ignoring files
1109 --------------
1110
1111 A project will often generate files that you do 'not' want to track with git.
1112 This typically includes files generated by a build process or temporary
1113 backup files made by your editor. Of course, 'not' tracking files with git
1114 is just a matter of 'not' calling "`git add`" on them. But it quickly becomes
1115 annoying to have these untracked files lying around; e.g. they make
1116 "`git add .`" and "`git commit -a`" practically useless, and they keep
1117 showing up in the output of "`git status`".
1118
1119 You can tell git to ignore certain files by creating a file called .gitignore
1120 in the top level of your working directory, with contents such as:
1121
1122 -------------------------------------------------
1123 # Lines starting with '#' are considered comments.
1124 # Ignore any file named foo.txt.
1125 foo.txt
1126 # Ignore (generated) html files,
1127 *.html
1128 # except foo.html which is maintained by hand.
1129 !foo.html
1130 # Ignore objects and archives.
1131 *.[oa]
1132 -------------------------------------------------
1133
1134 See gitlink:gitignore[5] for a detailed explanation of the syntax. You can
1135 also place .gitignore files in other directories in your working tree, and they
1136 will apply to those directories and their subdirectories. The `.gitignore`
1137 files can be added to your repository like any other files (just run `git add
1138 .gitignore` and `git commit`, as usual), which is convenient when the exclude
1139 patterns (such as patterns matching build output files) would also make sense
1140 for other users who clone your repository.
1141
1142 If you wish the exclude patterns to affect only certain repositories
1143 (instead of every repository for a given project), you may instead put
1144 them in a file in your repository named .git/info/exclude, or in any file
1145 specified by the `core.excludesfile` configuration variable. Some git
1146 commands can also take exclude patterns directly on the command line.
1147 See gitlink:gitignore[5] for the details.
1148
1149 [[how-to-merge]]
1150 How to merge
1151 ------------
1152
1153 You can rejoin two diverging branches of development using
1154 gitlink:git-merge[1]:
1155
1156 -------------------------------------------------
1157 $ git merge branchname
1158 -------------------------------------------------
1159
1160 merges the development in the branch "branchname" into the current
1161 branch. If there are conflicts--for example, if the same file is
1162 modified in two different ways in the remote branch and the local
1163 branch--then you are warned; the output may look something like this:
1164
1165 -------------------------------------------------
1166 $ git merge next
1167 100% (4/4) done
1168 Auto-merged file.txt
1169 CONFLICT (content): Merge conflict in file.txt
1170 Automatic merge failed; fix conflicts and then commit the result.
1171 -------------------------------------------------
1172
1173 Conflict markers are left in the problematic files, and after
1174 you resolve the conflicts manually, you can update the index
1175 with the contents and run git commit, as you normally would when
1176 creating a new file.
1177
1178 If you examine the resulting commit using gitk, you will see that it
1179 has two parents, one pointing to the top of the current branch, and
1180 one to the top of the other branch.
1181
1182 [[resolving-a-merge]]
1183 Resolving a merge
1184 -----------------
1185
1186 When a merge isn't resolved automatically, git leaves the index and
1187 the working tree in a special state that gives you all the
1188 information you need to help resolve the merge.
1189
1190 Files with conflicts are marked specially in the index, so until you
1191 resolve the problem and update the index, gitlink:git-commit[1] will
1192 fail:
1193
1194 -------------------------------------------------
1195 $ git commit
1196 file.txt: needs merge
1197 -------------------------------------------------
1198
1199 Also, gitlink:git-status[1] will list those files as "unmerged", and the
1200 files with conflicts will have conflict markers added, like this:
1201
1202 -------------------------------------------------
1203 <<<<<<< HEAD:file.txt
1204 Hello world
1205 =======
1206 Goodbye
1207 >>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1208 -------------------------------------------------
1209
1210 All you need to do is edit the files to resolve the conflicts, and then
1211
1212 -------------------------------------------------
1213 $ git add file.txt
1214 $ git commit
1215 -------------------------------------------------
1216
1217 Note that the commit message will already be filled in for you with
1218 some information about the merge. Normally you can just use this
1219 default message unchanged, but you may add additional commentary of
1220 your own if desired.
1221
1222 The above is all you need to know to resolve a simple merge. But git
1223 also provides more information to help resolve conflicts:
1224
1225 [[conflict-resolution]]
1226 Getting conflict-resolution help during a merge
1227 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1228
1229 All of the changes that git was able to merge automatically are
1230 already added to the index file, so gitlink:git-diff[1] shows only
1231 the conflicts. It uses an unusual syntax:
1232
1233 -------------------------------------------------
1234 $ git diff
1235 diff --cc file.txt
1236 index 802992c,2b60207..0000000
1237 --- a/file.txt
1238 +++ b/file.txt
1239 @@@ -1,1 -1,1 +1,5 @@@
1240 ++<<<<<<< HEAD:file.txt
1241 +Hello world
1242 ++=======
1243 + Goodbye
1244 ++>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1245 -------------------------------------------------
1246
1247 Recall that the commit which will be committed after we resolve this
1248 conflict will have two parents instead of the usual one: one parent
1249 will be HEAD, the tip of the current branch; the other will be the
1250 tip of the other branch, which is stored temporarily in MERGE_HEAD.
1251
1252 During the merge, the index holds three versions of each file. Each of
1253 these three "file stages" represents a different version of the file:
1254
1255 -------------------------------------------------
1256 $ git show :1:file.txt # the file in a common ancestor of both branches
1257 $ git show :2:file.txt # the version from HEAD, but including any
1258 # nonconflicting changes from MERGE_HEAD
1259 $ git show :3:file.txt # the version from MERGE_HEAD, but including any
1260 # nonconflicting changes from HEAD.
1261 -------------------------------------------------
1262
1263 Since the stage 2 and stage 3 versions have already been updated with
1264 nonconflicting changes, the only remaining differences between them are
1265 the important ones; thus gitlink:git-diff[1] can use the information in
1266 the index to show only those conflicts.
1267
1268 The diff above shows the differences between the working-tree version of
1269 file.txt and the stage 2 and stage 3 versions. So instead of preceding
1270 each line by a single "+" or "-", it now uses two columns: the first
1271 column is used for differences between the first parent and the working
1272 directory copy, and the second for differences between the second parent
1273 and the working directory copy. (See the "COMBINED DIFF FORMAT" section
1274 of gitlink:git-diff-files[1] for a details of the format.)
1275
1276 After resolving the conflict in the obvious way (but before updating the
1277 index), the diff will look like:
1278
1279 -------------------------------------------------
1280 $ git diff
1281 diff --cc file.txt
1282 index 802992c,2b60207..0000000
1283 --- a/file.txt
1284 +++ b/file.txt
1285 @@@ -1,1 -1,1 +1,1 @@@
1286 - Hello world
1287 -Goodbye
1288 ++Goodbye world
1289 -------------------------------------------------
1290
1291 This shows that our resolved version deleted "Hello world" from the
1292 first parent, deleted "Goodbye" from the second parent, and added
1293 "Goodbye world", which was previously absent from both.
1294
1295 Some special diff options allow diffing the working directory against
1296 any of these stages:
1297
1298 -------------------------------------------------
1299 $ git diff -1 file.txt # diff against stage 1
1300 $ git diff --base file.txt # same as the above
1301 $ git diff -2 file.txt # diff against stage 2
1302 $ git diff --ours file.txt # same as the above
1303 $ git diff -3 file.txt # diff against stage 3
1304 $ git diff --theirs file.txt # same as the above.
1305 -------------------------------------------------
1306
1307 The gitlink:git-log[1] and gitk[1] commands also provide special help
1308 for merges:
1309
1310 -------------------------------------------------
1311 $ git log --merge
1312 $ gitk --merge
1313 -------------------------------------------------
1314
1315 These will display all commits which exist only on HEAD or on
1316 MERGE_HEAD, and which touch an unmerged file.
1317
1318 You may also use gitlink:git-mergetool[1], which lets you merge the
1319 unmerged files using external tools such as emacs or kdiff3.
1320
1321 Each time you resolve the conflicts in a file and update the index:
1322
1323 -------------------------------------------------
1324 $ git add file.txt
1325 -------------------------------------------------
1326
1327 the different stages of that file will be "collapsed", after which
1328 git-diff will (by default) no longer show diffs for that file.
1329
1330 [[undoing-a-merge]]
1331 Undoing a merge
1332 ---------------
1333
1334 If you get stuck and decide to just give up and throw the whole mess
1335 away, you can always return to the pre-merge state with
1336
1337 -------------------------------------------------
1338 $ git reset --hard HEAD
1339 -------------------------------------------------
1340
1341 Or, if you've already committed the merge that you want to throw away,
1342
1343 -------------------------------------------------
1344 $ git reset --hard ORIG_HEAD
1345 -------------------------------------------------
1346
1347 However, this last command can be dangerous in some cases--never
1348 throw away a commit you have already committed if that commit may
1349 itself have been merged into another branch, as doing so may confuse
1350 further merges.
1351
1352 [[fast-forwards]]
1353 Fast-forward merges
1354 -------------------
1355
1356 There is one special case not mentioned above, which is treated
1357 differently. Normally, a merge results in a merge commit, with two
1358 parents, one pointing at each of the two lines of development that
1359 were merged.
1360
1361 However, if the current branch is a descendant of the other--so every
1362 commit present in the one is already contained in the other--then git
1363 just performs a "fast forward"; the head of the current branch is moved
1364 forward to point at the head of the merged-in branch, without any new
1365 commits being created.
1366
1367 [[fixing-mistakes]]
1368 Fixing mistakes
1369 ---------------
1370
1371 If you've messed up the working tree, but haven't yet committed your
1372 mistake, you can return the entire working tree to the last committed
1373 state with
1374
1375 -------------------------------------------------
1376 $ git reset --hard HEAD
1377 -------------------------------------------------
1378
1379 If you make a commit that you later wish you hadn't, there are two
1380 fundamentally different ways to fix the problem:
1381
1382 1. You can create a new commit that undoes whatever was done
1383 by the previous commit. This is the correct thing if your
1384 mistake has already been made public.
1385
1386 2. You can go back and modify the old commit. You should
1387 never do this if you have already made the history public;
1388 git does not normally expect the "history" of a project to
1389 change, and cannot correctly perform repeated merges from
1390 a branch that has had its history changed.
1391
1392 [[reverting-a-commit]]
1393 Fixing a mistake with a new commit
1394 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1395
1396 Creating a new commit that reverts an earlier change is very easy;
1397 just pass the gitlink:git-revert[1] command a reference to the bad
1398 commit; for example, to revert the most recent commit:
1399
1400 -------------------------------------------------
1401 $ git revert HEAD
1402 -------------------------------------------------
1403
1404 This will create a new commit which undoes the change in HEAD. You
1405 will be given a chance to edit the commit message for the new commit.
1406
1407 You can also revert an earlier change, for example, the next-to-last:
1408
1409 -------------------------------------------------
1410 $ git revert HEAD^
1411 -------------------------------------------------
1412
1413 In this case git will attempt to undo the old change while leaving
1414 intact any changes made since then. If more recent changes overlap
1415 with the changes to be reverted, then you will be asked to fix
1416 conflicts manually, just as in the case of <<resolving-a-merge,
1417 resolving a merge>>.
1418
1419 [[fixing-a-mistake-by-rewriting-history]]
1420 Fixing a mistake by rewriting history
1421 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1422
1423 If the problematic commit is the most recent commit, and you have not
1424 yet made that commit public, then you may just
1425 <<undoing-a-merge,destroy it using git-reset>>.
1426
1427 Alternatively, you
1428 can edit the working directory and update the index to fix your
1429 mistake, just as if you were going to <<how-to-make-a-commit,create a
1430 new commit>>, then run
1431
1432 -------------------------------------------------
1433 $ git commit --amend
1434 -------------------------------------------------
1435
1436 which will replace the old commit by a new commit incorporating your
1437 changes, giving you a chance to edit the old commit message first.
1438
1439 Again, you should never do this to a commit that may already have
1440 been merged into another branch; use gitlink:git-revert[1] instead in
1441 that case.
1442
1443 It is also possible to replace commits further back in the history, but
1444 this is an advanced topic to be left for
1445 <<cleaning-up-history,another chapter>>.
1446
1447 [[checkout-of-path]]
1448 Checking out an old version of a file
1449 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1450
1451 In the process of undoing a previous bad change, you may find it
1452 useful to check out an older version of a particular file using
1453 gitlink:git-checkout[1]. We've used git checkout before to switch
1454 branches, but it has quite different behavior if it is given a path
1455 name: the command
1456
1457 -------------------------------------------------
1458 $ git checkout HEAD^ path/to/file
1459 -------------------------------------------------
1460
1461 replaces path/to/file by the contents it had in the commit HEAD^, and
1462 also updates the index to match. It does not change branches.
1463
1464 If you just want to look at an old version of the file, without
1465 modifying the working directory, you can do that with
1466 gitlink:git-show[1]:
1467
1468 -------------------------------------------------
1469 $ git show HEAD^:path/to/file
1470 -------------------------------------------------
1471
1472 which will display the given version of the file.
1473
1474 [[interrupted-work]]
1475 Temporarily setting aside work in progress
1476 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1477
1478 While you are in the middle of working on something complicated, you
1479 find an unrelated but obvious and trivial bug. You would like to fix it
1480 before continuing. You can use gitlink:git-stash[1] to save the current
1481 state of your work, and after fixing the bug (or, optionally after doing
1482 so on a different branch and then coming back), unstash the
1483 work-in-progress changes.
1484
1485 ------------------------------------------------
1486 $ git stash "work in progress for foo feature"
1487 ------------------------------------------------
1488
1489 This command will save your changes away to the `stash`, and
1490 reset your working tree and the index to match the tip of your
1491 current branch. Then you can make your fix as usual.
1492
1493 ------------------------------------------------
1494 ... edit and test ...
1495 $ git commit -a -m "blorpl: typofix"
1496 ------------------------------------------------
1497
1498 After that, you can go back to what you were working on with
1499 `git stash apply`:
1500
1501 ------------------------------------------------
1502 $ git stash apply
1503 ------------------------------------------------
1504
1505
1506 [[ensuring-good-performance]]
1507 Ensuring good performance
1508 -------------------------
1509
1510 On large repositories, git depends on compression to keep the history
1511 information from taking up to much space on disk or in memory.
1512
1513 This compression is not performed automatically. Therefore you
1514 should occasionally run gitlink:git-gc[1]:
1515
1516 -------------------------------------------------
1517 $ git gc
1518 -------------------------------------------------
1519
1520 to recompress the archive. This can be very time-consuming, so
1521 you may prefer to run git-gc when you are not doing other work.
1522
1523
1524 [[ensuring-reliability]]
1525 Ensuring reliability
1526 --------------------
1527
1528 [[checking-for-corruption]]
1529 Checking the repository for corruption
1530 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1531
1532 The gitlink:git-fsck[1] command runs a number of self-consistency checks
1533 on the repository, and reports on any problems. This may take some
1534 time. The most common warning by far is about "dangling" objects:
1535
1536 -------------------------------------------------
1537 $ git fsck
1538 dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1539 dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1540 dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1541 dangling blob 218761f9d90712d37a9c5e36f406f92202db07eb
1542 dangling commit bf093535a34a4d35731aa2bd90fe6b176302f14f
1543 dangling commit 8e4bec7f2ddaa268bef999853c25755452100f8e
1544 dangling tree d50bb86186bf27b681d25af89d3b5b68382e4085
1545 dangling tree b24c2473f1fd3d91352a624795be026d64c8841f
1546 ...
1547 -------------------------------------------------
1548
1549 Dangling objects are not a problem. At worst they may take up a little
1550 extra disk space. They can sometimes provide a last-resort method for
1551 recovering lost work--see <<dangling-objects>> for details. However, if
1552 you wish, you can remove them with gitlink:git-prune[1] or the --prune
1553 option to gitlink:git-gc[1]:
1554
1555 -------------------------------------------------
1556 $ git gc --prune
1557 -------------------------------------------------
1558
1559 This may be time-consuming. Unlike most other git operations (including
1560 git-gc when run without any options), it is not safe to prune while
1561 other git operations are in progress in the same repository.
1562
1563 If gitlink:git-fsck[1] complains about sha1 mismatches or missing
1564 objects, you may have a much more serious problem; your best option is
1565 probably restoring from backups. See
1566 <<recovering-from-repository-corruption>> for a detailed discussion.
1567
1568 [[recovering-lost-changes]]
1569 Recovering lost changes
1570 ~~~~~~~~~~~~~~~~~~~~~~~
1571
1572 [[reflogs]]
1573 Reflogs
1574 ^^^^^^^
1575
1576 Say you modify a branch with gitlink:git-reset[1] --hard, and then
1577 realize that the branch was the only reference you had to that point in
1578 history.
1579
1580 Fortunately, git also keeps a log, called a "reflog", of all the
1581 previous values of each branch. So in this case you can still find the
1582 old history using, for example,
1583
1584 -------------------------------------------------
1585 $ git log master@{1}
1586 -------------------------------------------------
1587
1588 This lists the commits reachable from the previous version of the head.
1589 This syntax can be used to with any git command that accepts a commit,
1590 not just with git log. Some other examples:
1591
1592 -------------------------------------------------
1593 $ git show master@{2} # See where the branch pointed 2,
1594 $ git show master@{3} # 3, ... changes ago.
1595 $ gitk master@{yesterday} # See where it pointed yesterday,
1596 $ gitk master@{"1 week ago"} # ... or last week
1597 $ git log --walk-reflogs master # show reflog entries for master
1598 -------------------------------------------------
1599
1600 A separate reflog is kept for the HEAD, so
1601
1602 -------------------------------------------------
1603 $ git show HEAD@{"1 week ago"}
1604 -------------------------------------------------
1605
1606 will show what HEAD pointed to one week ago, not what the current branch
1607 pointed to one week ago. This allows you to see the history of what
1608 you've checked out.
1609
1610 The reflogs are kept by default for 30 days, after which they may be
1611 pruned. See gitlink:git-reflog[1] and gitlink:git-gc[1] to learn
1612 how to control this pruning, and see the "SPECIFYING REVISIONS"
1613 section of gitlink:git-rev-parse[1] for details.
1614
1615 Note that the reflog history is very different from normal git history.
1616 While normal history is shared by every repository that works on the
1617 same project, the reflog history is not shared: it tells you only about
1618 how the branches in your local repository have changed over time.
1619
1620 [[dangling-object-recovery]]
1621 Examining dangling objects
1622 ^^^^^^^^^^^^^^^^^^^^^^^^^^
1623
1624 In some situations the reflog may not be able to save you. For example,
1625 suppose you delete a branch, then realize you need the history it
1626 contained. The reflog is also deleted; however, if you have not yet
1627 pruned the repository, then you may still be able to find the lost
1628 commits in the dangling objects that git-fsck reports. See
1629 <<dangling-objects>> for the details.
1630
1631 -------------------------------------------------
1632 $ git fsck
1633 dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1634 dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1635 dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1636 ...
1637 -------------------------------------------------
1638
1639 You can examine
1640 one of those dangling commits with, for example,
1641
1642 ------------------------------------------------
1643 $ gitk 7281251ddd --not --all
1644 ------------------------------------------------
1645
1646 which does what it sounds like: it says that you want to see the commit
1647 history that is described by the dangling commit(s), but not the
1648 history that is described by all your existing branches and tags. Thus
1649 you get exactly the history reachable from that commit that is lost.
1650 (And notice that it might not be just one commit: we only report the
1651 "tip of the line" as being dangling, but there might be a whole deep
1652 and complex commit history that was dropped.)
1653
1654 If you decide you want the history back, you can always create a new
1655 reference pointing to it, for example, a new branch:
1656
1657 ------------------------------------------------
1658 $ git branch recovered-branch 7281251ddd
1659 ------------------------------------------------
1660
1661 Other types of dangling objects (blobs and trees) are also possible, and
1662 dangling objects can arise in other situations.
1663
1664
1665 [[sharing-development]]
1666 Sharing development with others
1667 ===============================
1668
1669 [[getting-updates-with-git-pull]]
1670 Getting updates with git pull
1671 -----------------------------
1672
1673 After you clone a repository and make a few changes of your own, you
1674 may wish to check the original repository for updates and merge them
1675 into your own work.
1676
1677 We have already seen <<Updating-a-repository-with-git-fetch,how to
1678 keep remote tracking branches up to date>> with gitlink:git-fetch[1],
1679 and how to merge two branches. So you can merge in changes from the
1680 original repository's master branch with:
1681
1682 -------------------------------------------------
1683 $ git fetch
1684 $ git merge origin/master
1685 -------------------------------------------------
1686
1687 However, the gitlink:git-pull[1] command provides a way to do this in
1688 one step:
1689
1690 -------------------------------------------------
1691 $ git pull origin master
1692 -------------------------------------------------
1693
1694 In fact, if you have "master" checked out, then by default "git pull"
1695 merges from the HEAD branch of the origin repository. So often you can
1696 accomplish the above with just a simple
1697
1698 -------------------------------------------------
1699 $ git pull
1700 -------------------------------------------------
1701
1702 More generally, a branch that is created from a remote branch will pull
1703 by default from that branch. See the descriptions of the
1704 branch.<name>.remote and branch.<name>.merge options in
1705 gitlink:git-config[1], and the discussion of the --track option in
1706 gitlink:git-checkout[1], to learn how to control these defaults.
1707
1708 In addition to saving you keystrokes, "git pull" also helps you by
1709 producing a default commit message documenting the branch and
1710 repository that you pulled from.
1711
1712 (But note that no such commit will be created in the case of a
1713 <<fast-forwards,fast forward>>; instead, your branch will just be
1714 updated to point to the latest commit from the upstream branch.)
1715
1716 The git-pull command can also be given "." as the "remote" repository,
1717 in which case it just merges in a branch from the current repository; so
1718 the commands
1719
1720 -------------------------------------------------
1721 $ git pull . branch
1722 $ git merge branch
1723 -------------------------------------------------
1724
1725 are roughly equivalent. The former is actually very commonly used.
1726
1727 [[submitting-patches]]
1728 Submitting patches to a project
1729 -------------------------------
1730
1731 If you just have a few changes, the simplest way to submit them may
1732 just be to send them as patches in email:
1733
1734 First, use gitlink:git-format-patch[1]; for example:
1735
1736 -------------------------------------------------
1737 $ git format-patch origin
1738 -------------------------------------------------
1739
1740 will produce a numbered series of files in the current directory, one
1741 for each patch in the current branch but not in origin/HEAD.
1742
1743 You can then import these into your mail client and send them by
1744 hand. However, if you have a lot to send at once, you may prefer to
1745 use the gitlink:git-send-email[1] script to automate the process.
1746 Consult the mailing list for your project first to determine how they
1747 prefer such patches be handled.
1748
1749 [[importing-patches]]
1750 Importing patches to a project
1751 ------------------------------
1752
1753 Git also provides a tool called gitlink:git-am[1] (am stands for
1754 "apply mailbox"), for importing such an emailed series of patches.
1755 Just save all of the patch-containing messages, in order, into a
1756 single mailbox file, say "patches.mbox", then run
1757
1758 -------------------------------------------------
1759 $ git am -3 patches.mbox
1760 -------------------------------------------------
1761
1762 Git will apply each patch in order; if any conflicts are found, it
1763 will stop, and you can fix the conflicts as described in
1764 "<<resolving-a-merge,Resolving a merge>>". (The "-3" option tells
1765 git to perform a merge; if you would prefer it just to abort and
1766 leave your tree and index untouched, you may omit that option.)
1767
1768 Once the index is updated with the results of the conflict
1769 resolution, instead of creating a new commit, just run
1770
1771 -------------------------------------------------
1772 $ git am --resolved
1773 -------------------------------------------------
1774
1775 and git will create the commit for you and continue applying the
1776 remaining patches from the mailbox.
1777
1778 The final result will be a series of commits, one for each patch in
1779 the original mailbox, with authorship and commit log message each
1780 taken from the message containing each patch.
1781
1782 [[public-repositories]]
1783 Public git repositories
1784 -----------------------
1785
1786 Another way to submit changes to a project is to tell the maintainer
1787 of that project to pull the changes from your repository using
1788 gitlink:git-pull[1]. In the section "<<getting-updates-with-git-pull,
1789 Getting updates with git pull>>" we described this as a way to get
1790 updates from the "main" repository, but it works just as well in the
1791 other direction.
1792
1793 If you and the maintainer both have accounts on the same machine, then
1794 you can just pull changes from each other's repositories directly;
1795 commands that accept repository URLs as arguments will also accept a
1796 local directory name:
1797
1798 -------------------------------------------------
1799 $ git clone /path/to/repository
1800 $ git pull /path/to/other/repository
1801 -------------------------------------------------
1802
1803 or an ssh url:
1804
1805 -------------------------------------------------
1806 $ git clone ssh://yourhost/~you/repository
1807 -------------------------------------------------
1808
1809 For projects with few developers, or for synchronizing a few private
1810 repositories, this may be all you need.
1811
1812 However, the more common way to do this is to maintain a separate public
1813 repository (usually on a different host) for others to pull changes
1814 from. This is usually more convenient, and allows you to cleanly
1815 separate private work in progress from publicly visible work.
1816
1817 You will continue to do your day-to-day work in your personal
1818 repository, but periodically "push" changes from your personal
1819 repository into your public repository, allowing other developers to
1820 pull from that repository. So the flow of changes, in a situation
1821 where there is one other developer with a public repository, looks
1822 like this:
1823
1824 you push
1825 your personal repo ------------------> your public repo
1826 ^ |
1827 | |
1828 | you pull | they pull
1829 | |
1830 | |
1831 | they push V
1832 their public repo <------------------- their repo
1833
1834 We explain how to do this in the following sections.
1835
1836 [[setting-up-a-public-repository]]
1837 Setting up a public repository
1838 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1839
1840 Assume your personal repository is in the directory ~/proj. We
1841 first create a new clone of the repository and tell git-daemon that it
1842 is meant to be public:
1843
1844 -------------------------------------------------
1845 $ git clone --bare ~/proj proj.git
1846 $ touch proj.git/git-daemon-export-ok
1847 -------------------------------------------------
1848
1849 The resulting directory proj.git contains a "bare" git repository--it is
1850 just the contents of the ".git" directory, without any files checked out
1851 around it.
1852
1853 Next, copy proj.git to the server where you plan to host the
1854 public repository. You can use scp, rsync, or whatever is most
1855 convenient.
1856
1857 [[exporting-via-git]]
1858 Exporting a git repository via the git protocol
1859 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1860
1861 This is the preferred method.
1862
1863 If someone else administers the server, they should tell you what
1864 directory to put the repository in, and what git:// url it will appear
1865 at. You can then skip to the section
1866 "<<pushing-changes-to-a-public-repository,Pushing changes to a public
1867 repository>>", below.
1868
1869 Otherwise, all you need to do is start gitlink:git-daemon[1]; it will
1870 listen on port 9418. By default, it will allow access to any directory
1871 that looks like a git directory and contains the magic file
1872 git-daemon-export-ok. Passing some directory paths as git-daemon
1873 arguments will further restrict the exports to those paths.
1874
1875 You can also run git-daemon as an inetd service; see the
1876 gitlink:git-daemon[1] man page for details. (See especially the
1877 examples section.)
1878
1879 [[exporting-via-http]]
1880 Exporting a git repository via http
1881 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1882
1883 The git protocol gives better performance and reliability, but on a
1884 host with a web server set up, http exports may be simpler to set up.
1885
1886 All you need to do is place the newly created bare git repository in
1887 a directory that is exported by the web server, and make some
1888 adjustments to give web clients some extra information they need:
1889
1890 -------------------------------------------------
1891 $ mv proj.git /home/you/public_html/proj.git
1892 $ cd proj.git
1893 $ git --bare update-server-info
1894 $ chmod a+x hooks/post-update
1895 -------------------------------------------------
1896
1897 (For an explanation of the last two lines, see
1898 gitlink:git-update-server-info[1], and the documentation
1899 link:hooks.html[Hooks used by git].)
1900
1901 Advertise the url of proj.git. Anybody else should then be able to
1902 clone or pull from that url, for example with a command line like:
1903
1904 -------------------------------------------------
1905 $ git clone http://yourserver.com/~you/proj.git
1906 -------------------------------------------------
1907
1908 (See also
1909 link:howto/setup-git-server-over-http.txt[setup-git-server-over-http]
1910 for a slightly more sophisticated setup using WebDAV which also
1911 allows pushing over http.)
1912
1913 [[pushing-changes-to-a-public-repository]]
1914 Pushing changes to a public repository
1915 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1916
1917 Note that the two techniques outlined above (exporting via
1918 <<exporting-via-http,http>> or <<exporting-via-git,git>>) allow other
1919 maintainers to fetch your latest changes, but they do not allow write
1920 access, which you will need to update the public repository with the
1921 latest changes created in your private repository.
1922
1923 The simplest way to do this is using gitlink:git-push[1] and ssh; to
1924 update the remote branch named "master" with the latest state of your
1925 branch named "master", run
1926
1927 -------------------------------------------------
1928 $ git push ssh://yourserver.com/~you/proj.git master:master
1929 -------------------------------------------------
1930
1931 or just
1932
1933 -------------------------------------------------
1934 $ git push ssh://yourserver.com/~you/proj.git master
1935 -------------------------------------------------
1936
1937 As with git-fetch, git-push will complain if this does not result in a
1938 <<fast-forwards,fast forward>>; see the following section for details on
1939 handling this case.
1940
1941 Note that the target of a "push" is normally a
1942 <<def_bare_repository,bare>> repository. You can also push to a
1943 repository that has a checked-out working tree, but the working tree
1944 will not be updated by the push. This may lead to unexpected results if
1945 the branch you push to is the currently checked-out branch!
1946
1947 As with git-fetch, you may also set up configuration options to
1948 save typing; so, for example, after
1949
1950 -------------------------------------------------
1951 $ cat >>.git/config <<EOF
1952 [remote "public-repo"]
1953 url = ssh://yourserver.com/~you/proj.git
1954 EOF
1955 -------------------------------------------------
1956
1957 you should be able to perform the above push with just
1958
1959 -------------------------------------------------
1960 $ git push public-repo master
1961 -------------------------------------------------
1962
1963 See the explanations of the remote.<name>.url, branch.<name>.remote,
1964 and remote.<name>.push options in gitlink:git-config[1] for
1965 details.
1966
1967 [[forcing-push]]
1968 What to do when a push fails
1969 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1970
1971 If a push would not result in a <<fast-forwards,fast forward>> of the
1972 remote branch, then it will fail with an error like:
1973
1974 -------------------------------------------------
1975 error: remote 'refs/heads/master' is not an ancestor of
1976 local 'refs/heads/master'.
1977 Maybe you are not up-to-date and need to pull first?
1978 error: failed to push to 'ssh://yourserver.com/~you/proj.git'
1979 -------------------------------------------------
1980
1981 This can happen, for example, if you:
1982
1983 - use `git reset --hard` to remove already-published commits, or
1984 - use `git commit --amend` to replace already-published commits
1985 (as in <<fixing-a-mistake-by-rewriting-history>>), or
1986 - use `git rebase` to rebase any already-published commits (as
1987 in <<using-git-rebase>>).
1988
1989 You may force git-push to perform the update anyway by preceding the
1990 branch name with a plus sign:
1991
1992 -------------------------------------------------
1993 $ git push ssh://yourserver.com/~you/proj.git +master
1994 -------------------------------------------------
1995
1996 Normally whenever a branch head in a public repository is modified, it
1997 is modified to point to a descendent of the commit that it pointed to
1998 before. By forcing a push in this situation, you break that convention.
1999 (See <<problems-with-rewriting-history>>.)
2000
2001 Nevertheless, this is a common practice for people that need a simple
2002 way to publish a work-in-progress patch series, and it is an acceptable
2003 compromise as long as you warn other developers that this is how you
2004 intend to manage the branch.
2005
2006 It's also possible for a push to fail in this way when other people have
2007 the right to push to the same repository. In that case, the correct
2008 solution is to retry the push after first updating your work by either a
2009 pull or a fetch followed by a rebase; see the
2010 <<setting-up-a-shared-repository,next section>> and
2011 link:cvs-migration.html[git for CVS users] for more.
2012
2013 [[setting-up-a-shared-repository]]
2014 Setting up a shared repository
2015 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2016
2017 Another way to collaborate is by using a model similar to that
2018 commonly used in CVS, where several developers with special rights
2019 all push to and pull from a single shared repository. See
2020 link:cvs-migration.html[git for CVS users] for instructions on how to
2021 set this up.
2022
2023 However, while there is nothing wrong with git's support for shared
2024 repositories, this mode of operation is not generally recommended,
2025 simply because the mode of collaboration that git supports--by
2026 exchanging patches and pulling from public repositories--has so many
2027 advantages over the central shared repository:
2028
2029 - Git's ability to quickly import and merge patches allows a
2030 single maintainer to process incoming changes even at very
2031 high rates. And when that becomes too much, git-pull provides
2032 an easy way for that maintainer to delegate this job to other
2033 maintainers while still allowing optional review of incoming
2034 changes.
2035 - Since every developer's repository has the same complete copy
2036 of the project history, no repository is special, and it is
2037 trivial for another developer to take over maintenance of a
2038 project, either by mutual agreement, or because a maintainer
2039 becomes unresponsive or difficult to work with.
2040 - The lack of a central group of "committers" means there is
2041 less need for formal decisions about who is "in" and who is
2042 "out".
2043
2044 [[setting-up-gitweb]]
2045 Allowing web browsing of a repository
2046 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2047
2048 The gitweb cgi script provides users an easy way to browse your
2049 project's files and history without having to install git; see the file
2050 gitweb/INSTALL in the git source tree for instructions on setting it up.
2051
2052 [[sharing-development-examples]]
2053 Examples
2054 --------
2055
2056 [[maintaining-topic-branches]]
2057 Maintaining topic branches for a Linux subsystem maintainer
2058 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2059
2060 This describes how Tony Luck uses git in his role as maintainer of the
2061 IA64 architecture for the Linux kernel.
2062
2063 He uses two public branches:
2064
2065 - A "test" tree into which patches are initially placed so that they
2066 can get some exposure when integrated with other ongoing development.
2067 This tree is available to Andrew for pulling into -mm whenever he
2068 wants.
2069
2070 - A "release" tree into which tested patches are moved for final sanity
2071 checking, and as a vehicle to send them upstream to Linus (by sending
2072 him a "please pull" request.)
2073
2074 He also uses a set of temporary branches ("topic branches"), each
2075 containing a logical grouping of patches.
2076
2077 To set this up, first create your work tree by cloning Linus's public
2078 tree:
2079
2080 -------------------------------------------------
2081 $ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git work
2082 $ cd work
2083 -------------------------------------------------
2084
2085 Linus's tree will be stored in the remote branch named origin/master,
2086 and can be updated using gitlink:git-fetch[1]; you can track other
2087 public trees using gitlink:git-remote[1] to set up a "remote" and
2088 gitlink:git-fetch[1] to keep them up-to-date; see
2089 <<repositories-and-branches>>.
2090
2091 Now create the branches in which you are going to work; these start out
2092 at the current tip of origin/master branch, and should be set up (using
2093 the --track option to gitlink:git-branch[1]) to merge changes in from
2094 Linus by default.
2095
2096 -------------------------------------------------
2097 $ git branch --track test origin/master
2098 $ git branch --track release origin/master
2099 -------------------------------------------------
2100
2101 These can be easily kept up to date using gitlink:git-pull[1]
2102
2103 -------------------------------------------------
2104 $ git checkout test && git pull
2105 $ git checkout release && git pull
2106 -------------------------------------------------
2107
2108 Important note! If you have any local changes in these branches, then
2109 this merge will create a commit object in the history (with no local
2110 changes git will simply do a "Fast forward" merge). Many people dislike
2111 the "noise" that this creates in the Linux history, so you should avoid
2112 doing this capriciously in the "release" branch, as these noisy commits
2113 will become part of the permanent history when you ask Linus to pull
2114 from the release branch.
2115
2116 A few configuration variables (see gitlink:git-config[1]) can
2117 make it easy to push both branches to your public tree. (See
2118 <<setting-up-a-public-repository>>.)
2119
2120 -------------------------------------------------
2121 $ cat >> .git/config <<EOF
2122 [remote "mytree"]
2123 url = master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6.git
2124 push = release
2125 push = test
2126 EOF
2127 -------------------------------------------------
2128
2129 Then you can push both the test and release trees using
2130 gitlink:git-push[1]:
2131
2132 -------------------------------------------------
2133 $ git push mytree
2134 -------------------------------------------------
2135
2136 or push just one of the test and release branches using:
2137
2138 -------------------------------------------------
2139 $ git push mytree test
2140 -------------------------------------------------
2141
2142 or
2143
2144 -------------------------------------------------
2145 $ git push mytree release
2146 -------------------------------------------------
2147
2148 Now to apply some patches from the community. Think of a short
2149 snappy name for a branch to hold this patch (or related group of
2150 patches), and create a new branch from the current tip of Linus's
2151 branch:
2152
2153 -------------------------------------------------
2154 $ git checkout -b speed-up-spinlocks origin
2155 -------------------------------------------------
2156
2157 Now you apply the patch(es), run some tests, and commit the change(s). If
2158 the patch is a multi-part series, then you should apply each as a separate
2159 commit to this branch.
2160
2161 -------------------------------------------------
2162 $ ... patch ... test ... commit [ ... patch ... test ... commit ]*
2163 -------------------------------------------------
2164
2165 When you are happy with the state of this change, you can pull it into the
2166 "test" branch in preparation to make it public:
2167
2168 -------------------------------------------------
2169 $ git checkout test && git pull . speed-up-spinlocks
2170 -------------------------------------------------
2171
2172 It is unlikely that you would have any conflicts here ... but you might if you
2173 spent a while on this step and had also pulled new versions from upstream.
2174
2175 Some time later when enough time has passed and testing done, you can pull the
2176 same branch into the "release" tree ready to go upstream. This is where you
2177 see the value of keeping each patch (or patch series) in its own branch. It
2178 means that the patches can be moved into the "release" tree in any order.
2179
2180 -------------------------------------------------
2181 $ git checkout release && git pull . speed-up-spinlocks
2182 -------------------------------------------------
2183
2184 After a while, you will have a number of branches, and despite the
2185 well chosen names you picked for each of them, you may forget what
2186 they are for, or what status they are in. To get a reminder of what
2187 changes are in a specific branch, use:
2188
2189 -------------------------------------------------
2190 $ git log linux..branchname | git-shortlog
2191 -------------------------------------------------
2192
2193 To see whether it has already been merged into the test or release branches
2194 use:
2195
2196 -------------------------------------------------
2197 $ git log test..branchname
2198 -------------------------------------------------
2199
2200 or
2201
2202 -------------------------------------------------
2203 $ git log release..branchname
2204 -------------------------------------------------
2205
2206 (If this branch has not yet been merged you will see some log entries.
2207 If it has been merged, then there will be no output.)
2208
2209 Once a patch completes the great cycle (moving from test to release,
2210 then pulled by Linus, and finally coming back into your local
2211 "origin/master" branch) the branch for this change is no longer needed.
2212 You detect this when the output from:
2213
2214 -------------------------------------------------
2215 $ git log origin..branchname
2216 -------------------------------------------------
2217
2218 is empty. At this point the branch can be deleted:
2219
2220 -------------------------------------------------
2221 $ git branch -d branchname
2222 -------------------------------------------------
2223
2224 Some changes are so trivial that it is not necessary to create a separate
2225 branch and then merge into each of the test and release branches. For
2226 these changes, just apply directly to the "release" branch, and then
2227 merge that into the "test" branch.
2228
2229 To create diffstat and shortlog summaries of changes to include in a "please
2230 pull" request to Linus you can use:
2231
2232 -------------------------------------------------
2233 $ git diff --stat origin..release
2234 -------------------------------------------------
2235
2236 and
2237
2238 -------------------------------------------------
2239 $ git log -p origin..release | git shortlog
2240 -------------------------------------------------
2241
2242 Here are some of the scripts that simplify all this even further.
2243
2244 -------------------------------------------------
2245 ==== update script ====
2246 # Update a branch in my GIT tree. If the branch to be updated
2247 # is origin, then pull from kernel.org. Otherwise merge
2248 # origin/master branch into test|release branch
2249
2250 case "$1" in
2251 test|release)
2252 git checkout $1 && git pull . origin
2253 ;;
2254 origin)
2255 before=$(git rev-parse refs/remotes/origin/master)
2256 git fetch origin
2257 after=$(git rev-parse refs/remotes/origin/master)
2258 if [ $before != $after ]
2259 then
2260 git log $before..$after | git shortlog
2261 fi
2262 ;;
2263 *)
2264 echo "Usage: $0 origin|test|release" 1>&2
2265 exit 1
2266 ;;
2267 esac
2268 -------------------------------------------------
2269
2270 -------------------------------------------------
2271 ==== merge script ====
2272 # Merge a branch into either the test or release branch
2273
2274 pname=$0
2275
2276 usage()
2277 {
2278 echo "Usage: $pname branch test|release" 1>&2
2279 exit 1
2280 }
2281
2282 git show-ref -q --verify -- refs/heads/"$1" || {
2283 echo "Can't see branch <$1>" 1>&2
2284 usage
2285 }
2286
2287 case "$2" in
2288 test|release)
2289 if [ $(git log $2..$1 | wc -c) -eq 0 ]
2290 then
2291 echo $1 already merged into $2 1>&2
2292 exit 1
2293 fi
2294 git checkout $2 && git pull . $1
2295 ;;
2296 *)
2297 usage
2298 ;;
2299 esac
2300 -------------------------------------------------
2301
2302 -------------------------------------------------
2303 ==== status script ====
2304 # report on status of my ia64 GIT tree
2305
2306 gb=$(tput setab 2)
2307 rb=$(tput setab 1)
2308 restore=$(tput setab 9)
2309
2310 if [ `git rev-list test..release | wc -c` -gt 0 ]
2311 then
2312 echo $rb Warning: commits in release that are not in test $restore
2313 git log test..release
2314 fi
2315
2316 for branch in `git show-ref --heads | sed 's|^.*/||'`
2317 do
2318 if [ $branch = test -o $branch = release ]
2319 then
2320 continue
2321 fi
2322
2323 echo -n $gb ======= $branch ====== $restore " "
2324 status=
2325 for ref in test release origin/master
2326 do
2327 if [ `git rev-list $ref..$branch | wc -c` -gt 0 ]
2328 then
2329 status=$status${ref:0:1}
2330 fi
2331 done
2332 case $status in
2333 trl)
2334 echo $rb Need to pull into test $restore
2335 ;;
2336 rl)
2337 echo "In test"
2338 ;;
2339 l)
2340 echo "Waiting for linus"
2341 ;;
2342 "")
2343 echo $rb All done $restore
2344 ;;
2345 *)
2346 echo $rb "<$status>" $restore
2347 ;;
2348 esac
2349 git log origin/master..$branch | git shortlog
2350 done
2351 -------------------------------------------------
2352
2353
2354 [[cleaning-up-history]]
2355 Rewriting history and maintaining patch series
2356 ==============================================
2357
2358 Normally commits are only added to a project, never taken away or
2359 replaced. Git is designed with this assumption, and violating it will
2360 cause git's merge machinery (for example) to do the wrong thing.
2361
2362 However, there is a situation in which it can be useful to violate this
2363 assumption.
2364
2365 [[patch-series]]
2366 Creating the perfect patch series
2367 ---------------------------------
2368
2369 Suppose you are a contributor to a large project, and you want to add a
2370 complicated feature, and to present it to the other developers in a way
2371 that makes it easy for them to read your changes, verify that they are
2372 correct, and understand why you made each change.
2373
2374 If you present all of your changes as a single patch (or commit), they
2375 may find that it is too much to digest all at once.
2376
2377 If you present them with the entire history of your work, complete with
2378 mistakes, corrections, and dead ends, they may be overwhelmed.
2379
2380 So the ideal is usually to produce a series of patches such that:
2381
2382 1. Each patch can be applied in order.
2383
2384 2. Each patch includes a single logical change, together with a
2385 message explaining the change.
2386
2387 3. No patch introduces a regression: after applying any initial
2388 part of the series, the resulting project still compiles and
2389 works, and has no bugs that it didn't have before.
2390
2391 4. The complete series produces the same end result as your own
2392 (probably much messier!) development process did.
2393
2394 We will introduce some tools that can help you do this, explain how to
2395 use them, and then explain some of the problems that can arise because
2396 you are rewriting history.
2397
2398 [[using-git-rebase]]
2399 Keeping a patch series up to date using git-rebase
2400 --------------------------------------------------
2401
2402 Suppose that you create a branch "mywork" on a remote-tracking branch
2403 "origin", and create some commits on top of it:
2404
2405 -------------------------------------------------
2406 $ git checkout -b mywork origin
2407 $ vi file.txt
2408 $ git commit
2409 $ vi otherfile.txt
2410 $ git commit
2411 ...
2412 -------------------------------------------------
2413
2414 You have performed no merges into mywork, so it is just a simple linear
2415 sequence of patches on top of "origin":
2416
2417 ................................................
2418 o--o--o <-- origin
2419 \
2420 o--o--o <-- mywork
2421 ................................................
2422
2423 Some more interesting work has been done in the upstream project, and
2424 "origin" has advanced:
2425
2426 ................................................
2427 o--o--O--o--o--o <-- origin
2428 \
2429 a--b--c <-- mywork
2430 ................................................
2431
2432 At this point, you could use "pull" to merge your changes back in;
2433 the result would create a new merge commit, like this:
2434
2435 ................................................
2436 o--o--O--o--o--o <-- origin
2437 \ \
2438 a--b--c--m <-- mywork
2439 ................................................
2440
2441 However, if you prefer to keep the history in mywork a simple series of
2442 commits without any merges, you may instead choose to use
2443 gitlink:git-rebase[1]:
2444
2445 -------------------------------------------------
2446 $ git checkout mywork
2447 $ git rebase origin
2448 -------------------------------------------------
2449
2450 This will remove each of your commits from mywork, temporarily saving
2451 them as patches (in a directory named ".dotest"), update mywork to
2452 point at the latest version of origin, then apply each of the saved
2453 patches to the new mywork. The result will look like:
2454
2455
2456 ................................................
2457 o--o--O--o--o--o <-- origin
2458 \
2459 a'--b'--c' <-- mywork
2460 ................................................
2461
2462 In the process, it may discover conflicts. In that case it will stop
2463 and allow you to fix the conflicts; after fixing conflicts, use "git
2464 add" to update the index with those contents, and then, instead of
2465 running git-commit, just run
2466
2467 -------------------------------------------------
2468 $ git rebase --continue
2469 -------------------------------------------------
2470
2471 and git will continue applying the rest of the patches.
2472
2473 At any point you may use the --abort option to abort this process and
2474 return mywork to the state it had before you started the rebase:
2475
2476 -------------------------------------------------
2477 $ git rebase --abort
2478 -------------------------------------------------
2479
2480 [[rewriting-one-commit]]
2481 Rewriting a single commit
2482 -------------------------
2483
2484 We saw in <<fixing-a-mistake-by-rewriting-history>> that you can replace the
2485 most recent commit using
2486
2487 -------------------------------------------------
2488 $ git commit --amend
2489 -------------------------------------------------
2490
2491 which will replace the old commit by a new commit incorporating your
2492 changes, giving you a chance to edit the old commit message first.
2493
2494 You can also use a combination of this and gitlink:git-rebase[1] to
2495 replace a commit further back in your history and recreate the
2496 intervening changes on top of it. First, tag the problematic commit
2497 with
2498
2499 -------------------------------------------------
2500 $ git tag bad mywork~5
2501 -------------------------------------------------
2502
2503 (Either gitk or git-log may be useful for finding the commit.)
2504
2505 Then check out that commit, edit it, and rebase the rest of the series
2506 on top of it (note that we could check out the commit on a temporary
2507 branch, but instead we're using a <<detached-head,detached head>>):
2508
2509 -------------------------------------------------
2510 $ git checkout bad
2511 $ # make changes here and update the index
2512 $ git commit --amend
2513 $ git rebase --onto HEAD bad mywork
2514 -------------------------------------------------
2515
2516 When you're done, you'll be left with mywork checked out, with the top
2517 patches on mywork reapplied on top of your modified commit. You can
2518 then clean up with
2519
2520 -------------------------------------------------
2521 $ git tag -d bad
2522 -------------------------------------------------
2523
2524 Note that the immutable nature of git history means that you haven't really
2525 "modified" existing commits; instead, you have replaced the old commits with
2526 new commits having new object names.
2527
2528 [[reordering-patch-series]]
2529 Reordering or selecting from a patch series
2530 -------------------------------------------
2531
2532 Given one existing commit, the gitlink:git-cherry-pick[1] command
2533 allows you to apply the change introduced by that commit and create a
2534 new commit that records it. So, for example, if "mywork" points to a
2535 series of patches on top of "origin", you might do something like:
2536
2537 -------------------------------------------------
2538 $ git checkout -b mywork-new origin
2539 $ gitk origin..mywork &
2540 -------------------------------------------------
2541
2542 And browse through the list of patches in the mywork branch using gitk,
2543 applying them (possibly in a different order) to mywork-new using
2544 cherry-pick, and possibly modifying them as you go using commit --amend.
2545 The gitlink:git-gui[1] command may also help as it allows you to
2546 individually select diff hunks for inclusion in the index (by
2547 right-clicking on the diff hunk and choosing "Stage Hunk for Commit").
2548
2549 Another technique is to use git-format-patch to create a series of
2550 patches, then reset the state to before the patches:
2551
2552 -------------------------------------------------
2553 $ git format-patch origin
2554 $ git reset --hard origin
2555 -------------------------------------------------
2556
2557 Then modify, reorder, or eliminate patches as preferred before applying
2558 them again with gitlink:git-am[1].
2559
2560 [[patch-series-tools]]
2561 Other tools
2562 -----------
2563
2564 There are numerous other tools, such as StGIT, which exist for the
2565 purpose of maintaining a patch series. These are outside of the scope of
2566 this manual.
2567
2568 [[problems-with-rewriting-history]]
2569 Problems with rewriting history
2570 -------------------------------
2571
2572 The primary problem with rewriting the history of a branch has to do
2573 with merging. Suppose somebody fetches your branch and merges it into
2574 their branch, with a result something like this:
2575
2576 ................................................
2577 o--o--O--o--o--o <-- origin
2578 \ \
2579 t--t--t--m <-- their branch:
2580 ................................................
2581
2582 Then suppose you modify the last three commits:
2583
2584 ................................................
2585 o--o--o <-- new head of origin
2586 /
2587 o--o--O--o--o--o <-- old head of origin
2588 ................................................
2589
2590 If we examined all this history together in one repository, it will
2591 look like:
2592
2593 ................................................
2594 o--o--o <-- new head of origin
2595 /
2596 o--o--O--o--o--o <-- old head of origin
2597 \ \
2598 t--t--t--m <-- their branch:
2599 ................................................
2600
2601 Git has no way of knowing that the new head is an updated version of
2602 the old head; it treats this situation exactly the same as it would if
2603 two developers had independently done the work on the old and new heads
2604 in parallel. At this point, if someone attempts to merge the new head
2605 in to their branch, git will attempt to merge together the two (old and
2606 new) lines of development, instead of trying to replace the old by the
2607 new. The results are likely to be unexpected.
2608
2609 You may still choose to publish branches whose history is rewritten,
2610 and it may be useful for others to be able to fetch those branches in
2611 order to examine or test them, but they should not attempt to pull such
2612 branches into their own work.
2613
2614 For true distributed development that supports proper merging,
2615 published branches should never be rewritten.
2616
2617 [[bisect-merges]]
2618 Why bisecting merge commits can be harder than bisecting linear history
2619 -----------------------------------------------------------------------
2620
2621 The gitlink:git-bisect[1] command correctly handles history that
2622 includes merge commits. However, when the commit that it finds is a
2623 merge commit, the user may need to work harder than usual to figure out
2624 why that commit introduced a problem.
2625
2626 Imagine this history:
2627
2628 ................................................
2629 ---Z---o---X---...---o---A---C---D
2630 \ /
2631 o---o---Y---...---o---B
2632 ................................................
2633
2634 Suppose that on the upper line of development, the meaning of one
2635 of the functions that exists at Z is changed at commit X. The
2636 commits from Z leading to A change both the function's
2637 implementation and all calling sites that exist at Z, as well
2638 as new calling sites they add, to be consistent. There is no
2639 bug at A.
2640
2641 Suppose that in the meantime on the lower line of development somebody
2642 adds a new calling site for that function at commit Y. The
2643 commits from Z leading to B all assume the old semantics of that
2644 function and the callers and the callee are consistent with each
2645 other. There is no bug at B, either.
2646
2647 Suppose further that the two development lines merge cleanly at C,
2648 so no conflict resolution is required.
2649
2650 Nevertheless, the code at C is broken, because the callers added
2651 on the lower line of development have not been converted to the new
2652 semantics introduced on the upper line of development. So if all
2653 you know is that D is bad, that Z is good, and that
2654 gitlink:git-bisect[1] identifies C as the culprit, how will you
2655 figure out that the problem is due to this change in semantics?
2656
2657 When the result of a git-bisect is a non-merge commit, you should
2658 normally be able to discover the problem by examining just that commit.
2659 Developers can make this easy by breaking their changes into small
2660 self-contained commits. That won't help in the case above, however,
2661 because the problem isn't obvious from examination of any single
2662 commit; instead, a global view of the development is required. To
2663 make matters worse, the change in semantics in the problematic
2664 function may be just one small part of the changes in the upper
2665 line of development.
2666
2667 On the other hand, if instead of merging at C you had rebased the
2668 history between Z to B on top of A, you would have gotten this
2669 linear history:
2670
2671 ................................................................
2672 ---Z---o---X--...---o---A---o---o---Y*--...---o---B*--D*
2673 ................................................................
2674
2675 Bisecting between Z and D* would hit a single culprit commit Y*,
2676 and understanding why Y* was broken would probably be easier.
2677
2678 Partly for this reason, many experienced git users, even when
2679 working on an otherwise merge-heavy project, keep the history
2680 linear by rebasing against the latest upstream version before
2681 publishing.
2682
2683 [[advanced-branch-management]]
2684 Advanced branch management
2685 ==========================
2686
2687 [[fetching-individual-branches]]
2688 Fetching individual branches
2689 ----------------------------
2690
2691 Instead of using gitlink:git-remote[1], you can also choose just
2692 to update one branch at a time, and to store it locally under an
2693 arbitrary name:
2694
2695 -------------------------------------------------
2696 $ git fetch origin todo:my-todo-work
2697 -------------------------------------------------
2698
2699 The first argument, "origin", just tells git to fetch from the
2700 repository you originally cloned from. The second argument tells git
2701 to fetch the branch named "todo" from the remote repository, and to
2702 store it locally under the name refs/heads/my-todo-work.
2703
2704 You can also fetch branches from other repositories; so
2705
2706 -------------------------------------------------
2707 $ git fetch git://example.com/proj.git master:example-master
2708 -------------------------------------------------
2709
2710 will create a new branch named "example-master" and store in it the
2711 branch named "master" from the repository at the given URL. If you
2712 already have a branch named example-master, it will attempt to
2713 <<fast-forwards,fast-forward>> to the commit given by example.com's
2714 master branch. In more detail:
2715
2716 [[fetch-fast-forwards]]
2717 git fetch and fast-forwards
2718 ---------------------------
2719
2720 In the previous example, when updating an existing branch, "git
2721 fetch" checks to make sure that the most recent commit on the remote
2722 branch is a descendant of the most recent commit on your copy of the
2723 branch before updating your copy of the branch to point at the new
2724 commit. Git calls this process a <<fast-forwards,fast forward>>.
2725
2726 A fast forward looks something like this:
2727
2728 ................................................
2729 o--o--o--o <-- old head of the branch
2730 \
2731 o--o--o <-- new head of the branch
2732 ................................................
2733
2734
2735 In some cases it is possible that the new head will *not* actually be
2736 a descendant of the old head. For example, the developer may have
2737 realized she made a serious mistake, and decided to backtrack,
2738 resulting in a situation like:
2739
2740 ................................................
2741 o--o--o--o--a--b <-- old head of the branch
2742 \
2743 o--o--o <-- new head of the branch
2744 ................................................
2745
2746 In this case, "git fetch" will fail, and print out a warning.
2747
2748 In that case, you can still force git to update to the new head, as
2749 described in the following section. However, note that in the
2750 situation above this may mean losing the commits labeled "a" and "b",
2751 unless you've already created a reference of your own pointing to
2752 them.
2753
2754 [[forcing-fetch]]
2755 Forcing git fetch to do non-fast-forward updates
2756 ------------------------------------------------
2757
2758 If git fetch fails because the new head of a branch is not a
2759 descendant of the old head, you may force the update with:
2760
2761 -------------------------------------------------
2762 $ git fetch git://example.com/proj.git +master:refs/remotes/example/master
2763 -------------------------------------------------
2764
2765 Note the addition of the "+" sign. Alternatively, you can use the "-f"
2766 flag to force updates of all the fetched branches, as in:
2767
2768 -------------------------------------------------
2769 $ git fetch -f origin
2770 -------------------------------------------------
2771
2772 Be aware that commits that the old version of example/master pointed at
2773 may be lost, as we saw in the previous section.
2774
2775 [[remote-branch-configuration]]
2776 Configuring remote branches
2777 ---------------------------
2778
2779 We saw above that "origin" is just a shortcut to refer to the
2780 repository that you originally cloned from. This information is
2781 stored in git configuration variables, which you can see using
2782 gitlink:git-config[1]:
2783
2784 -------------------------------------------------
2785 $ git config -l
2786 core.repositoryformatversion=0
2787 core.filemode=true
2788 core.logallrefupdates=true
2789 remote.origin.url=git://git.kernel.org/pub/scm/git/git.git
2790 remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*
2791 branch.master.remote=origin
2792 branch.master.merge=refs/heads/master
2793 -------------------------------------------------
2794
2795 If there are other repositories that you also use frequently, you can
2796 create similar configuration options to save typing; for example,
2797 after
2798
2799 -------------------------------------------------
2800 $ git config remote.example.url git://example.com/proj.git
2801 -------------------------------------------------
2802
2803 then the following two commands will do the same thing:
2804
2805 -------------------------------------------------
2806 $ git fetch git://example.com/proj.git master:refs/remotes/example/master
2807 $ git fetch example master:refs/remotes/example/master
2808 -------------------------------------------------
2809
2810 Even better, if you add one more option:
2811
2812 -------------------------------------------------
2813 $ git config remote.example.fetch master:refs/remotes/example/master
2814 -------------------------------------------------
2815
2816 then the following commands will all do the same thing:
2817
2818 -------------------------------------------------
2819 $ git fetch git://example.com/proj.git master:refs/remotes/example/master
2820 $ git fetch example master:refs/remotes/example/master
2821 $ git fetch example
2822 -------------------------------------------------
2823
2824 You can also add a "+" to force the update each time:
2825
2826 -------------------------------------------------
2827 $ git config remote.example.fetch +master:ref/remotes/example/master
2828 -------------------------------------------------
2829
2830 Don't do this unless you're sure you won't mind "git fetch" possibly
2831 throwing away commits on mybranch.
2832
2833 Also note that all of the above configuration can be performed by
2834 directly editing the file .git/config instead of using
2835 gitlink:git-config[1].
2836
2837 See gitlink:git-config[1] for more details on the configuration
2838 options mentioned above.
2839
2840
2841 [[git-concepts]]
2842 Git concepts
2843 ============
2844
2845 Git is built on a small number of simple but powerful ideas. While it
2846 is possible to get things done without understanding them, you will find
2847 git much more intuitive if you do.
2848
2849 We start with the most important, the <<def_object_database,object
2850 database>> and the <<def_index,index>>.
2851
2852 [[the-object-database]]
2853 The Object Database
2854 -------------------
2855
2856
2857 We already saw in <<understanding-commits>> that all commits are stored
2858 under a 40-digit "object name". In fact, all the information needed to
2859 represent the history of a project is stored in objects with such names.
2860 In each case the name is calculated by taking the SHA1 hash of the
2861 contents of the object. The SHA1 hash is a cryptographic hash function.
2862 What that means to us is that it is impossible to find two different
2863 objects with the same name. This has a number of advantages; among
2864 others:
2865
2866 - Git can quickly determine whether two objects are identical or not,
2867 just by comparing names.
2868 - Since object names are computed the same way in ever repository, the
2869 same content stored in two repositories will always be stored under
2870 the same name.
2871 - Git can detect errors when it reads an object, by checking that the
2872 object's name is still the SHA1 hash of its contents.
2873
2874 (See <<object-details>> for the details of the object formatting and
2875 SHA1 calculation.)
2876
2877 There are four different types of objects: "blob", "tree", "commit", and
2878 "tag".
2879
2880 - A <<def_blob_object,"blob" object>> is used to store file data.
2881 - A <<def_tree_object,"tree" object>> is an object that ties one or more
2882 "blob" objects into a directory structure. In addition, a tree object
2883 can refer to other tree objects, thus creating a directory hierarchy.
2884 - A <<def_commit_object,"commit" object>> ties such directory hierarchies
2885 together into a <<def_DAG,directed acyclic graph>> of revisions - each
2886 commit contains the object name of exactly one tree designating the
2887 directory hierarchy at the time of the commit. In addition, a commit
2888 refers to "parent" commit objects that describe the history of how we
2889 arrived at that directory hierarchy.
2890 - A <<def_tag_object,"tag" object>> symbolically identifies and can be
2891 used to sign other objects. It contains the object name and type of
2892 another object, a symbolic name (of course!) and, optionally, a
2893 signature.
2894
2895 The object types in some more detail:
2896
2897 [[commit-object]]
2898 Commit Object
2899 ~~~~~~~~~~~~~
2900
2901 The "commit" object links a physical state of a tree with a description
2902 of how we got there and why. Use the --pretty=raw option to
2903 gitlink:git-show[1] or gitlink:git-log[1] to examine your favorite
2904 commit:
2905
2906 ------------------------------------------------
2907 $ git show -s --pretty=raw 2be7fcb476
2908 commit 2be7fcb4764f2dbcee52635b91fedb1b3dcf7ab4
2909 tree fb3a8bdd0ceddd019615af4d57a53f43d8cee2bf
2910 parent 257a84d9d02e90447b149af58b271c19405edb6a
2911 author Dave Watson <dwatson@mimvista.com> 1187576872 -0400
2912 committer Junio C Hamano <gitster@pobox.com> 1187591163 -0700
2913
2914 Fix misspelling of 'suppress' in docs
2915
2916 Signed-off-by: Junio C Hamano <gitster@pobox.com>
2917 ------------------------------------------------
2918
2919 As you can see, a commit is defined by:
2920
2921 - a tree: The SHA1 name of a tree object (as defined below), representing
2922 the contents of a directory at a certain point in time.
2923 - parent(s): The SHA1 name of some number of commits which represent the
2924 immediately prevoius step(s) in the history of the project. The
2925 example above has one parent; merge commits may have more than
2926 one. A commit with no parents is called a "root" commit, and
2927 represents the initial revision of a project. Each project must have
2928 at least one root. A project can also have multiple roots, though
2929 that isn't common (or necessarily a good idea).
2930 - an author: The name of the person responsible for this change, together
2931 with its date.
2932 - a committer: The name of the person who actually created the commit,
2933 with the date it was done. This may be different from the author, for
2934 example, if the author was someone who wrote a patch and emailed it
2935 to the person who used it to create the commit.
2936 - a comment describing this commit.
2937
2938 Note that a commit does not itself contain any information about what
2939 actually changed; all changes are calculated by comparing the contents
2940 of the tree referred to by this commit with the trees associated with
2941 its parents. In particular, git does not attempt to record file renames
2942 explicitly, though it can identify cases where the existence of the same
2943 file data at changing paths suggests a rename. (See, for example, the
2944 -M option to gitlink:git-diff[1]).
2945
2946 A commit is usually created by gitlink:git-commit[1], which creates a
2947 commit whose parent is normally the current HEAD, and whose tree is
2948 taken from the content currently stored in the index.
2949
2950 [[tree-object]]
2951 Tree Object
2952 ~~~~~~~~~~~
2953
2954 The ever-versatile gitlink:git-show[1] command can also be used to
2955 examine tree objects, but gitlink:git-ls-tree[1] will give you more
2956 details:
2957
2958 ------------------------------------------------
2959 $ git ls-tree fb3a8bdd0ce
2960 100644 blob 63c918c667fa005ff12ad89437f2fdc80926e21c .gitignore
2961 100644 blob 5529b198e8d14decbe4ad99db3f7fb632de0439d .mailmap
2962 100644 blob 6ff87c4664981e4397625791c8ea3bbb5f2279a3 COPYING
2963 040000 tree 2fb783e477100ce076f6bf57e4a6f026013dc745 Documentation
2964 100755 blob 3c0032cec592a765692234f1cba47dfdcc3a9200 GIT-VERSION-GEN
2965 100644 blob 289b046a443c0647624607d471289b2c7dcd470b INSTALL
2966 100644 blob 4eb463797adc693dc168b926b6932ff53f17d0b1 Makefile
2967 100644 blob 548142c327a6790ff8821d67c2ee1eff7a656b52 README
2968 ...
2969 ------------------------------------------------
2970
2971 As you can see, a tree object contains a list of entries, each with a
2972 mode, object type, SHA1 name, and name, sorted by name. It represents
2973 the contents of a single directory tree.
2974
2975 The object type may be a blob, representing the contents of a file, or
2976 another tree, representing the contents of a subdirectory. Since trees
2977 and blobs, like all other objects, are named by the SHA1 hash of their
2978 contents, two trees have the same SHA1 name if and only if their
2979 contents (including, recursively, the contents of all subdirectories)
2980 are identical. This allows git to quickly determine the differences
2981 between two related tree objects, since it can ignore any entries with
2982 identical object names.
2983
2984 (Note: in the presence of submodules, trees may also have commits as
2985 entries. See <<submodules>> for documentation.)
2986
2987 Note that the files all have mode 644 or 755: git actually only pays
2988 attention to the executable bit.
2989
2990 [[blob-object]]
2991 Blob Object
2992 ~~~~~~~~~~~
2993
2994 You can use gitlink:git-show[1] to examine the contents of a blob; take,
2995 for example, the blob in the entry for "COPYING" from the tree above:
2996
2997 ------------------------------------------------
2998 $ git show 6ff87c4664
2999
3000 Note that the only valid version of the GPL as far as this project
3001 is concerned is _this_ particular version of the license (ie v2, not
3002 v2.2 or v3.x or whatever), unless explicitly otherwise stated.
3003 ...
3004 ------------------------------------------------
3005
3006 A "blob" object is nothing but a binary blob of data. It doesn't refer
3007 to anything else or have attributes of any kind.
3008
3009 Since the blob is entirely defined by its data, if two files in a
3010 directory tree (or in multiple different versions of the repository)
3011 have the same contents, they will share the same blob object. The object
3012 is totally independent of its location in the directory tree, and
3013 renaming a file does not change the object that file is associated with.
3014
3015 Note that any tree or blob object can be examined using
3016 gitlink:git-show[1] with the <revision>:<path> syntax. This can
3017 sometimes be useful for browsing the contents of a tree that is not
3018 currently checked out.
3019
3020 [[trust]]
3021 Trust
3022 ~~~~~
3023
3024 If you receive the SHA1 name of a blob from one source, and its contents
3025 from another (possibly untrusted) source, you can still trust that those
3026 contents are correct as long as the SHA1 name agrees. This is because
3027 the SHA1 is designed so that it is infeasible to find different contents
3028 that produce the same hash.
3029
3030 Similarly, you need only trust the SHA1 name of a top-level tree object
3031 to trust the contents of the entire directory that it refers to, and if
3032 you receive the SHA1 name of a commit from a trusted source, then you
3033 can easily verify the entire history of commits reachable through
3034 parents of that commit, and all of those contents of the trees referred
3035 to by those commits.
3036
3037 So to introduce some real trust in the system, the only thing you need
3038 to do is to digitally sign just 'one' special note, which includes the
3039 name of a top-level commit. Your digital signature shows others
3040 that you trust that commit, and the immutability of the history of
3041 commits tells others that they can trust the whole history.
3042
3043 In other words, you can easily validate a whole archive by just
3044 sending out a single email that tells the people the name (SHA1 hash)
3045 of the top commit, and digitally sign that email using something
3046 like GPG/PGP.
3047
3048 To assist in this, git also provides the tag object...
3049
3050 [[tag-object]]
3051 Tag Object
3052 ~~~~~~~~~~
3053
3054 A tag object contains an object, object type, tag name, the name of the
3055 person ("tagger") who created the tag, and a message, which may contain
3056 a signature, as can be seen using the gitlink:git-cat-file[1]:
3057
3058 ------------------------------------------------
3059 $ git cat-file tag v1.5.0
3060 object 437b1b20df4b356c9342dac8d38849f24ef44f27
3061 type commit
3062 tag v1.5.0
3063 tagger Junio C Hamano <junkio@cox.net> 1171411200 +0000
3064
3065 GIT 1.5.0
3066 -----BEGIN PGP SIGNATURE-----
3067 Version: GnuPG v1.4.6 (GNU/Linux)
3068
3069 iD8DBQBF0lGqwMbZpPMRm5oRAuRiAJ9ohBLd7s2kqjkKlq1qqC57SbnmzQCdG4ui
3070 nLE/L9aUXdWeTFPron96DLA=
3071 =2E+0
3072 -----END PGP SIGNATURE-----
3073 ------------------------------------------------
3074
3075 See the gitlink:git-tag[1] command to learn how to create and verify tag
3076 objects. (Note that gitlink:git-tag[1] can also be used to create
3077 "lightweight tags", which are not tag objects at all, but just simple
3078 references whose names begin with "refs/tags/").
3079
3080 [[pack-files]]
3081 How git stores objects efficiently: pack files
3082 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3083
3084 Newly created objects are initially created in a file named after the
3085 object's SHA1 hash (stored in .git/objects).
3086
3087 Unfortunately this system becomes inefficient once a project has a
3088 lot of objects. Try this on an old project:
3089
3090 ------------------------------------------------
3091 $ git count-objects
3092 6930 objects, 47620 kilobytes
3093 ------------------------------------------------
3094
3095 The first number is the number of objects which are kept in
3096 individual files. The second is the amount of space taken up by
3097 those "loose" objects.
3098
3099 You can save space and make git faster by moving these loose objects in
3100 to a "pack file", which stores a group of objects in an efficient
3101 compressed format; the details of how pack files are formatted can be
3102 found in link:technical/pack-format.txt[technical/pack-format.txt].
3103
3104 To put the loose objects into a pack, just run git repack:
3105
3106 ------------------------------------------------
3107 $ git repack
3108 Generating pack...
3109 Done counting 6020 objects.
3110 Deltifying 6020 objects.
3111 100% (6020/6020) done
3112 Writing 6020 objects.
3113 100% (6020/6020) done
3114 Total 6020, written 6020 (delta 4070), reused 0 (delta 0)
3115 Pack pack-3e54ad29d5b2e05838c75df582c65257b8d08e1c created.
3116 ------------------------------------------------
3117
3118 You can then run
3119
3120 ------------------------------------------------
3121 $ git prune
3122 ------------------------------------------------
3123
3124 to remove any of the "loose" objects that are now contained in the
3125 pack. This will also remove any unreferenced objects (which may be
3126 created when, for example, you use "git reset" to remove a commit).
3127 You can verify that the loose objects are gone by looking at the
3128 .git/objects directory or by running
3129
3130 ------------------------------------------------
3131 $ git count-objects
3132 0 objects, 0 kilobytes
3133 ------------------------------------------------
3134
3135 Although the object files are gone, any commands that refer to those
3136 objects will work exactly as they did before.
3137
3138 The gitlink:git-gc[1] command performs packing, pruning, and more for
3139 you, so is normally the only high-level command you need.
3140
3141 [[dangling-objects]]
3142 Dangling objects
3143 ~~~~~~~~~~~~~~~~
3144
3145 The gitlink:git-fsck[1] command will sometimes complain about dangling
3146 objects. They are not a problem.
3147
3148 The most common cause of dangling objects is that you've rebased a
3149 branch, or you have pulled from somebody else who rebased a branch--see
3150 <<cleaning-up-history>>. In that case, the old head of the original
3151 branch still exists, as does everything it pointed to. The branch
3152 pointer itself just doesn't, since you replaced it with another one.
3153
3154 There are also other situations that cause dangling objects. For
3155 example, a "dangling blob" may arise because you did a "git add" of a
3156 file, but then, before you actually committed it and made it part of the
3157 bigger picture, you changed something else in that file and committed
3158 that *updated* thing - the old state that you added originally ends up
3159 not being pointed to by any commit or tree, so it's now a dangling blob
3160 object.
3161
3162 Similarly, when the "recursive" merge strategy runs, and finds that
3163 there are criss-cross merges and thus more than one merge base (which is
3164 fairly unusual, but it does happen), it will generate one temporary
3165 midway tree (or possibly even more, if you had lots of criss-crossing
3166 merges and more than two merge bases) as a temporary internal merge
3167 base, and again, those are real objects, but the end result will not end
3168 up pointing to them, so they end up "dangling" in your repository.
3169
3170 Generally, dangling objects aren't anything to worry about. They can
3171 even be very useful: if you screw something up, the dangling objects can
3172 be how you recover your old tree (say, you did a rebase, and realized
3173 that you really didn't want to - you can look at what dangling objects
3174 you have, and decide to reset your head to some old dangling state).
3175
3176 For commits, you can just use:
3177
3178 ------------------------------------------------
3179 $ gitk <dangling-commit-sha-goes-here> --not --all
3180 ------------------------------------------------
3181
3182 This asks for all the history reachable from the given commit but not
3183 from any branch, tag, or other reference. If you decide it's something
3184 you want, you can always create a new reference to it, e.g.,
3185
3186 ------------------------------------------------
3187 $ git branch recovered-branch <dangling-commit-sha-goes-here>
3188 ------------------------------------------------
3189
3190 For blobs and trees, you can't do the same, but you can still examine
3191 them. You can just do
3192
3193 ------------------------------------------------
3194 $ git show <dangling-blob/tree-sha-goes-here>
3195 ------------------------------------------------
3196
3197 to show what the contents of the blob were (or, for a tree, basically
3198 what the "ls" for that directory was), and that may give you some idea
3199 of what the operation was that left that dangling object.
3200
3201 Usually, dangling blobs and trees aren't very interesting. They're
3202 almost always the result of either being a half-way mergebase (the blob
3203 will often even have the conflict markers from a merge in it, if you
3204 have had conflicting merges that you fixed up by hand), or simply
3205 because you interrupted a "git fetch" with ^C or something like that,
3206 leaving _some_ of the new objects in the object database, but just
3207 dangling and useless.
3208
3209 Anyway, once you are sure that you're not interested in any dangling
3210 state, you can just prune all unreachable objects:
3211
3212 ------------------------------------------------
3213 $ git prune
3214 ------------------------------------------------
3215
3216 and they'll be gone. But you should only run "git prune" on a quiescent
3217 repository - it's kind of like doing a filesystem fsck recovery: you
3218 don't want to do that while the filesystem is mounted.
3219
3220 (The same is true of "git-fsck" itself, btw - but since
3221 git-fsck never actually *changes* the repository, it just reports
3222 on what it found, git-fsck itself is never "dangerous" to run.
3223 Running it while somebody is actually changing the repository can cause
3224 confusing and scary messages, but it won't actually do anything bad. In
3225 contrast, running "git prune" while somebody is actively changing the
3226 repository is a *BAD* idea).
3227
3228 [[recovering-from-repository-corruption]]
3229 Recovering from repository corruption
3230 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3231
3232 By design, git treats data trusted to it with caution. However, even in
3233 the absence of bugs in git itself, it is still possible that hardware or
3234 operating system errors could corrupt data.
3235
3236 The first defense against such problems is backups. You can back up a
3237 git directory using clone, or just using cp, tar, or any other backup
3238 mechanism.
3239
3240 As a last resort, you can search for the corrupted objects and attempt
3241 to replace them by hand. Back up your repository before attempting this
3242 in case you corrupt things even more in the process.
3243
3244 We'll assume that the problem is a single missing or corrupted blob,
3245 which is sometimes a solveable problem. (Recovering missing trees and
3246 especially commits is *much* harder).
3247
3248 Before starting, verify that there is corruption, and figure out where
3249 it is with gitlink:git-fsck[1]; this may be time-consuming.
3250
3251 Assume the output looks like this:
3252
3253 ------------------------------------------------
3254 $ git-fsck --full
3255 broken link from tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff8
3256 to blob 4b9458b3786228369c63936db65827de3cc06200
3257 missing blob 4b9458b3786228369c63936db65827de3cc06200
3258 ------------------------------------------------
3259
3260 (Typically there will be some "dangling object" messages too, but they
3261 aren't interesting.)
3262
3263 Now you know that blob 4b9458b3 is missing, and that the tree 2d9263c6
3264 points to it. If you could find just one copy of that missing blob
3265 object, possibly in some other repository, you could move it into
3266 .git/objects/4b/9458b3... and be done. Suppose you can't. You can
3267 still examine the tree that pointed to it with gitlink:git-ls-tree[1],
3268 which might output something like:
3269
3270 ------------------------------------------------
3271 $ git ls-tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff8
3272 100644 blob 8d14531846b95bfa3564b58ccfb7913a034323b8 .gitignore
3273 100644 blob ebf9bf84da0aab5ed944264a5db2a65fe3a3e883 .mailmap
3274 100644 blob ca442d313d86dc67e0a2e5d584b465bd382cbf5c COPYING
3275 ...
3276 100644 blob 4b9458b3786228369c63936db65827de3cc06200 myfile
3277 ...
3278 ------------------------------------------------
3279
3280 So now you know that the missing blob was the data for a file named
3281 "myfile". And chances are you can also identify the directory--let's
3282 say it's in "somedirectory". If you're lucky the missing copy might be
3283 the same as the copy you have checked out in your working tree at
3284 "somedirectory/myfile"; you can test whether that's right with
3285 gitlink:git-hash-object[1]:
3286
3287 ------------------------------------------------
3288 $ git hash-object -w somedirectory/myfile
3289 ------------------------------------------------
3290
3291 which will create and store a blob object with the contents of
3292 somedirectory/myfile, and output the sha1 of that object. if you're
3293 extremely lucky it might be 4b9458b3786228369c63936db65827de3cc06200, in
3294 which case you've guessed right, and the corruption is fixed!
3295
3296 Otherwise, you need more information. How do you tell which version of
3297 the file has been lost?
3298
3299 The easiest way to do this is with:
3300
3301 ------------------------------------------------
3302 $ git log --raw --all --full-history -- somedirectory/myfile
3303 ------------------------------------------------
3304
3305 Because you're asking for raw output, you'll now get something like
3306
3307 ------------------------------------------------
3308 commit abc
3309 Author:
3310 Date:
3311 ...
3312 :100644 100644 4b9458b... newsha... M somedirectory/myfile
3313
3314
3315 commit xyz
3316 Author:
3317 Date:
3318
3319 ...
3320 :100644 100644 oldsha... 4b9458b... M somedirectory/myfile
3321 ------------------------------------------------
3322
3323 This tells you that the immediately preceding version of the file was
3324 "newsha", and that the immediately following version was "oldsha".
3325 You also know the commit messages that went with the change from oldsha
3326 to 4b9458b and with the change from 4b9458b to newsha.
3327
3328 If you've been committing small enough changes, you may now have a good
3329 shot at reconstructing the contents of the in-between state 4b9458b.
3330
3331 If you can do that, you can now recreate the missing object with
3332
3333 ------------------------------------------------
3334 $ git hash-object -w <recreated-file>
3335 ------------------------------------------------
3336
3337 and your repository is good again!
3338
3339 (Btw, you could have ignored the fsck, and started with doing a
3340
3341 ------------------------------------------------
3342 $ git log --raw --all
3343 ------------------------------------------------
3344
3345 and just looked for the sha of the missing object (4b9458b..) in that
3346 whole thing. It's up to you - git does *have* a lot of information, it is
3347 just missing one particular blob version.
3348
3349 [[the-index]]
3350 The index
3351 -----------
3352
3353 The index is a binary file (generally kept in .git/index) containing a
3354 sorted list of path names, each with permissions and the SHA1 of a blob
3355 object; gitlink:git-ls-files[1] can show you the contents of the index:
3356
3357 -------------------------------------------------
3358 $ git ls-files --stage
3359 100644 63c918c667fa005ff12ad89437f2fdc80926e21c 0 .gitignore
3360 100644 5529b198e8d14decbe4ad99db3f7fb632de0439d 0 .mailmap
3361 100644 6ff87c4664981e4397625791c8ea3bbb5f2279a3 0 COPYING
3362 100644 a37b2152bd26be2c2289e1f57a292534a51a93c7 0 Documentation/.gitignore
3363 100644 fbefe9a45b00a54b58d94d06eca48b03d40a50e0 0 Documentation/Makefile
3364 ...
3365 100644 2511aef8d89ab52be5ec6a5e46236b4b6bcd07ea 0 xdiff/xtypes.h
3366 100644 2ade97b2574a9f77e7ae4002a4e07a6a38e46d07 0 xdiff/xutils.c
3367 100644 d5de8292e05e7c36c4b68857c1cf9855e3d2f70a 0 xdiff/xutils.h
3368 -------------------------------------------------
3369
3370 Note that in older documentation you may see the index called the
3371 "current directory cache" or just the "cache". It has three important
3372 properties:
3373
3374 1. The index contains all the information necessary to generate a single
3375 (uniquely determined) tree object.
3376 +
3377 For example, running gitlink:git-commit[1] generates this tree object
3378 from the index, stores it in the object database, and uses it as the
3379 tree object associated with the new commit.
3380
3381 2. The index enables fast comparisons between the tree object it defines
3382 and the working tree.
3383 +
3384 It does this by storing some additional data for each entry (such as
3385 the last modified time). This data is not displayed above, and is not
3386 stored in the created tree object, but it can be used to determine
3387 quickly which files in the working directory differ from what was
3388 stored in the index, and thus save git from having to read all of the
3389 data from such files to look for changes.
3390
3391 3. It can efficiently represent information about merge conflicts
3392 between different tree objects, allowing each pathname to be
3393 associated with sufficient information about the trees involved that
3394 you can create a three-way merge between them.
3395 +
3396 We saw in <<conflict-resolution>> that during a merge the index can
3397 store multiple versions of a single file (called "stages"). The third
3398 column in the gitlink:git-ls-files[1] output above is the stage
3399 number, and will take on values other than 0 for files with merge
3400 conflicts.
3401
3402 The index is thus a sort of temporary staging area, which is filled with
3403 a tree which you are in the process of working on.
3404
3405 If you blow the index away entirely, you generally haven't lost any
3406 information as long as you have the name of the tree that it described.
3407
3408 [[submodules]]
3409 Submodules
3410 ==========
3411
3412 Large projects are often composed of smaller, self-contained modules. For
3413 example, an embedded Linux distribution's source tree would include every
3414 piece of software in the distribution with some local modifications; a movie
3415 player might need to build against a specific, known-working version of a
3416 decompression library; several independent programs might all share the same
3417 build scripts.
3418
3419 With centralized revision control systems this is often accomplished by
3420 including every module in one single repository. Developers can check out
3421 all modules or only the modules they need to work with. They can even modify
3422 files across several modules in a single commit while moving things around
3423 or updating APIs and translations.
3424
3425 Git does not allow partial checkouts, so duplicating this approach in Git
3426 would force developers to keep a local copy of modules they are not
3427 interested in touching. Commits in an enormous checkout would be slower
3428 than you'd expect as Git would have to scan every directory for changes.
3429 If modules have a lot of local history, clones would take forever.
3430
3431 On the plus side, distributed revision control systems can much better
3432 integrate with external sources. In a centralized model, a single arbitrary
3433 snapshot of the external project is exported from its own revision control
3434 and then imported into the local revision control on a vendor branch. All
3435 the history is hidden. With distributed revision control you can clone the
3436 entire external history and much more easily follow development and re-merge
3437 local changes.
3438
3439 Git's submodule support allows a repository to contain, as a subdirectory, a
3440 checkout of an external project. Submodules maintain their own identity;
3441 the submodule support just stores the submodule repository location and
3442 commit ID, so other developers who clone the containing project
3443 ("superproject") can easily clone all the submodules at the same revision.
3444 Partial checkouts of the superproject are possible: you can tell Git to
3445 clone none, some or all of the submodules.
3446
3447 The gitlink:git-submodule[1] command is available since Git 1.5.3. Users
3448 with Git 1.5.2 can look up the submodule commits in the repository and
3449 manually check them out; earlier versions won't recognize the submodules at
3450 all.
3451
3452 To see how submodule support works, create (for example) four example
3453 repositories that can be used later as a submodule:
3454
3455 -------------------------------------------------
3456 $ mkdir ~/git
3457 $ cd ~/git
3458 $ for i in a b c d
3459 do
3460 mkdir $i
3461 cd $i
3462 git init
3463 echo "module $i" > $i.txt
3464 git add $i.txt
3465 git commit -m "Initial commit, submodule $i"
3466 cd ..
3467 done
3468 -------------------------------------------------
3469
3470 Now create the superproject and add all the submodules:
3471
3472 -------------------------------------------------
3473 $ mkdir super
3474 $ cd super
3475 $ git init
3476 $ for i in a b c d
3477 do
3478 git submodule add ~/git/$i
3479 done
3480 -------------------------------------------------
3481
3482 NOTE: Do not use local URLs here if you plan to publish your superproject!
3483
3484 See what files `git submodule` created:
3485
3486 -------------------------------------------------
3487 $ ls -a
3488 . .. .git .gitmodules a b c d
3489 -------------------------------------------------
3490
3491 The `git submodule add` command does a couple of things:
3492
3493 - It clones the submodule under the current directory and by default checks out
3494 the master branch.
3495 - It adds the submodule's clone path to the gitlink:gitmodules[5] file and
3496 adds this file to the index, ready to be committed.
3497 - It adds the submodule's current commit ID to the index, ready to be
3498 committed.
3499
3500 Commit the superproject:
3501
3502 -------------------------------------------------
3503 $ git commit -m "Add submodules a, b, c and d."
3504 -------------------------------------------------
3505
3506 Now clone the superproject:
3507
3508 -------------------------------------------------
3509 $ cd ..
3510 $ git clone super cloned
3511 $ cd cloned
3512 -------------------------------------------------
3513
3514 The submodule directories are there, but they're empty:
3515
3516 -------------------------------------------------
3517 $ ls -a a
3518 . ..
3519 $ git submodule status
3520 -d266b9873ad50488163457f025db7cdd9683d88b a
3521 -e81d457da15309b4fef4249aba9b50187999670d b
3522 -c1536a972b9affea0f16e0680ba87332dc059146 c
3523 -d96249ff5d57de5de093e6baff9e0aafa5276a74 d
3524 -------------------------------------------------
3525
3526 NOTE: The commit object names shown above would be different for you, but they
3527 should match the HEAD commit object names of your repositories. You can check
3528 it by running `git ls-remote ../a`.
3529
3530 Pulling down the submodules is a two-step process. First run `git submodule
3531 init` to add the submodule repository URLs to `.git/config`:
3532
3533 -------------------------------------------------
3534 $ git submodule init
3535 -------------------------------------------------
3536
3537 Now use `git submodule update` to clone the repositories and check out the
3538 commits specified in the superproject:
3539
3540 -------------------------------------------------
3541 $ git submodule update
3542 $ cd a
3543 $ ls -a
3544 . .. .git a.txt
3545 -------------------------------------------------
3546
3547 One major difference between `git submodule update` and `git submodule add` is
3548 that `git submodule update` checks out a specific commit, rather than the tip
3549 of a branch. It's like checking out a tag: the head is detached, so you're not
3550 working on a branch.
3551
3552 -------------------------------------------------
3553 $ git branch
3554 * (no branch)
3555 master
3556 -------------------------------------------------
3557
3558 If you want to make a change within a submodule and you have a detached head,
3559 then you should create or checkout a branch, make your changes, publish the
3560 change within the submodule, and then update the superproject to reference the
3561 new commit:
3562
3563 -------------------------------------------------
3564 $ git checkout master
3565 -------------------------------------------------
3566
3567 or
3568
3569 -------------------------------------------------
3570 $ git checkout -b fix-up
3571 -------------------------------------------------
3572
3573 then
3574
3575 -------------------------------------------------
3576 $ echo "adding a line again" >> a.txt
3577 $ git commit -a -m "Updated the submodule from within the superproject."
3578 $ git push
3579 $ cd ..
3580 $ git diff
3581 diff --git a/a b/a
3582 index d266b98..261dfac 160000
3583 --- a/a
3584 +++ b/a
3585 @@ -1 +1 @@
3586 -Subproject commit d266b9873ad50488163457f025db7cdd9683d88b
3587 +Subproject commit 261dfac35cb99d380eb966e102c1197139f7fa24
3588 $ git add a
3589 $ git commit -m "Updated submodule a."
3590 $ git push
3591 -------------------------------------------------
3592
3593 You have to run `git submodule update` after `git pull` if you want to update
3594 submodules, too.
3595
3596 Pitfalls with submodules
3597 ------------------------
3598
3599 Always publish the submodule change before publishing the change to the
3600 superproject that references it. If you forget to publish the submodule change,
3601 others won't be able to clone the repository:
3602
3603 -------------------------------------------------
3604 $ cd ~/git/super/a
3605 $ echo i added another line to this file >> a.txt
3606 $ git commit -a -m "doing it wrong this time"
3607 $ cd ..
3608 $ git add a
3609 $ git commit -m "Updated submodule a again."
3610 $ git push
3611 $ cd ~/git/cloned
3612 $ git pull
3613 $ git submodule update
3614 error: pathspec '261dfac35cb99d380eb966e102c1197139f7fa24' did not match any file(s) known to git.
3615 Did you forget to 'git add'?
3616 Unable to checkout '261dfac35cb99d380eb966e102c1197139f7fa24' in submodule path 'a'
3617 -------------------------------------------------
3618
3619 You also should not rewind branches in a submodule beyond commits that were
3620 ever recorded in any superproject.
3621
3622 It's not safe to run `git submodule update` if you've made and committed
3623 changes within a submodule without checking out a branch first. They will be
3624 silently overwritten:
3625
3626 -------------------------------------------------
3627 $ cat a.txt
3628 module a
3629 $ echo line added from private2 >> a.txt
3630 $ git commit -a -m "line added inside private2"
3631 $ cd ..
3632 $ git submodule update
3633 Submodule path 'a': checked out 'd266b9873ad50488163457f025db7cdd9683d88b'
3634 $ cd a
3635 $ cat a.txt
3636 module a
3637 -------------------------------------------------
3638
3639 NOTE: The changes are still visible in the submodule's reflog.
3640
3641 This is not the case if you did not commit your changes.
3642
3643 [[low-level-operations]]
3644 Low-level git operations
3645 ========================
3646
3647 Many of the higher-level commands were originally implemented as shell
3648 scripts using a smaller core of low-level git commands. These can still
3649 be useful when doing unusual things with git, or just as a way to
3650 understand its inner workings.
3651
3652 [[object-manipulation]]
3653 Object access and manipulation
3654 ------------------------------
3655
3656 The gitlink:git-cat-file[1] command can show the contents of any object,
3657 though the higher-level gitlink:git-show[1] is usually more useful.
3658
3659 The gitlink:git-commit-tree[1] command allows constructing commits with
3660 arbitrary parents and trees.
3661
3662 A tree can be created with gitlink:git-write-tree[1] and its data can be
3663 accessed by gitlink:git-ls-tree[1]. Two trees can be compared with
3664 gitlink:git-diff-tree[1].
3665
3666 A tag is created with gitlink:git-mktag[1], and the signature can be
3667 verified by gitlink:git-verify-tag[1], though it is normally simpler to
3668 use gitlink:git-tag[1] for both.
3669
3670 [[the-workflow]]
3671 The Workflow
3672 ------------
3673
3674 High-level operations such as gitlink:git-commit[1],
3675 gitlink:git-checkout[1] and git-reset[1] work by moving data between the
3676 working tree, the index, and the object database. Git provides
3677 low-level operations which perform each of these steps individually.
3678
3679 Generally, all "git" operations work on the index file. Some operations
3680 work *purely* on the index file (showing the current state of the
3681 index), but most operations move data between the index file and either
3682 the database or the working directory. Thus there are four main
3683 combinations:
3684
3685 [[working-directory-to-index]]
3686 working directory -> index
3687 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3688
3689 The gitlink:git-update-index[1] command updates the index with
3690 information from the working directory. You generally update the
3691 index information by just specifying the filename you want to update,
3692 like so:
3693
3694 -------------------------------------------------
3695 $ git update-index filename
3696 -------------------------------------------------
3697
3698 but to avoid common mistakes with filename globbing etc, the command
3699 will not normally add totally new entries or remove old entries,
3700 i.e. it will normally just update existing cache entries.
3701
3702 To tell git that yes, you really do realize that certain files no
3703 longer exist, or that new files should be added, you
3704 should use the `--remove` and `--add` flags respectively.
3705
3706 NOTE! A `--remove` flag does 'not' mean that subsequent filenames will
3707 necessarily be removed: if the files still exist in your directory
3708 structure, the index will be updated with their new status, not
3709 removed. The only thing `--remove` means is that update-cache will be
3710 considering a removed file to be a valid thing, and if the file really
3711 does not exist any more, it will update the index accordingly.
3712
3713 As a special case, you can also do `git-update-index --refresh`, which
3714 will refresh the "stat" information of each index to match the current
3715 stat information. It will 'not' update the object status itself, and
3716 it will only update the fields that are used to quickly test whether
3717 an object still matches its old backing store object.
3718
3719 The previously introduced gitlink:git-add[1] is just a wrapper for
3720 gitlink:git-update-index[1].
3721
3722 [[index-to-object-database]]
3723 index -> object database
3724 ~~~~~~~~~~~~~~~~~~~~~~~~
3725
3726 You write your current index file to a "tree" object with the program
3727
3728 -------------------------------------------------
3729 $ git write-tree
3730 -------------------------------------------------
3731
3732 that doesn't come with any options - it will just write out the
3733 current index into the set of tree objects that describe that state,
3734 and it will return the name of the resulting top-level tree. You can
3735 use that tree to re-generate the index at any time by going in the
3736 other direction:
3737
3738 [[object-database-to-index]]
3739 object database -> index
3740 ~~~~~~~~~~~~~~~~~~~~~~~~
3741
3742 You read a "tree" file from the object database, and use that to
3743 populate (and overwrite - don't do this if your index contains any
3744 unsaved state that you might want to restore later!) your current
3745 index. Normal operation is just
3746
3747 -------------------------------------------------
3748 $ git-read-tree <sha1 of tree>
3749 -------------------------------------------------
3750
3751 and your index file will now be equivalent to the tree that you saved
3752 earlier. However, that is only your 'index' file: your working
3753 directory contents have not been modified.
3754
3755 [[index-to-working-directory]]
3756 index -> working directory
3757 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3758
3759 You update your working directory from the index by "checking out"
3760 files. This is not a very common operation, since normally you'd just
3761 keep your files updated, and rather than write to your working
3762 directory, you'd tell the index files about the changes in your
3763 working directory (i.e. `git-update-index`).
3764
3765 However, if you decide to jump to a new version, or check out somebody
3766 else's version, or just restore a previous tree, you'd populate your
3767 index file with read-tree, and then you need to check out the result
3768 with
3769
3770 -------------------------------------------------
3771 $ git-checkout-index filename
3772 -------------------------------------------------
3773
3774 or, if you want to check out all of the index, use `-a`.
3775
3776 NOTE! git-checkout-index normally refuses to overwrite old files, so
3777 if you have an old version of the tree already checked out, you will
3778 need to use the "-f" flag ('before' the "-a" flag or the filename) to
3779 'force' the checkout.
3780
3781
3782 Finally, there are a few odds and ends which are not purely moving
3783 from one representation to the other:
3784
3785 [[tying-it-all-together]]
3786 Tying it all together
3787 ~~~~~~~~~~~~~~~~~~~~~
3788
3789 To commit a tree you have instantiated with "git-write-tree", you'd
3790 create a "commit" object that refers to that tree and the history
3791 behind it - most notably the "parent" commits that preceded it in
3792 history.
3793
3794 Normally a "commit" has one parent: the previous state of the tree
3795 before a certain change was made. However, sometimes it can have two
3796 or more parent commits, in which case we call it a "merge", due to the
3797 fact that such a commit brings together ("merges") two or more
3798 previous states represented by other commits.
3799
3800 In other words, while a "tree" represents a particular directory state
3801 of a working directory, a "commit" represents that state in "time",
3802 and explains how we got there.
3803
3804 You create a commit object by giving it the tree that describes the
3805 state at the time of the commit, and a list of parents:
3806
3807 -------------------------------------------------
3808 $ git-commit-tree <tree> -p <parent> [-p <parent2> ..]
3809 -------------------------------------------------
3810
3811 and then giving the reason for the commit on stdin (either through
3812 redirection from a pipe or file, or by just typing it at the tty).
3813
3814 git-commit-tree will return the name of the object that represents
3815 that commit, and you should save it away for later use. Normally,
3816 you'd commit a new `HEAD` state, and while git doesn't care where you
3817 save the note about that state, in practice we tend to just write the
3818 result to the file pointed at by `.git/HEAD`, so that we can always see
3819 what the last committed state was.
3820
3821 Here is an ASCII art by Jon Loeliger that illustrates how
3822 various pieces fit together.
3823
3824 ------------
3825
3826 commit-tree
3827 commit obj
3828 +----+
3829 | |
3830 | |
3831 V V
3832 +-----------+
3833 | Object DB |
3834 | Backing |
3835 | Store |
3836 +-----------+
3837 ^
3838 write-tree | |
3839 tree obj | |
3840 | | read-tree
3841 | | tree obj
3842 V
3843 +-----------+
3844 | Index |
3845 | "cache" |
3846 +-----------+
3847 update-index ^
3848 blob obj | |
3849 | |
3850 checkout-index -u | | checkout-index
3851 stat | | blob obj
3852 V
3853 +-----------+
3854 | Working |
3855 | Directory |
3856 +-----------+
3857
3858 ------------
3859
3860
3861 [[examining-the-data]]
3862 Examining the data
3863 ------------------
3864
3865 You can examine the data represented in the object database and the
3866 index with various helper tools. For every object, you can use
3867 gitlink:git-cat-file[1] to examine details about the
3868 object:
3869
3870 -------------------------------------------------
3871 $ git-cat-file -t <objectname>
3872 -------------------------------------------------
3873
3874 shows the type of the object, and once you have the type (which is
3875 usually implicit in where you find the object), you can use
3876
3877 -------------------------------------------------
3878 $ git-cat-file blob|tree|commit|tag <objectname>
3879 -------------------------------------------------
3880
3881 to show its contents. NOTE! Trees have binary content, and as a result
3882 there is a special helper for showing that content, called
3883 `git-ls-tree`, which turns the binary content into a more easily
3884 readable form.
3885
3886 It's especially instructive to look at "commit" objects, since those
3887 tend to be small and fairly self-explanatory. In particular, if you
3888 follow the convention of having the top commit name in `.git/HEAD`,
3889 you can do
3890
3891 -------------------------------------------------
3892 $ git-cat-file commit HEAD
3893 -------------------------------------------------
3894
3895 to see what the top commit was.
3896
3897 [[merging-multiple-trees]]
3898 Merging multiple trees
3899 ----------------------
3900
3901 Git helps you do a three-way merge, which you can expand to n-way by
3902 repeating the merge procedure arbitrary times until you finally
3903 "commit" the state. The normal situation is that you'd only do one
3904 three-way merge (two parents), and commit it, but if you like to, you
3905 can do multiple parents in one go.
3906
3907 To do a three-way merge, you need the two sets of "commit" objects
3908 that you want to merge, use those to find the closest common parent (a
3909 third "commit" object), and then use those commit objects to find the
3910 state of the directory ("tree" object) at these points.
3911
3912 To get the "base" for the merge, you first look up the common parent
3913 of two commits with
3914
3915 -------------------------------------------------
3916 $ git-merge-base <commit1> <commit2>
3917 -------------------------------------------------
3918
3919 which will return you the commit they are both based on. You should
3920 now look up the "tree" objects of those commits, which you can easily
3921 do with (for example)
3922
3923 -------------------------------------------------
3924 $ git-cat-file commit <commitname> | head -1
3925 -------------------------------------------------
3926
3927 since the tree object information is always the first line in a commit
3928 object.
3929
3930 Once you know the three trees you are going to merge (the one "original"
3931 tree, aka the common tree, and the two "result" trees, aka the branches
3932 you want to merge), you do a "merge" read into the index. This will
3933 complain if it has to throw away your old index contents, so you should
3934 make sure that you've committed those - in fact you would normally
3935 always do a merge against your last commit (which should thus match what
3936 you have in your current index anyway).
3937
3938 To do the merge, do
3939
3940 -------------------------------------------------
3941 $ git-read-tree -m -u <origtree> <yourtree> <targettree>
3942 -------------------------------------------------
3943
3944 which will do all trivial merge operations for you directly in the
3945 index file, and you can just write the result out with
3946 `git-write-tree`.
3947
3948
3949 [[merging-multiple-trees-2]]
3950 Merging multiple trees, continued
3951 ---------------------------------
3952
3953 Sadly, many merges aren't trivial. If there are files that have
3954 been added.moved or removed, or if both branches have modified the
3955 same file, you will be left with an index tree that contains "merge
3956 entries" in it. Such an index tree can 'NOT' be written out to a tree
3957 object, and you will have to resolve any such merge clashes using
3958 other tools before you can write out the result.
3959
3960 You can examine such index state with `git-ls-files --unmerged`
3961 command. An example:
3962
3963 ------------------------------------------------
3964 $ git-read-tree -m $orig HEAD $target
3965 $ git-ls-files --unmerged
3966 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello.c
3967 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello.c
3968 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello.c
3969 ------------------------------------------------
3970
3971 Each line of the `git-ls-files --unmerged` output begins with
3972 the blob mode bits, blob SHA1, 'stage number', and the
3973 filename. The 'stage number' is git's way to say which tree it
3974 came from: stage 1 corresponds to `$orig` tree, stage 2 `HEAD`
3975 tree, and stage3 `$target` tree.
3976
3977 Earlier we said that trivial merges are done inside
3978 `git-read-tree -m`. For example, if the file did not change
3979 from `$orig` to `HEAD` nor `$target`, or if the file changed
3980 from `$orig` to `HEAD` and `$orig` to `$target` the same way,
3981 obviously the final outcome is what is in `HEAD`. What the
3982 above example shows is that file `hello.c` was changed from
3983 `$orig` to `HEAD` and `$orig` to `$target` in a different way.
3984 You could resolve this by running your favorite 3-way merge
3985 program, e.g. `diff3`, `merge`, or git's own merge-file, on
3986 the blob objects from these three stages yourself, like this:
3987
3988 ------------------------------------------------
3989 $ git-cat-file blob 263414f... >hello.c~1
3990 $ git-cat-file blob 06fa6a2... >hello.c~2
3991 $ git-cat-file blob cc44c73... >hello.c~3
3992 $ git merge-file hello.c~2 hello.c~1 hello.c~3
3993 ------------------------------------------------
3994
3995 This would leave the merge result in `hello.c~2` file, along
3996 with conflict markers if there are conflicts. After verifying
3997 the merge result makes sense, you can tell git what the final
3998 merge result for this file is by:
3999
4000 -------------------------------------------------
4001 $ mv -f hello.c~2 hello.c
4002 $ git-update-index hello.c
4003 -------------------------------------------------
4004
4005 When a path is in unmerged state, running `git-update-index` for
4006 that path tells git to mark the path resolved.
4007
4008 The above is the description of a git merge at the lowest level,
4009 to help you understand what conceptually happens under the hood.
4010 In practice, nobody, not even git itself, uses three `git-cat-file`
4011 for this. There is `git-merge-index` program that extracts the
4012 stages to temporary files and calls a "merge" script on it:
4013
4014 -------------------------------------------------
4015 $ git-merge-index git-merge-one-file hello.c
4016 -------------------------------------------------
4017
4018 and that is what higher level `git merge -s resolve` is implemented with.
4019
4020 [[hacking-git]]
4021 Hacking git
4022 ===========
4023
4024 This chapter covers internal details of the git implementation which
4025 probably only git developers need to understand.
4026
4027 [[object-details]]
4028 Object storage format
4029 ---------------------
4030
4031 All objects have a statically determined "type" which identifies the
4032 format of the object (i.e. how it is used, and how it can refer to other
4033 objects). There are currently four different object types: "blob",
4034 "tree", "commit", and "tag".
4035
4036 Regardless of object type, all objects share the following
4037 characteristics: they are all deflated with zlib, and have a header
4038 that not only specifies their type, but also provides size information
4039 about the data in the object. It's worth noting that the SHA1 hash
4040 that is used to name the object is the hash of the original data
4041 plus this header, so `sha1sum` 'file' does not match the object name
4042 for 'file'.
4043 (Historical note: in the dawn of the age of git the hash
4044 was the sha1 of the 'compressed' object.)
4045
4046 As a result, the general consistency of an object can always be tested
4047 independently of the contents or the type of the object: all objects can
4048 be validated by verifying that (a) their hashes match the content of the
4049 file and (b) the object successfully inflates to a stream of bytes that
4050 forms a sequence of <ascii type without space> {plus} <space> {plus} <ascii decimal
4051 size> {plus} <byte\0> {plus} <binary object data>.
4052
4053 The structured objects can further have their structure and
4054 connectivity to other objects verified. This is generally done with
4055 the `git-fsck` program, which generates a full dependency graph
4056 of all objects, and verifies their internal consistency (in addition
4057 to just verifying their superficial consistency through the hash).
4058
4059 [[birdview-on-the-source-code]]
4060 A birds-eye view of Git's source code
4061 -------------------------------------
4062
4063 It is not always easy for new developers to find their way through Git's
4064 source code. This section gives you a little guidance to show where to
4065 start.
4066
4067 A good place to start is with the contents of the initial commit, with:
4068
4069 ----------------------------------------------------
4070 $ git checkout e83c5163
4071 ----------------------------------------------------
4072
4073 The initial revision lays the foundation for almost everything git has
4074 today, but is small enough to read in one sitting.
4075
4076 Note that terminology has changed since that revision. For example, the
4077 README in that revision uses the word "changeset" to describe what we
4078 now call a <<def_commit_object,commit>>.
4079
4080 Also, we do not call it "cache" any more, but "index", however, the
4081 file is still called `cache.h`. Remark: Not much reason to change it now,
4082 especially since there is no good single name for it anyway, because it is
4083 basically _the_ header file which is included by _all_ of Git's C sources.
4084
4085 If you grasp the ideas in that initial commit, you should check out a
4086 more recent version and skim `cache.h`, `object.h` and `commit.h`.
4087
4088 In the early days, Git (in the tradition of UNIX) was a bunch of programs
4089 which were extremely simple, and which you used in scripts, piping the
4090 output of one into another. This turned out to be good for initial
4091 development, since it was easier to test new things. However, recently
4092 many of these parts have become builtins, and some of the core has been
4093 "libified", i.e. put into libgit.a for performance, portability reasons,
4094 and to avoid code duplication.
4095
4096 By now, you know what the index is (and find the corresponding data
4097 structures in `cache.h`), and that there are just a couple of object types
4098 (blobs, trees, commits and tags) which inherit their common structure from
4099 `struct object`, which is their first member (and thus, you can cast e.g.
4100 `(struct object *)commit` to achieve the _same_ as `&commit->object`, i.e.
4101 get at the object name and flags).
4102
4103 Now is a good point to take a break to let this information sink in.
4104
4105 Next step: get familiar with the object naming. Read <<naming-commits>>.
4106 There are quite a few ways to name an object (and not only revisions!).
4107 All of these are handled in `sha1_name.c`. Just have a quick look at
4108 the function `get_sha1()`. A lot of the special handling is done by
4109 functions like `get_sha1_basic()` or the likes.
4110
4111 This is just to get you into the groove for the most libified part of Git:
4112 the revision walker.
4113
4114 Basically, the initial version of `git log` was a shell script:
4115
4116 ----------------------------------------------------------------
4117 $ git-rev-list --pretty $(git-rev-parse --default HEAD "$@") | \
4118 LESS=-S ${PAGER:-less}
4119 ----------------------------------------------------------------
4120
4121 What does this mean?
4122
4123 `git-rev-list` is the original version of the revision walker, which
4124 _always_ printed a list of revisions to stdout. It is still functional,
4125 and needs to, since most new Git programs start out as scripts using
4126 `git-rev-list`.
4127
4128 `git-rev-parse` is not as important any more; it was only used to filter out
4129 options that were relevant for the different plumbing commands that were
4130 called by the script.
4131
4132 Most of what `git-rev-list` did is contained in `revision.c` and
4133 `revision.h`. It wraps the options in a struct named `rev_info`, which
4134 controls how and what revisions are walked, and more.
4135
4136 The original job of `git-rev-parse` is now taken by the function
4137 `setup_revisions()`, which parses the revisions and the common command line
4138 options for the revision walker. This information is stored in the struct
4139 `rev_info` for later consumption. You can do your own command line option
4140 parsing after calling `setup_revisions()`. After that, you have to call
4141 `prepare_revision_walk()` for initialization, and then you can get the
4142 commits one by one with the function `get_revision()`.
4143
4144 If you are interested in more details of the revision walking process,
4145 just have a look at the first implementation of `cmd_log()`; call
4146 `git-show v1.3.0~155^2~4` and scroll down to that function (note that you
4147 no longer need to call `setup_pager()` directly).
4148
4149 Nowadays, `git log` is a builtin, which means that it is _contained_ in the
4150 command `git`. The source side of a builtin is
4151
4152 - a function called `cmd_<bla>`, typically defined in `builtin-<bla>.c`,
4153 and declared in `builtin.h`,
4154
4155 - an entry in the `commands[]` array in `git.c`, and
4156
4157 - an entry in `BUILTIN_OBJECTS` in the `Makefile`.
4158
4159 Sometimes, more than one builtin is contained in one source file. For
4160 example, `cmd_whatchanged()` and `cmd_log()` both reside in `builtin-log.c`,
4161 since they share quite a bit of code. In that case, the commands which are
4162 _not_ named like the `.c` file in which they live have to be listed in
4163 `BUILT_INS` in the `Makefile`.
4164
4165 `git log` looks more complicated in C than it does in the original script,
4166 but that allows for a much greater flexibility and performance.
4167
4168 Here again it is a good point to take a pause.
4169
4170 Lesson three is: study the code. Really, it is the best way to learn about
4171 the organization of Git (after you know the basic concepts).
4172
4173 So, think about something which you are interested in, say, "how can I
4174 access a blob just knowing the object name of it?". The first step is to
4175 find a Git command with which you can do it. In this example, it is either
4176 `git show` or `git cat-file`.
4177
4178 For the sake of clarity, let's stay with `git cat-file`, because it
4179
4180 - is plumbing, and
4181
4182 - was around even in the initial commit (it literally went only through
4183 some 20 revisions as `cat-file.c`, was renamed to `builtin-cat-file.c`
4184 when made a builtin, and then saw less than 10 versions).
4185
4186 So, look into `builtin-cat-file.c`, search for `cmd_cat_file()` and look what
4187 it does.
4188
4189 ------------------------------------------------------------------
4190 git_config(git_default_config);
4191 if (argc != 3)
4192 usage("git-cat-file [-t|-s|-e|-p|<type>] <sha1>");
4193 if (get_sha1(argv[2], sha1))
4194 die("Not a valid object name %s", argv[2]);
4195 ------------------------------------------------------------------
4196
4197 Let's skip over the obvious details; the only really interesting part
4198 here is the call to `get_sha1()`. It tries to interpret `argv[2]` as an
4199 object name, and if it refers to an object which is present in the current
4200 repository, it writes the resulting SHA-1 into the variable `sha1`.
4201
4202 Two things are interesting here:
4203
4204 - `get_sha1()` returns 0 on _success_. This might surprise some new
4205 Git hackers, but there is a long tradition in UNIX to return different
4206 negative numbers in case of different errors -- and 0 on success.
4207
4208 - the variable `sha1` in the function signature of `get_sha1()` is `unsigned
4209 char \*`, but is actually expected to be a pointer to `unsigned
4210 char[20]`. This variable will contain the 160-bit SHA-1 of the given
4211 commit. Note that whenever a SHA-1 is passed as `unsigned char \*`, it
4212 is the binary representation, as opposed to the ASCII representation in
4213 hex characters, which is passed as `char *`.
4214
4215 You will see both of these things throughout the code.
4216
4217 Now, for the meat:
4218
4219 -----------------------------------------------------------------------------
4220 case 0:
4221 buf = read_object_with_reference(sha1, argv[1], &size, NULL);
4222 -----------------------------------------------------------------------------
4223
4224 This is how you read a blob (actually, not only a blob, but any type of
4225 object). To know how the function `read_object_with_reference()` actually
4226 works, find the source code for it (something like `git grep
4227 read_object_with | grep ":[a-z]"` in the git repository), and read
4228 the source.
4229
4230 To find out how the result can be used, just read on in `cmd_cat_file()`:
4231
4232 -----------------------------------
4233 write_or_die(1, buf, size);
4234 -----------------------------------
4235
4236 Sometimes, you do not know where to look for a feature. In many such cases,
4237 it helps to search through the output of `git log`, and then `git show` the
4238 corresponding commit.
4239
4240 Example: If you know that there was some test case for `git bundle`, but
4241 do not remember where it was (yes, you _could_ `git grep bundle t/`, but that
4242 does not illustrate the point!):
4243
4244 ------------------------
4245 $ git log --no-merges t/
4246 ------------------------
4247
4248 In the pager (`less`), just search for "bundle", go a few lines back,
4249 and see that it is in commit 18449ab0... Now just copy this object name,
4250 and paste it into the command line
4251
4252 -------------------
4253 $ git show 18449ab0
4254 -------------------
4255
4256 Voila.
4257
4258 Another example: Find out what to do in order to make some script a
4259 builtin:
4260
4261 -------------------------------------------------
4262 $ git log --no-merges --diff-filter=A builtin-*.c
4263 -------------------------------------------------
4264
4265 You see, Git is actually the best tool to find out about the source of Git
4266 itself!
4267
4268 [[glossary]]
4269 include::glossary.txt[]
4270
4271 [[git-quick-start]]
4272 Appendix A: Git Quick Reference
4273 ===============================
4274
4275 This is a quick summary of the major commands; the previous chapters
4276 explain how these work in more detail.
4277
4278 [[quick-creating-a-new-repository]]
4279 Creating a new repository
4280 -------------------------
4281
4282 From a tarball:
4283
4284 -----------------------------------------------
4285 $ tar xzf project.tar.gz
4286 $ cd project
4287 $ git init
4288 Initialized empty Git repository in .git/
4289 $ git add .
4290 $ git commit
4291 -----------------------------------------------
4292
4293 From a remote repository:
4294
4295 -----------------------------------------------
4296 $ git clone git://example.com/pub/project.git
4297 $ cd project
4298 -----------------------------------------------
4299
4300 [[managing-branches]]
4301 Managing branches
4302 -----------------
4303
4304 -----------------------------------------------
4305 $ git branch # list all local branches in this repo
4306 $ git checkout test # switch working directory to branch "test"
4307 $ git branch new # create branch "new" starting at current HEAD
4308 $ git branch -d new # delete branch "new"
4309 -----------------------------------------------
4310
4311 Instead of basing new branch on current HEAD (the default), use:
4312
4313 -----------------------------------------------
4314 $ git branch new test # branch named "test"
4315 $ git branch new v2.6.15 # tag named v2.6.15
4316 $ git branch new HEAD^ # commit before the most recent
4317 $ git branch new HEAD^^ # commit before that
4318 $ git branch new test~10 # ten commits before tip of branch "test"
4319 -----------------------------------------------
4320
4321 Create and switch to a new branch at the same time:
4322
4323 -----------------------------------------------
4324 $ git checkout -b new v2.6.15
4325 -----------------------------------------------
4326
4327 Update and examine branches from the repository you cloned from:
4328
4329 -----------------------------------------------
4330 $ git fetch # update
4331 $ git branch -r # list
4332 origin/master
4333 origin/next
4334 ...
4335 $ git checkout -b masterwork origin/master
4336 -----------------------------------------------
4337
4338 Fetch a branch from a different repository, and give it a new
4339 name in your repository:
4340
4341 -----------------------------------------------
4342 $ git fetch git://example.com/project.git theirbranch:mybranch
4343 $ git fetch git://example.com/project.git v2.6.15:mybranch
4344 -----------------------------------------------
4345
4346 Keep a list of repositories you work with regularly:
4347
4348 -----------------------------------------------
4349 $ git remote add example git://example.com/project.git
4350 $ git remote # list remote repositories
4351 example
4352 origin
4353 $ git remote show example # get details
4354 * remote example
4355 URL: git://example.com/project.git
4356 Tracked remote branches
4357 master next ...
4358 $ git fetch example # update branches from example
4359 $ git branch -r # list all remote branches
4360 -----------------------------------------------
4361
4362
4363 [[exploring-history]]
4364 Exploring history
4365 -----------------
4366
4367 -----------------------------------------------
4368 $ gitk # visualize and browse history
4369 $ git log # list all commits
4370 $ git log src/ # ...modifying src/
4371 $ git log v2.6.15..v2.6.16 # ...in v2.6.16, not in v2.6.15
4372 $ git log master..test # ...in branch test, not in branch master
4373 $ git log test..master # ...in branch master, but not in test
4374 $ git log test...master # ...in one branch, not in both
4375 $ git log -S'foo()' # ...where difference contain "foo()"
4376 $ git log --since="2 weeks ago"
4377 $ git log -p # show patches as well
4378 $ git show # most recent commit
4379 $ git diff v2.6.15..v2.6.16 # diff between two tagged versions
4380 $ git diff v2.6.15..HEAD # diff with current head
4381 $ git grep "foo()" # search working directory for "foo()"
4382 $ git grep v2.6.15 "foo()" # search old tree for "foo()"
4383 $ git show v2.6.15:a.txt # look at old version of a.txt
4384 -----------------------------------------------
4385
4386 Search for regressions:
4387
4388 -----------------------------------------------
4389 $ git bisect start
4390 $ git bisect bad # current version is bad
4391 $ git bisect good v2.6.13-rc2 # last known good revision
4392 Bisecting: 675 revisions left to test after this
4393 # test here, then:
4394 $ git bisect good # if this revision is good, or
4395 $ git bisect bad # if this revision is bad.
4396 # repeat until done.
4397 -----------------------------------------------
4398
4399 [[making-changes]]
4400 Making changes
4401 --------------
4402
4403 Make sure git knows who to blame:
4404
4405 ------------------------------------------------
4406 $ cat >>~/.gitconfig <<\EOF
4407 [user]
4408 name = Your Name Comes Here
4409 email = you@yourdomain.example.com
4410 EOF
4411 ------------------------------------------------
4412
4413 Select file contents to include in the next commit, then make the
4414 commit:
4415
4416 -----------------------------------------------
4417 $ git add a.txt # updated file
4418 $ git add b.txt # new file
4419 $ git rm c.txt # old file
4420 $ git commit
4421 -----------------------------------------------
4422
4423 Or, prepare and create the commit in one step:
4424
4425 -----------------------------------------------
4426 $ git commit d.txt # use latest content only of d.txt
4427 $ git commit -a # use latest content of all tracked files
4428 -----------------------------------------------
4429
4430 [[merging]]
4431 Merging
4432 -------
4433
4434 -----------------------------------------------
4435 $ git merge test # merge branch "test" into the current branch
4436 $ git pull git://example.com/project.git master
4437 # fetch and merge in remote branch
4438 $ git pull . test # equivalent to git merge test
4439 -----------------------------------------------
4440
4441 [[sharing-your-changes]]
4442 Sharing your changes
4443 --------------------
4444
4445 Importing or exporting patches:
4446
4447 -----------------------------------------------
4448 $ git format-patch origin..HEAD # format a patch for each commit
4449 # in HEAD but not in origin
4450 $ git am mbox # import patches from the mailbox "mbox"
4451 -----------------------------------------------
4452
4453 Fetch a branch in a different git repository, then merge into the
4454 current branch:
4455
4456 -----------------------------------------------
4457 $ git pull git://example.com/project.git theirbranch
4458 -----------------------------------------------
4459
4460 Store the fetched branch into a local branch before merging into the
4461 current branch:
4462
4463 -----------------------------------------------
4464 $ git pull git://example.com/project.git theirbranch:mybranch
4465 -----------------------------------------------
4466
4467 After creating commits on a local branch, update the remote
4468 branch with your commits:
4469
4470 -----------------------------------------------
4471 $ git push ssh://example.com/project.git mybranch:theirbranch
4472 -----------------------------------------------
4473
4474 When remote and local branch are both named "test":
4475
4476 -----------------------------------------------
4477 $ git push ssh://example.com/project.git test
4478 -----------------------------------------------
4479
4480 Shortcut version for a frequently used remote repository:
4481
4482 -----------------------------------------------
4483 $ git remote add example ssh://example.com/project.git
4484 $ git push example test
4485 -----------------------------------------------
4486
4487 [[repository-maintenance]]
4488 Repository maintenance
4489 ----------------------
4490
4491 Check for corruption:
4492
4493 -----------------------------------------------
4494 $ git fsck
4495 -----------------------------------------------
4496
4497 Recompress, remove unused cruft:
4498
4499 -----------------------------------------------
4500 $ git gc
4501 -----------------------------------------------
4502
4503
4504 [[todo]]
4505 Appendix B: Notes and todo list for this manual
4506 ===============================================
4507
4508 This is a work in progress.
4509
4510 The basic requirements:
4511
4512 - It must be readable in order, from beginning to end, by someone
4513 intelligent with a basic grasp of the UNIX command line, but without
4514 any special knowledge of git. If necessary, any other prerequisites
4515 should be specifically mentioned as they arise.
4516 - Whenever possible, section headings should clearly describe the task
4517 they explain how to do, in language that requires no more knowledge
4518 than necessary: for example, "importing patches into a project" rather
4519 than "the git-am command"
4520
4521 Think about how to create a clear chapter dependency graph that will
4522 allow people to get to important topics without necessarily reading
4523 everything in between.
4524
4525 Scan Documentation/ for other stuff left out; in particular:
4526
4527 - howto's
4528 - some of technical/?
4529 - hooks
4530 - list of commands in gitlink:git[1]
4531
4532 Scan email archives for other stuff left out
4533
4534 Scan man pages to see if any assume more background than this manual
4535 provides.
4536
4537 Simplify beginning by suggesting disconnected head instead of
4538 temporary branch creation?
4539
4540 Add more good examples. Entire sections of just cookbook examples
4541 might be a good idea; maybe make an "advanced examples" section a
4542 standard end-of-chapter section?
4543
4544 Include cross-references to the glossary, where appropriate.
4545
4546 Document shallow clones? See draft 1.5.0 release notes for some
4547 documentation.
4548
4549 Add a section on working with other version control systems, including
4550 CVS, Subversion, and just imports of series of release tarballs.
4551
4552 More details on gitweb?
4553
4554 Write a chapter on using plumbing and writing scripts.
4555
4556 Alternates, clone -reference, etc.
4557
4558 More on recovery from repository corruption. See:
4559 http://marc.theaimsgroup.com/?l=git&m=117263864820799&w=2
4560 http://marc.theaimsgroup.com/?l=git&m=117147855503798&w=2
4561 http://marc.theaimsgroup.com/?l=git&m=117147855503798&w=2