user-manual: Reorganize the reroll sections, adding 'git rebase -i'
[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, or linkgit:git-help[1] command. For example, for the command
22 "git clone <repo>", you can either use:
23
24 ------------------------------------------------
25 $ man git-clone
26 ------------------------------------------------
27
28 or:
29
30 ------------------------------------------------
31 $ git help clone
32 ------------------------------------------------
33
34 With the latter, you can use the manual viewer of your choice; see
35 linkgit:git-help[1] for more information.
36
37 See also <<git-quick-start>> for a brief overview of git commands,
38 without any explanation.
39
40 Finally, see <<todo>> for ways that you can help make this manual more
41 complete.
42
43
44 [[repositories-and-branches]]
45 Repositories and Branches
46 =========================
47
48 [[how-to-get-a-git-repository]]
49 How to get a git repository
50 ---------------------------
51
52 It will be useful to have a git repository to experiment with as you
53 read this manual.
54
55 The best way to get one is by using the linkgit:git-clone[1] command to
56 download a copy of an existing repository. If you don't already have a
57 project in mind, here are some interesting examples:
58
59 ------------------------------------------------
60 # git itself (approx. 10MB download):
61 $ git clone git://git.kernel.org/pub/scm/git/git.git
62 # the Linux kernel (approx. 150MB download):
63 $ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
64 ------------------------------------------------
65
66 The initial clone may be time-consuming for a large project, but you
67 will only need to clone once.
68
69 The clone command creates a new directory named after the project ("git"
70 or "linux-2.6" in the examples above). After you cd into this
71 directory, you will see that it contains a copy of the project files,
72 called the <<def_working_tree,working tree>>, together with a special
73 top-level directory named ".git", which contains all the information
74 about the history of the project.
75
76 [[how-to-check-out]]
77 How to check out a different version of a project
78 -------------------------------------------------
79
80 Git is best thought of as a tool for storing the history of a collection
81 of files. It stores the history as a compressed collection of
82 interrelated snapshots of the project's contents. In git each such
83 version is called a <<def_commit,commit>>.
84
85 Those snapshots aren't necessarily all arranged in a single line from
86 oldest to newest; instead, work may simultaneously proceed along
87 parallel lines of development, called <<def_branch,branches>>, which may
88 merge and diverge.
89
90 A single git repository can track development on multiple branches. It
91 does this by keeping a list of <<def_head,heads>> which reference the
92 latest commit on each branch; the linkgit:git-branch[1] command shows
93 you the list of branch heads:
94
95 ------------------------------------------------
96 $ git branch
97 * master
98 ------------------------------------------------
99
100 A freshly cloned repository contains a single branch head, by default
101 named "master", with the working directory initialized to the state of
102 the project referred to by that branch head.
103
104 Most projects also use <<def_tag,tags>>. Tags, like heads, are
105 references into the project's history, and can be listed using the
106 linkgit:git-tag[1] command:
107
108 ------------------------------------------------
109 $ git tag -l
110 v2.6.11
111 v2.6.11-tree
112 v2.6.12
113 v2.6.12-rc2
114 v2.6.12-rc3
115 v2.6.12-rc4
116 v2.6.12-rc5
117 v2.6.12-rc6
118 v2.6.13
119 ...
120 ------------------------------------------------
121
122 Tags are expected to always point at the same version of a project,
123 while heads are expected to advance as development progresses.
124
125 Create a new branch head pointing to one of these versions and check it
126 out using linkgit:git-checkout[1]:
127
128 ------------------------------------------------
129 $ git checkout -b new v2.6.13
130 ------------------------------------------------
131
132 The working directory then reflects the contents that the project had
133 when it was tagged v2.6.13, and linkgit:git-branch[1] shows two
134 branches, with an asterisk marking the currently checked-out branch:
135
136 ------------------------------------------------
137 $ git branch
138 master
139 * new
140 ------------------------------------------------
141
142 If you decide that you'd rather see version 2.6.17, you can modify
143 the current branch to point at v2.6.17 instead, with
144
145 ------------------------------------------------
146 $ git reset --hard v2.6.17
147 ------------------------------------------------
148
149 Note that if the current branch head was your only reference to a
150 particular point in history, then resetting that branch may leave you
151 with no way to find the history it used to point to; so use this command
152 carefully.
153
154 [[understanding-commits]]
155 Understanding History: Commits
156 ------------------------------
157
158 Every change in the history of a project is represented by a commit.
159 The linkgit:git-show[1] command shows the most recent commit on the
160 current branch:
161
162 ------------------------------------------------
163 $ git show
164 commit 17cf781661e6d38f737f15f53ab552f1e95960d7
165 Author: Linus Torvalds <torvalds@ppc970.osdl.org.(none)>
166 Date: Tue Apr 19 14:11:06 2005 -0700
167
168 Remove duplicate getenv(DB_ENVIRONMENT) call
169
170 Noted by Tony Luck.
171
172 diff --git a/init-db.c b/init-db.c
173 index 65898fa..b002dc6 100644
174 --- a/init-db.c
175 +++ b/init-db.c
176 @@ -7,7 +7,7 @@
177
178 int main(int argc, char **argv)
179 {
180 - char *sha1_dir = getenv(DB_ENVIRONMENT), *path;
181 + char *sha1_dir, *path;
182 int len, i;
183
184 if (mkdir(".git", 0755) < 0) {
185 ------------------------------------------------
186
187 As you can see, a commit shows who made the latest change, what they
188 did, and why.
189
190 Every commit has a 40-hexdigit id, sometimes called the "object name" or the
191 "SHA-1 id", shown on the first line of the "git show" output. You can usually
192 refer to a commit by a shorter name, such as a tag or a branch name, but this
193 longer name can also be useful. Most importantly, it is a globally unique
194 name for this commit: so if you tell somebody else the object name (for
195 example in email), then you are guaranteed that name will refer to the same
196 commit in their repository that it does in yours (assuming their repository
197 has that commit at all). Since the object name is computed as a hash over the
198 contents of the commit, you are guaranteed that the commit can never change
199 without its name also changing.
200
201 In fact, in <<git-concepts>> we shall see that everything stored in git
202 history, including file data and directory contents, is stored in an object
203 with a name that is a hash of its contents.
204
205 [[understanding-reachability]]
206 Understanding history: commits, parents, and reachability
207 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
208
209 Every commit (except the very first commit in a project) also has a
210 parent commit which shows what happened before this commit.
211 Following the chain of parents will eventually take you back to the
212 beginning of the project.
213
214 However, the commits do not form a simple list; git allows lines of
215 development to diverge and then reconverge, and the point where two
216 lines of development reconverge is called a "merge". The commit
217 representing a merge can therefore have more than one parent, with
218 each parent representing the most recent commit on one of the lines
219 of development leading to that point.
220
221 The best way to see how this works is using the linkgit:gitk[1]
222 command; running gitk now on a git repository and looking for merge
223 commits will help understand how the git organizes history.
224
225 In the following, we say that commit X is "reachable" from commit Y
226 if commit X is an ancestor of commit Y. Equivalently, you could say
227 that Y is a descendant of X, or that there is a chain of parents
228 leading from commit Y to commit X.
229
230 [[history-diagrams]]
231 Understanding history: History diagrams
232 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
233
234 We will sometimes represent git history using diagrams like the one
235 below. Commits are shown as "o", and the links between them with
236 lines drawn with - / and \. Time goes left to right:
237
238
239 ................................................
240 o--o--o <-- Branch A
241 /
242 o--o--o <-- master
243 \
244 o--o--o <-- Branch B
245 ................................................
246
247 If we need to talk about a particular commit, the character "o" may
248 be replaced with another letter or number.
249
250 [[what-is-a-branch]]
251 Understanding history: What is a branch?
252 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
253
254 When we need to be precise, we will use the word "branch" to mean a line
255 of development, and "branch head" (or just "head") to mean a reference
256 to the most recent commit on a branch. In the example above, the branch
257 head named "A" is a pointer to one particular commit, but we refer to
258 the line of three commits leading up to that point as all being part of
259 "branch A".
260
261 However, when no confusion will result, we often just use the term
262 "branch" both for branches and for branch heads.
263
264 [[manipulating-branches]]
265 Manipulating branches
266 ---------------------
267
268 Creating, deleting, and modifying branches is quick and easy; here's
269 a summary of the commands:
270
271 git branch::
272 list all branches
273 git branch <branch>::
274 create a new branch named <branch>, referencing the same
275 point in history as the current branch
276 git branch <branch> <start-point>::
277 create a new branch named <branch>, referencing
278 <start-point>, which may be specified any way you like,
279 including using a branch name or a tag name
280 git branch -d <branch>::
281 delete the branch <branch>; if the branch you are deleting
282 points to a commit which is not reachable from the current
283 branch, this command will fail with a warning.
284 git branch -D <branch>::
285 even if the branch points to a commit not reachable
286 from the current branch, you may know that that commit
287 is still reachable from some other branch or tag. In that
288 case it is safe to use this command to force git to delete
289 the branch.
290 git checkout <branch>::
291 make the current branch <branch>, updating the working
292 directory to reflect the version referenced by <branch>
293 git checkout -b <new> <start-point>::
294 create a new branch <new> referencing <start-point>, and
295 check it out.
296
297 The special symbol "HEAD" can always be used to refer to the current
298 branch. In fact, git uses a file named "HEAD" in the .git directory to
299 remember which branch is current:
300
301 ------------------------------------------------
302 $ cat .git/HEAD
303 ref: refs/heads/master
304 ------------------------------------------------
305
306 [[detached-head]]
307 Examining an old version without creating a new branch
308 ------------------------------------------------------
309
310 The `git checkout` command normally expects a branch head, but will also
311 accept an arbitrary commit; for example, you can check out the commit
312 referenced by a tag:
313
314 ------------------------------------------------
315 $ git checkout v2.6.17
316 Note: moving to "v2.6.17" which isn't a local branch
317 If you want to create a new branch from this checkout, you may do so
318 (now or later) by using -b with the checkout command again. Example:
319 git checkout -b <new_branch_name>
320 HEAD is now at 427abfa... Linux v2.6.17
321 ------------------------------------------------
322
323 The HEAD then refers to the SHA-1 of the commit instead of to a branch,
324 and git branch shows that you are no longer on a branch:
325
326 ------------------------------------------------
327 $ cat .git/HEAD
328 427abfa28afedffadfca9dd8b067eb6d36bac53f
329 $ git branch
330 * (no branch)
331 master
332 ------------------------------------------------
333
334 In this case we say that the HEAD is "detached".
335
336 This is an easy way to check out a particular version without having to
337 make up a name for the new branch. You can still create a new branch
338 (or tag) for this version later if you decide to.
339
340 [[examining-remote-branches]]
341 Examining branches from a remote repository
342 -------------------------------------------
343
344 The "master" branch that was created at the time you cloned is a copy
345 of the HEAD in the repository that you cloned from. That repository
346 may also have had other branches, though, and your local repository
347 keeps branches which track each of those remote branches, called
348 remote-tracking branches, which you
349 can view using the "-r" option to linkgit:git-branch[1]:
350
351 ------------------------------------------------
352 $ git branch -r
353 origin/HEAD
354 origin/html
355 origin/maint
356 origin/man
357 origin/master
358 origin/next
359 origin/pu
360 origin/todo
361 ------------------------------------------------
362
363 In this example, "origin" is called a remote repository, or "remote"
364 for short. The branches of this repository are called "remote
365 branches" from our point of view. The remote-tracking branches listed
366 above were created based on the remote branches at clone time and will
367 be updated by "git fetch" (hence "git pull") and "git push". See
368 <<Updating-a-repository-With-git-fetch>> for details.
369
370 You might want to build on one of these remote-tracking branches
371 on a branch of your own, just as you would for a tag:
372
373 ------------------------------------------------
374 $ git checkout -b my-todo-copy origin/todo
375 ------------------------------------------------
376
377 You can also check out "origin/todo" directly to examine it or
378 write a one-off patch. See <<detached-head,detached head>>.
379
380 Note that the name "origin" is just the name that git uses by default
381 to refer to the repository that you cloned from.
382
383 [[how-git-stores-references]]
384 Naming branches, tags, and other references
385 -------------------------------------------
386
387 Branches, remote-tracking branches, and tags are all references to
388 commits. All references are named with a slash-separated path name
389 starting with "refs"; the names we've been using so far are actually
390 shorthand:
391
392 - The branch "test" is short for "refs/heads/test".
393 - The tag "v2.6.18" is short for "refs/tags/v2.6.18".
394 - "origin/master" is short for "refs/remotes/origin/master".
395
396 The full name is occasionally useful if, for example, there ever
397 exists a tag and a branch with the same name.
398
399 (Newly created refs are actually stored in the .git/refs directory,
400 under the path given by their name. However, for efficiency reasons
401 they may also be packed together in a single file; see
402 linkgit:git-pack-refs[1]).
403
404 As another useful shortcut, the "HEAD" of a repository can be referred
405 to just using the name of that repository. So, for example, "origin"
406 is usually a shortcut for the HEAD branch in the repository "origin".
407
408 For the complete list of paths which git checks for references, and
409 the order it uses to decide which to choose when there are multiple
410 references with the same shorthand name, see the "SPECIFYING
411 REVISIONS" section of linkgit:gitrevisions[7].
412
413 [[Updating-a-repository-With-git-fetch]]
414 Updating a repository with git fetch
415 ------------------------------------
416
417 Eventually the developer cloned from will do additional work in her
418 repository, creating new commits and advancing the branches to point
419 at the new commits.
420
421 The command "git fetch", with no arguments, will update all of the
422 remote-tracking branches to the latest version found in her
423 repository. It will not touch any of your own branches--not even the
424 "master" branch that was created for you on clone.
425
426 [[fetching-branches]]
427 Fetching branches from other repositories
428 -----------------------------------------
429
430 You can also track branches from repositories other than the one you
431 cloned from, using linkgit:git-remote[1]:
432
433 -------------------------------------------------
434 $ git remote add linux-nfs git://linux-nfs.org/pub/nfs-2.6.git
435 $ git fetch linux-nfs
436 * refs/remotes/linux-nfs/master: storing branch 'master' ...
437 commit: bf81b46
438 -------------------------------------------------
439
440 New remote-tracking branches will be stored under the shorthand name
441 that you gave "git remote add", in this case linux-nfs:
442
443 -------------------------------------------------
444 $ git branch -r
445 linux-nfs/master
446 origin/master
447 -------------------------------------------------
448
449 If you run "git fetch <remote>" later, the remote-tracking branches for the
450 named <remote> will be updated.
451
452 If you examine the file .git/config, you will see that git has added
453 a new stanza:
454
455 -------------------------------------------------
456 $ cat .git/config
457 ...
458 [remote "linux-nfs"]
459 url = git://linux-nfs.org/pub/nfs-2.6.git
460 fetch = +refs/heads/*:refs/remotes/linux-nfs/*
461 ...
462 -------------------------------------------------
463
464 This is what causes git to track the remote's branches; you may modify
465 or delete these configuration options by editing .git/config with a
466 text editor. (See the "CONFIGURATION FILE" section of
467 linkgit:git-config[1] for details.)
468
469 [[exploring-git-history]]
470 Exploring git history
471 =====================
472
473 Git is best thought of as a tool for storing the history of a
474 collection of files. It does this by storing compressed snapshots of
475 the contents of a file hierarchy, together with "commits" which show
476 the relationships between these snapshots.
477
478 Git provides extremely flexible and fast tools for exploring the
479 history of a project.
480
481 We start with one specialized tool that is useful for finding the
482 commit that introduced a bug into a project.
483
484 [[using-bisect]]
485 How to use bisect to find a regression
486 --------------------------------------
487
488 Suppose version 2.6.18 of your project worked, but the version at
489 "master" crashes. Sometimes the best way to find the cause of such a
490 regression is to perform a brute-force search through the project's
491 history to find the particular commit that caused the problem. The
492 linkgit:git-bisect[1] command can help you do this:
493
494 -------------------------------------------------
495 $ git bisect start
496 $ git bisect good v2.6.18
497 $ git bisect bad master
498 Bisecting: 3537 revisions left to test after this
499 [65934a9a028b88e83e2b0f8b36618fe503349f8e] BLOCK: Make USB storage depend on SCSI rather than selecting it [try #6]
500 -------------------------------------------------
501
502 If you run "git branch" at this point, you'll see that git has
503 temporarily moved you in "(no branch)". HEAD is now detached from any
504 branch and points directly to a commit (with commit id 65934...) that
505 is reachable from "master" but not from v2.6.18. Compile and test it,
506 and see whether it crashes. Assume it does crash. Then:
507
508 -------------------------------------------------
509 $ git bisect bad
510 Bisecting: 1769 revisions left to test after this
511 [7eff82c8b1511017ae605f0c99ac275a7e21b867] i2c-core: Drop useless bitmaskings
512 -------------------------------------------------
513
514 checks out an older version. Continue like this, telling git at each
515 stage whether the version it gives you is good or bad, and notice
516 that the number of revisions left to test is cut approximately in
517 half each time.
518
519 After about 13 tests (in this case), it will output the commit id of
520 the guilty commit. You can then examine the commit with
521 linkgit:git-show[1], find out who wrote it, and mail them your bug
522 report with the commit id. Finally, run
523
524 -------------------------------------------------
525 $ git bisect reset
526 -------------------------------------------------
527
528 to return you to the branch you were on before.
529
530 Note that the version which `git bisect` checks out for you at each
531 point is just a suggestion, and you're free to try a different
532 version if you think it would be a good idea. For example,
533 occasionally you may land on a commit that broke something unrelated;
534 run
535
536 -------------------------------------------------
537 $ git bisect visualize
538 -------------------------------------------------
539
540 which will run gitk and label the commit it chose with a marker that
541 says "bisect". Choose a safe-looking commit nearby, note its commit
542 id, and check it out with:
543
544 -------------------------------------------------
545 $ git reset --hard fb47ddb2db...
546 -------------------------------------------------
547
548 then test, run "bisect good" or "bisect bad" as appropriate, and
549 continue.
550
551 Instead of "git bisect visualize" and then "git reset --hard
552 fb47ddb2db...", you might just want to tell git that you want to skip
553 the current commit:
554
555 -------------------------------------------------
556 $ git bisect skip
557 -------------------------------------------------
558
559 In this case, though, git may not eventually be able to tell the first
560 bad one between some first skipped commits and a later bad commit.
561
562 There are also ways to automate the bisecting process if you have a
563 test script that can tell a good from a bad commit. See
564 linkgit:git-bisect[1] for more information about this and other "git
565 bisect" features.
566
567 [[naming-commits]]
568 Naming commits
569 --------------
570
571 We have seen several ways of naming commits already:
572
573 - 40-hexdigit object name
574 - branch name: refers to the commit at the head of the given
575 branch
576 - tag name: refers to the commit pointed to by the given tag
577 (we've seen branches and tags are special cases of
578 <<how-git-stores-references,references>>).
579 - HEAD: refers to the head of the current branch
580
581 There are many more; see the "SPECIFYING REVISIONS" section of the
582 linkgit:gitrevisions[7] man page for the complete list of ways to
583 name revisions. Some examples:
584
585 -------------------------------------------------
586 $ git show fb47ddb2 # the first few characters of the object name
587 # are usually enough to specify it uniquely
588 $ git show HEAD^ # the parent of the HEAD commit
589 $ git show HEAD^^ # the grandparent
590 $ git show HEAD~4 # the great-great-grandparent
591 -------------------------------------------------
592
593 Recall that merge commits may have more than one parent; by default,
594 ^ and ~ follow the first parent listed in the commit, but you can
595 also choose:
596
597 -------------------------------------------------
598 $ git show HEAD^1 # show the first parent of HEAD
599 $ git show HEAD^2 # show the second parent of HEAD
600 -------------------------------------------------
601
602 In addition to HEAD, there are several other special names for
603 commits:
604
605 Merges (to be discussed later), as well as operations such as
606 `git reset`, which change the currently checked-out commit, generally
607 set ORIG_HEAD to the value HEAD had before the current operation.
608
609 The `git fetch` operation always stores the head of the last fetched
610 branch in FETCH_HEAD. For example, if you run `git fetch` without
611 specifying a local branch as the target of the operation
612
613 -------------------------------------------------
614 $ git fetch git://example.com/proj.git theirbranch
615 -------------------------------------------------
616
617 the fetched commits will still be available from FETCH_HEAD.
618
619 When we discuss merges we'll also see the special name MERGE_HEAD,
620 which refers to the other branch that we're merging in to the current
621 branch.
622
623 The linkgit:git-rev-parse[1] command is a low-level command that is
624 occasionally useful for translating some name for a commit to the object
625 name for that commit:
626
627 -------------------------------------------------
628 $ git rev-parse origin
629 e05db0fd4f31dde7005f075a84f96b360d05984b
630 -------------------------------------------------
631
632 [[creating-tags]]
633 Creating tags
634 -------------
635
636 We can also create a tag to refer to a particular commit; after
637 running
638
639 -------------------------------------------------
640 $ git tag stable-1 1b2e1d63ff
641 -------------------------------------------------
642
643 You can use stable-1 to refer to the commit 1b2e1d63ff.
644
645 This creates a "lightweight" tag. If you would also like to include a
646 comment with the tag, and possibly sign it cryptographically, then you
647 should create a tag object instead; see the linkgit:git-tag[1] man page
648 for details.
649
650 [[browsing-revisions]]
651 Browsing revisions
652 ------------------
653
654 The linkgit:git-log[1] command can show lists of commits. On its
655 own, it shows all commits reachable from the parent commit; but you
656 can also make more specific requests:
657
658 -------------------------------------------------
659 $ git log v2.5.. # commits since (not reachable from) v2.5
660 $ git log test..master # commits reachable from master but not test
661 $ git log master..test # ...reachable from test but not master
662 $ git log master...test # ...reachable from either test or master,
663 # but not both
664 $ git log --since="2 weeks ago" # commits from the last 2 weeks
665 $ git log Makefile # commits which modify Makefile
666 $ git log fs/ # ... which modify any file under fs/
667 $ git log -S'foo()' # commits which add or remove any file data
668 # matching the string 'foo()'
669 -------------------------------------------------
670
671 And of course you can combine all of these; the following finds
672 commits since v2.5 which touch the Makefile or any file under fs:
673
674 -------------------------------------------------
675 $ git log v2.5.. Makefile fs/
676 -------------------------------------------------
677
678 You can also ask git log to show patches:
679
680 -------------------------------------------------
681 $ git log -p
682 -------------------------------------------------
683
684 See the "--pretty" option in the linkgit:git-log[1] man page for more
685 display options.
686
687 Note that git log starts with the most recent commit and works
688 backwards through the parents; however, since git history can contain
689 multiple independent lines of development, the particular order that
690 commits are listed in may be somewhat arbitrary.
691
692 [[generating-diffs]]
693 Generating diffs
694 ----------------
695
696 You can generate diffs between any two versions using
697 linkgit:git-diff[1]:
698
699 -------------------------------------------------
700 $ git diff master..test
701 -------------------------------------------------
702
703 That will produce the diff between the tips of the two branches. If
704 you'd prefer to find the diff from their common ancestor to test, you
705 can use three dots instead of two:
706
707 -------------------------------------------------
708 $ git diff master...test
709 -------------------------------------------------
710
711 Sometimes what you want instead is a set of patches; for this you can
712 use linkgit:git-format-patch[1]:
713
714 -------------------------------------------------
715 $ git format-patch master..test
716 -------------------------------------------------
717
718 will generate a file with a patch for each commit reachable from test
719 but not from master.
720
721 [[viewing-old-file-versions]]
722 Viewing old file versions
723 -------------------------
724
725 You can always view an old version of a file by just checking out the
726 correct revision first. But sometimes it is more convenient to be
727 able to view an old version of a single file without checking
728 anything out; this command does that:
729
730 -------------------------------------------------
731 $ git show v2.5:fs/locks.c
732 -------------------------------------------------
733
734 Before the colon may be anything that names a commit, and after it
735 may be any path to a file tracked by git.
736
737 [[history-examples]]
738 Examples
739 --------
740
741 [[counting-commits-on-a-branch]]
742 Counting the number of commits on a branch
743 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
744
745 Suppose you want to know how many commits you've made on "mybranch"
746 since it diverged from "origin":
747
748 -------------------------------------------------
749 $ git log --pretty=oneline origin..mybranch | wc -l
750 -------------------------------------------------
751
752 Alternatively, you may often see this sort of thing done with the
753 lower-level command linkgit:git-rev-list[1], which just lists the SHA-1's
754 of all the given commits:
755
756 -------------------------------------------------
757 $ git rev-list origin..mybranch | wc -l
758 -------------------------------------------------
759
760 [[checking-for-equal-branches]]
761 Check whether two branches point at the same history
762 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
763
764 Suppose you want to check whether two branches point at the same point
765 in history.
766
767 -------------------------------------------------
768 $ git diff origin..master
769 -------------------------------------------------
770
771 will tell you whether the contents of the project are the same at the
772 two branches; in theory, however, it's possible that the same project
773 contents could have been arrived at by two different historical
774 routes. You could compare the object names:
775
776 -------------------------------------------------
777 $ git rev-list origin
778 e05db0fd4f31dde7005f075a84f96b360d05984b
779 $ git rev-list master
780 e05db0fd4f31dde7005f075a84f96b360d05984b
781 -------------------------------------------------
782
783 Or you could recall that the ... operator selects all commits
784 contained reachable from either one reference or the other but not
785 both; so
786
787 -------------------------------------------------
788 $ git log origin...master
789 -------------------------------------------------
790
791 will return no commits when the two branches are equal.
792
793 [[finding-tagged-descendants]]
794 Find first tagged version including a given fix
795 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
796
797 Suppose you know that the commit e05db0fd fixed a certain problem.
798 You'd like to find the earliest tagged release that contains that
799 fix.
800
801 Of course, there may be more than one answer--if the history branched
802 after commit e05db0fd, then there could be multiple "earliest" tagged
803 releases.
804
805 You could just visually inspect the commits since e05db0fd:
806
807 -------------------------------------------------
808 $ gitk e05db0fd..
809 -------------------------------------------------
810
811 Or you can use linkgit:git-name-rev[1], which will give the commit a
812 name based on any tag it finds pointing to one of the commit's
813 descendants:
814
815 -------------------------------------------------
816 $ git name-rev --tags e05db0fd
817 e05db0fd tags/v1.5.0-rc1^0~23
818 -------------------------------------------------
819
820 The linkgit:git-describe[1] command does the opposite, naming the
821 revision using a tag on which the given commit is based:
822
823 -------------------------------------------------
824 $ git describe e05db0fd
825 v1.5.0-rc0-260-ge05db0f
826 -------------------------------------------------
827
828 but that may sometimes help you guess which tags might come after the
829 given commit.
830
831 If you just want to verify whether a given tagged version contains a
832 given commit, you could use linkgit:git-merge-base[1]:
833
834 -------------------------------------------------
835 $ git merge-base e05db0fd v1.5.0-rc1
836 e05db0fd4f31dde7005f075a84f96b360d05984b
837 -------------------------------------------------
838
839 The merge-base command finds a common ancestor of the given commits,
840 and always returns one or the other in the case where one is a
841 descendant of the other; so the above output shows that e05db0fd
842 actually is an ancestor of v1.5.0-rc1.
843
844 Alternatively, note that
845
846 -------------------------------------------------
847 $ git log v1.5.0-rc1..e05db0fd
848 -------------------------------------------------
849
850 will produce empty output if and only if v1.5.0-rc1 includes e05db0fd,
851 because it outputs only commits that are not reachable from v1.5.0-rc1.
852
853 As yet another alternative, the linkgit:git-show-branch[1] command lists
854 the commits reachable from its arguments with a display on the left-hand
855 side that indicates which arguments that commit is reachable from. So,
856 you can run something like
857
858 -------------------------------------------------
859 $ git show-branch e05db0fd v1.5.0-rc0 v1.5.0-rc1 v1.5.0-rc2
860 ! [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
861 available
862 ! [v1.5.0-rc0] GIT v1.5.0 preview
863 ! [v1.5.0-rc1] GIT v1.5.0-rc1
864 ! [v1.5.0-rc2] GIT v1.5.0-rc2
865 ...
866 -------------------------------------------------
867
868 then search for a line that looks like
869
870 -------------------------------------------------
871 + ++ [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
872 available
873 -------------------------------------------------
874
875 Which shows that e05db0fd is reachable from itself, from v1.5.0-rc1, and
876 from v1.5.0-rc2, but not from v1.5.0-rc0.
877
878 [[showing-commits-unique-to-a-branch]]
879 Showing commits unique to a given branch
880 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
881
882 Suppose you would like to see all the commits reachable from the branch
883 head named "master" but not from any other head in your repository.
884
885 We can list all the heads in this repository with
886 linkgit:git-show-ref[1]:
887
888 -------------------------------------------------
889 $ git show-ref --heads
890 bf62196b5e363d73353a9dcf094c59595f3153b7 refs/heads/core-tutorial
891 db768d5504c1bb46f63ee9d6e1772bd047e05bf9 refs/heads/maint
892 a07157ac624b2524a059a3414e99f6f44bebc1e7 refs/heads/master
893 24dbc180ea14dc1aebe09f14c8ecf32010690627 refs/heads/tutorial-2
894 1e87486ae06626c2f31eaa63d26fc0fd646c8af2 refs/heads/tutorial-fixes
895 -------------------------------------------------
896
897 We can get just the branch-head names, and remove "master", with
898 the help of the standard utilities cut and grep:
899
900 -------------------------------------------------
901 $ git show-ref --heads | cut -d' ' -f2 | grep -v '^refs/heads/master'
902 refs/heads/core-tutorial
903 refs/heads/maint
904 refs/heads/tutorial-2
905 refs/heads/tutorial-fixes
906 -------------------------------------------------
907
908 And then we can ask to see all the commits reachable from master
909 but not from these other heads:
910
911 -------------------------------------------------
912 $ gitk master --not $( git show-ref --heads | cut -d' ' -f2 |
913 grep -v '^refs/heads/master' )
914 -------------------------------------------------
915
916 Obviously, endless variations are possible; for example, to see all
917 commits reachable from some head but not from any tag in the repository:
918
919 -------------------------------------------------
920 $ gitk $( git show-ref --heads ) --not $( git show-ref --tags )
921 -------------------------------------------------
922
923 (See linkgit:gitrevisions[7] for explanations of commit-selecting
924 syntax such as `--not`.)
925
926 [[making-a-release]]
927 Creating a changelog and tarball for a software release
928 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
929
930 The linkgit:git-archive[1] command can create a tar or zip archive from
931 any version of a project; for example:
932
933 -------------------------------------------------
934 $ git archive -o latest.tar.gz --prefix=project/ HEAD
935 -------------------------------------------------
936
937 will use HEAD to produce a gzipped tar archive in which each filename
938 is preceded by `project/`. The output file format is inferred from
939 the output file extension if possible, see linkgit:git-archive[1] for
940 details.
941
942 Versions of Git older than 1.7.7 don't know about the 'tar.gz' format,
943 you'll need to use gzip explicitly:
944
945 -------------------------------------------------
946 $ git archive --format=tar --prefix=project/ HEAD | gzip >latest.tar.gz
947 -------------------------------------------------
948
949 If you're releasing a new version of a software project, you may want
950 to simultaneously make a changelog to include in the release
951 announcement.
952
953 Linus Torvalds, for example, makes new kernel releases by tagging them,
954 then running:
955
956 -------------------------------------------------
957 $ release-script 2.6.12 2.6.13-rc6 2.6.13-rc7
958 -------------------------------------------------
959
960 where release-script is a shell script that looks like:
961
962 -------------------------------------------------
963 #!/bin/sh
964 stable="$1"
965 last="$2"
966 new="$3"
967 echo "# git tag v$new"
968 echo "git archive --prefix=linux-$new/ v$new | gzip -9 > ../linux-$new.tar.gz"
969 echo "git diff v$stable v$new | gzip -9 > ../patch-$new.gz"
970 echo "git log --no-merges v$new ^v$last > ../ChangeLog-$new"
971 echo "git shortlog --no-merges v$new ^v$last > ../ShortLog"
972 echo "git diff --stat --summary -M v$last v$new > ../diffstat-$new"
973 -------------------------------------------------
974
975 and then he just cut-and-pastes the output commands after verifying that
976 they look OK.
977
978 [[Finding-commits-With-given-Content]]
979 Finding commits referencing a file with given content
980 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
981
982 Somebody hands you a copy of a file, and asks which commits modified a
983 file such that it contained the given content either before or after the
984 commit. You can find out with this:
985
986 -------------------------------------------------
987 $ git log --raw --abbrev=40 --pretty=oneline |
988 grep -B 1 `git hash-object filename`
989 -------------------------------------------------
990
991 Figuring out why this works is left as an exercise to the (advanced)
992 student. The linkgit:git-log[1], linkgit:git-diff-tree[1], and
993 linkgit:git-hash-object[1] man pages may prove helpful.
994
995 [[Developing-With-git]]
996 Developing with git
997 ===================
998
999 [[telling-git-your-name]]
1000 Telling git your name
1001 ---------------------
1002
1003 Before creating any commits, you should introduce yourself to Git.
1004 The easiest way to do so is to use linkgit:git-config[1]:
1005
1006 ------------------------------------------------
1007 $ git config --global user.name 'Your Name Comes Here'
1008 $ git config --global user.email 'you@yourdomain.example.com'
1009 ------------------------------------------------
1010
1011 Which will add the following to a file named `.gitconfig` in your
1012 home directory:
1013
1014 ------------------------------------------------
1015 [user]
1016 name = Your Name Comes Here
1017 email = you@yourdomain.example.com
1018 ------------------------------------------------
1019
1020 See the "CONFIGURATION FILE" section of linkgit:git-config[1] for
1021 details on the configuration file. The file is plain text, so you can
1022 also edit it with your favorite editor.
1023
1024
1025 [[creating-a-new-repository]]
1026 Creating a new repository
1027 -------------------------
1028
1029 Creating a new repository from scratch is very easy:
1030
1031 -------------------------------------------------
1032 $ mkdir project
1033 $ cd project
1034 $ git init
1035 -------------------------------------------------
1036
1037 If you have some initial content (say, a tarball):
1038
1039 -------------------------------------------------
1040 $ tar xzvf project.tar.gz
1041 $ cd project
1042 $ git init
1043 $ git add . # include everything below ./ in the first commit:
1044 $ git commit
1045 -------------------------------------------------
1046
1047 [[how-to-make-a-commit]]
1048 How to make a commit
1049 --------------------
1050
1051 Creating a new commit takes three steps:
1052
1053 1. Making some changes to the working directory using your
1054 favorite editor.
1055 2. Telling git about your changes.
1056 3. Creating the commit using the content you told git about
1057 in step 2.
1058
1059 In practice, you can interleave and repeat steps 1 and 2 as many
1060 times as you want: in order to keep track of what you want committed
1061 at step 3, git maintains a snapshot of the tree's contents in a
1062 special staging area called "the index."
1063
1064 At the beginning, the content of the index will be identical to
1065 that of the HEAD. The command "git diff --cached", which shows
1066 the difference between the HEAD and the index, should therefore
1067 produce no output at that point.
1068
1069 Modifying the index is easy:
1070
1071 To update the index with the new contents of a modified file, use
1072
1073 -------------------------------------------------
1074 $ git add path/to/file
1075 -------------------------------------------------
1076
1077 To add the contents of a new file to the index, use
1078
1079 -------------------------------------------------
1080 $ git add path/to/file
1081 -------------------------------------------------
1082
1083 To remove a file from the index and from the working tree,
1084
1085 -------------------------------------------------
1086 $ git rm path/to/file
1087 -------------------------------------------------
1088
1089 After each step you can verify that
1090
1091 -------------------------------------------------
1092 $ git diff --cached
1093 -------------------------------------------------
1094
1095 always shows the difference between the HEAD and the index file--this
1096 is what you'd commit if you created the commit now--and that
1097
1098 -------------------------------------------------
1099 $ git diff
1100 -------------------------------------------------
1101
1102 shows the difference between the working tree and the index file.
1103
1104 Note that "git add" always adds just the current contents of a file
1105 to the index; further changes to the same file will be ignored unless
1106 you run `git add` on the file again.
1107
1108 When you're ready, just run
1109
1110 -------------------------------------------------
1111 $ git commit
1112 -------------------------------------------------
1113
1114 and git will prompt you for a commit message and then create the new
1115 commit. Check to make sure it looks like what you expected with
1116
1117 -------------------------------------------------
1118 $ git show
1119 -------------------------------------------------
1120
1121 As a special shortcut,
1122
1123 -------------------------------------------------
1124 $ git commit -a
1125 -------------------------------------------------
1126
1127 will update the index with any files that you've modified or removed
1128 and create a commit, all in one step.
1129
1130 A number of commands are useful for keeping track of what you're
1131 about to commit:
1132
1133 -------------------------------------------------
1134 $ git diff --cached # difference between HEAD and the index; what
1135 # would be committed if you ran "commit" now.
1136 $ git diff # difference between the index file and your
1137 # working directory; changes that would not
1138 # be included if you ran "commit" now.
1139 $ git diff HEAD # difference between HEAD and working tree; what
1140 # would be committed if you ran "commit -a" now.
1141 $ git status # a brief per-file summary of the above.
1142 -------------------------------------------------
1143
1144 You can also use linkgit:git-gui[1] to create commits, view changes in
1145 the index and the working tree files, and individually select diff hunks
1146 for inclusion in the index (by right-clicking on the diff hunk and
1147 choosing "Stage Hunk For Commit").
1148
1149 [[creating-good-commit-messages]]
1150 Creating good commit messages
1151 -----------------------------
1152
1153 Though not required, it's a good idea to begin the commit message
1154 with a single short (less than 50 character) line summarizing the
1155 change, followed by a blank line and then a more thorough
1156 description. The text up to the first blank line in a commit
1157 message is treated as the commit title, and that title is used
1158 throughout git. For example, linkgit:git-format-patch[1] turns a
1159 commit into email, and it uses the title on the Subject line and the
1160 rest of the commit in the body.
1161
1162
1163 [[ignoring-files]]
1164 Ignoring files
1165 --------------
1166
1167 A project will often generate files that you do 'not' want to track with git.
1168 This typically includes files generated by a build process or temporary
1169 backup files made by your editor. Of course, 'not' tracking files with git
1170 is just a matter of 'not' calling `git add` on them. But it quickly becomes
1171 annoying to have these untracked files lying around; e.g. they make
1172 `git add .` practically useless, and they keep showing up in the output of
1173 `git status`.
1174
1175 You can tell git to ignore certain files by creating a file called .gitignore
1176 in the top level of your working directory, with contents such as:
1177
1178 -------------------------------------------------
1179 # Lines starting with '#' are considered comments.
1180 # Ignore any file named foo.txt.
1181 foo.txt
1182 # Ignore (generated) html files,
1183 *.html
1184 # except foo.html which is maintained by hand.
1185 !foo.html
1186 # Ignore objects and archives.
1187 *.[oa]
1188 -------------------------------------------------
1189
1190 See linkgit:gitignore[5] for a detailed explanation of the syntax. You can
1191 also place .gitignore files in other directories in your working tree, and they
1192 will apply to those directories and their subdirectories. The `.gitignore`
1193 files can be added to your repository like any other files (just run `git add
1194 .gitignore` and `git commit`, as usual), which is convenient when the exclude
1195 patterns (such as patterns matching build output files) would also make sense
1196 for other users who clone your repository.
1197
1198 If you wish the exclude patterns to affect only certain repositories
1199 (instead of every repository for a given project), you may instead put
1200 them in a file in your repository named .git/info/exclude, or in any file
1201 specified by the `core.excludesfile` configuration variable. Some git
1202 commands can also take exclude patterns directly on the command line.
1203 See linkgit:gitignore[5] for the details.
1204
1205 [[how-to-merge]]
1206 How to merge
1207 ------------
1208
1209 You can rejoin two diverging branches of development using
1210 linkgit:git-merge[1]:
1211
1212 -------------------------------------------------
1213 $ git merge branchname
1214 -------------------------------------------------
1215
1216 merges the development in the branch "branchname" into the current
1217 branch.
1218
1219 A merge is made by combining the changes made in "branchname" and the
1220 changes made up to the latest commit in your current branch since
1221 their histories forked. The work tree is overwritten by the result of
1222 the merge when this combining is done cleanly, or overwritten by a
1223 half-merged results when this combining results in conflicts.
1224 Therefore, if you have uncommitted changes touching the same files as
1225 the ones impacted by the merge, Git will refuse to proceed. Most of
1226 the time, you will want to commit your changes before you can merge,
1227 and if you don't, then linkgit:git-stash[1] can take these changes
1228 away while you're doing the merge, and reapply them afterwards.
1229
1230 If the changes are independent enough, Git will automatically complete
1231 the merge and commit the result (or reuse an existing commit in case
1232 of <<fast-forwards,fast-forward>>, see below). On the other hand,
1233 if there are conflicts--for example, if the same file is
1234 modified in two different ways in the remote branch and the local
1235 branch--then you are warned; the output may look something like this:
1236
1237 -------------------------------------------------
1238 $ git merge next
1239 100% (4/4) done
1240 Auto-merged file.txt
1241 CONFLICT (content): Merge conflict in file.txt
1242 Automatic merge failed; fix conflicts and then commit the result.
1243 -------------------------------------------------
1244
1245 Conflict markers are left in the problematic files, and after
1246 you resolve the conflicts manually, you can update the index
1247 with the contents and run git commit, as you normally would when
1248 creating a new file.
1249
1250 If you examine the resulting commit using gitk, you will see that it
1251 has two parents, one pointing to the top of the current branch, and
1252 one to the top of the other branch.
1253
1254 [[resolving-a-merge]]
1255 Resolving a merge
1256 -----------------
1257
1258 When a merge isn't resolved automatically, git leaves the index and
1259 the working tree in a special state that gives you all the
1260 information you need to help resolve the merge.
1261
1262 Files with conflicts are marked specially in the index, so until you
1263 resolve the problem and update the index, linkgit:git-commit[1] will
1264 fail:
1265
1266 -------------------------------------------------
1267 $ git commit
1268 file.txt: needs merge
1269 -------------------------------------------------
1270
1271 Also, linkgit:git-status[1] will list those files as "unmerged", and the
1272 files with conflicts will have conflict markers added, like this:
1273
1274 -------------------------------------------------
1275 <<<<<<< HEAD:file.txt
1276 Hello world
1277 =======
1278 Goodbye
1279 >>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1280 -------------------------------------------------
1281
1282 All you need to do is edit the files to resolve the conflicts, and then
1283
1284 -------------------------------------------------
1285 $ git add file.txt
1286 $ git commit
1287 -------------------------------------------------
1288
1289 Note that the commit message will already be filled in for you with
1290 some information about the merge. Normally you can just use this
1291 default message unchanged, but you may add additional commentary of
1292 your own if desired.
1293
1294 The above is all you need to know to resolve a simple merge. But git
1295 also provides more information to help resolve conflicts:
1296
1297 [[conflict-resolution]]
1298 Getting conflict-resolution help during a merge
1299 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1300
1301 All of the changes that git was able to merge automatically are
1302 already added to the index file, so linkgit:git-diff[1] shows only
1303 the conflicts. It uses an unusual syntax:
1304
1305 -------------------------------------------------
1306 $ git diff
1307 diff --cc file.txt
1308 index 802992c,2b60207..0000000
1309 --- a/file.txt
1310 +++ b/file.txt
1311 @@@ -1,1 -1,1 +1,5 @@@
1312 ++<<<<<<< HEAD:file.txt
1313 +Hello world
1314 ++=======
1315 + Goodbye
1316 ++>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1317 -------------------------------------------------
1318
1319 Recall that the commit which will be committed after we resolve this
1320 conflict will have two parents instead of the usual one: one parent
1321 will be HEAD, the tip of the current branch; the other will be the
1322 tip of the other branch, which is stored temporarily in MERGE_HEAD.
1323
1324 During the merge, the index holds three versions of each file. Each of
1325 these three "file stages" represents a different version of the file:
1326
1327 -------------------------------------------------
1328 $ git show :1:file.txt # the file in a common ancestor of both branches
1329 $ git show :2:file.txt # the version from HEAD.
1330 $ git show :3:file.txt # the version from MERGE_HEAD.
1331 -------------------------------------------------
1332
1333 When you ask linkgit:git-diff[1] to show the conflicts, it runs a
1334 three-way diff between the conflicted merge results in the work tree with
1335 stages 2 and 3 to show only hunks whose contents come from both sides,
1336 mixed (in other words, when a hunk's merge results come only from stage 2,
1337 that part is not conflicting and is not shown. Same for stage 3).
1338
1339 The diff above shows the differences between the working-tree version of
1340 file.txt and the stage 2 and stage 3 versions. So instead of preceding
1341 each line by a single "+" or "-", it now uses two columns: the first
1342 column is used for differences between the first parent and the working
1343 directory copy, and the second for differences between the second parent
1344 and the working directory copy. (See the "COMBINED DIFF FORMAT" section
1345 of linkgit:git-diff-files[1] for a details of the format.)
1346
1347 After resolving the conflict in the obvious way (but before updating the
1348 index), the diff will look like:
1349
1350 -------------------------------------------------
1351 $ git diff
1352 diff --cc file.txt
1353 index 802992c,2b60207..0000000
1354 --- a/file.txt
1355 +++ b/file.txt
1356 @@@ -1,1 -1,1 +1,1 @@@
1357 - Hello world
1358 -Goodbye
1359 ++Goodbye world
1360 -------------------------------------------------
1361
1362 This shows that our resolved version deleted "Hello world" from the
1363 first parent, deleted "Goodbye" from the second parent, and added
1364 "Goodbye world", which was previously absent from both.
1365
1366 Some special diff options allow diffing the working directory against
1367 any of these stages:
1368
1369 -------------------------------------------------
1370 $ git diff -1 file.txt # diff against stage 1
1371 $ git diff --base file.txt # same as the above
1372 $ git diff -2 file.txt # diff against stage 2
1373 $ git diff --ours file.txt # same as the above
1374 $ git diff -3 file.txt # diff against stage 3
1375 $ git diff --theirs file.txt # same as the above.
1376 -------------------------------------------------
1377
1378 The linkgit:git-log[1] and linkgit:gitk[1] commands also provide special help
1379 for merges:
1380
1381 -------------------------------------------------
1382 $ git log --merge
1383 $ gitk --merge
1384 -------------------------------------------------
1385
1386 These will display all commits which exist only on HEAD or on
1387 MERGE_HEAD, and which touch an unmerged file.
1388
1389 You may also use linkgit:git-mergetool[1], which lets you merge the
1390 unmerged files using external tools such as Emacs or kdiff3.
1391
1392 Each time you resolve the conflicts in a file and update the index:
1393
1394 -------------------------------------------------
1395 $ git add file.txt
1396 -------------------------------------------------
1397
1398 the different stages of that file will be "collapsed", after which
1399 `git diff` will (by default) no longer show diffs for that file.
1400
1401 [[undoing-a-merge]]
1402 Undoing a merge
1403 ---------------
1404
1405 If you get stuck and decide to just give up and throw the whole mess
1406 away, you can always return to the pre-merge state with
1407
1408 -------------------------------------------------
1409 $ git reset --hard HEAD
1410 -------------------------------------------------
1411
1412 Or, if you've already committed the merge that you want to throw away,
1413
1414 -------------------------------------------------
1415 $ git reset --hard ORIG_HEAD
1416 -------------------------------------------------
1417
1418 However, this last command can be dangerous in some cases--never
1419 throw away a commit you have already committed if that commit may
1420 itself have been merged into another branch, as doing so may confuse
1421 further merges.
1422
1423 [[fast-forwards]]
1424 Fast-forward merges
1425 -------------------
1426
1427 There is one special case not mentioned above, which is treated
1428 differently. Normally, a merge results in a merge commit, with two
1429 parents, one pointing at each of the two lines of development that
1430 were merged.
1431
1432 However, if the current branch is a descendant of the other--so every
1433 commit present in the one is already contained in the other--then git
1434 just performs a "fast-forward"; the head of the current branch is moved
1435 forward to point at the head of the merged-in branch, without any new
1436 commits being created.
1437
1438 [[fixing-mistakes]]
1439 Fixing mistakes
1440 ---------------
1441
1442 If you've messed up the working tree, but haven't yet committed your
1443 mistake, you can return the entire working tree to the last committed
1444 state with
1445
1446 -------------------------------------------------
1447 $ git reset --hard HEAD
1448 -------------------------------------------------
1449
1450 If you make a commit that you later wish you hadn't, there are two
1451 fundamentally different ways to fix the problem:
1452
1453 1. You can create a new commit that undoes whatever was done
1454 by the old commit. This is the correct thing if your
1455 mistake has already been made public.
1456
1457 2. You can go back and modify the old commit. You should
1458 never do this if you have already made the history public;
1459 git does not normally expect the "history" of a project to
1460 change, and cannot correctly perform repeated merges from
1461 a branch that has had its history changed.
1462
1463 [[reverting-a-commit]]
1464 Fixing a mistake with a new commit
1465 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1466
1467 Creating a new commit that reverts an earlier change is very easy;
1468 just pass the linkgit:git-revert[1] command a reference to the bad
1469 commit; for example, to revert the most recent commit:
1470
1471 -------------------------------------------------
1472 $ git revert HEAD
1473 -------------------------------------------------
1474
1475 This will create a new commit which undoes the change in HEAD. You
1476 will be given a chance to edit the commit message for the new commit.
1477
1478 You can also revert an earlier change, for example, the next-to-last:
1479
1480 -------------------------------------------------
1481 $ git revert HEAD^
1482 -------------------------------------------------
1483
1484 In this case git will attempt to undo the old change while leaving
1485 intact any changes made since then. If more recent changes overlap
1486 with the changes to be reverted, then you will be asked to fix
1487 conflicts manually, just as in the case of <<resolving-a-merge,
1488 resolving a merge>>.
1489
1490 [[fixing-a-mistake-by-rewriting-history]]
1491 Fixing a mistake by rewriting history
1492 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1493
1494 If the problematic commit is the most recent commit, and you have not
1495 yet made that commit public, then you may just
1496 <<undoing-a-merge,destroy it using `git reset`>>.
1497
1498 Alternatively, you
1499 can edit the working directory and update the index to fix your
1500 mistake, just as if you were going to <<how-to-make-a-commit,create a
1501 new commit>>, then run
1502
1503 -------------------------------------------------
1504 $ git commit --amend
1505 -------------------------------------------------
1506
1507 which will replace the old commit by a new commit incorporating your
1508 changes, giving you a chance to edit the old commit message first.
1509
1510 Again, you should never do this to a commit that may already have
1511 been merged into another branch; use linkgit:git-revert[1] instead in
1512 that case.
1513
1514 It is also possible to replace commits further back in the history, but
1515 this is an advanced topic to be left for
1516 <<cleaning-up-history,another chapter>>.
1517
1518 [[checkout-of-path]]
1519 Checking out an old version of a file
1520 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1521
1522 In the process of undoing a previous bad change, you may find it
1523 useful to check out an older version of a particular file using
1524 linkgit:git-checkout[1]. We've used `git checkout` before to switch
1525 branches, but it has quite different behavior if it is given a path
1526 name: the command
1527
1528 -------------------------------------------------
1529 $ git checkout HEAD^ path/to/file
1530 -------------------------------------------------
1531
1532 replaces path/to/file by the contents it had in the commit HEAD^, and
1533 also updates the index to match. It does not change branches.
1534
1535 If you just want to look at an old version of the file, without
1536 modifying the working directory, you can do that with
1537 linkgit:git-show[1]:
1538
1539 -------------------------------------------------
1540 $ git show HEAD^:path/to/file
1541 -------------------------------------------------
1542
1543 which will display the given version of the file.
1544
1545 [[interrupted-work]]
1546 Temporarily setting aside work in progress
1547 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1548
1549 While you are in the middle of working on something complicated, you
1550 find an unrelated but obvious and trivial bug. You would like to fix it
1551 before continuing. You can use linkgit:git-stash[1] to save the current
1552 state of your work, and after fixing the bug (or, optionally after doing
1553 so on a different branch and then coming back), unstash the
1554 work-in-progress changes.
1555
1556 ------------------------------------------------
1557 $ git stash save "work in progress for foo feature"
1558 ------------------------------------------------
1559
1560 This command will save your changes away to the `stash`, and
1561 reset your working tree and the index to match the tip of your
1562 current branch. Then you can make your fix as usual.
1563
1564 ------------------------------------------------
1565 ... edit and test ...
1566 $ git commit -a -m "blorpl: typofix"
1567 ------------------------------------------------
1568
1569 After that, you can go back to what you were working on with
1570 `git stash pop`:
1571
1572 ------------------------------------------------
1573 $ git stash pop
1574 ------------------------------------------------
1575
1576
1577 [[ensuring-good-performance]]
1578 Ensuring good performance
1579 -------------------------
1580
1581 On large repositories, Git depends on compression to keep the history
1582 information from taking up too much space on disk or in memory. Some
1583 git commands may automatically run linkgit:git-gc[1], so you don't
1584 have to worry about running it manually. However, compressing a large
1585 repository may take a while, so you may want to call `gc` explicitly
1586 to avoid automatic compression kicking in when it is not convenient.
1587
1588
1589 [[ensuring-reliability]]
1590 Ensuring reliability
1591 --------------------
1592
1593 [[checking-for-corruption]]
1594 Checking the repository for corruption
1595 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1596
1597 The linkgit:git-fsck[1] command runs a number of self-consistency checks
1598 on the repository, and reports on any problems. This may take some
1599 time.
1600
1601 -------------------------------------------------
1602 $ git fsck
1603 dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1604 dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1605 dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1606 dangling blob 218761f9d90712d37a9c5e36f406f92202db07eb
1607 dangling commit bf093535a34a4d35731aa2bd90fe6b176302f14f
1608 dangling commit 8e4bec7f2ddaa268bef999853c25755452100f8e
1609 dangling tree d50bb86186bf27b681d25af89d3b5b68382e4085
1610 dangling tree b24c2473f1fd3d91352a624795be026d64c8841f
1611 ...
1612 -------------------------------------------------
1613
1614 You will see informational messages on dangling objects. They are objects
1615 that still exist in the repository but are no longer referenced by any of
1616 your branches, and can (and will) be removed after a while with "gc".
1617 You can run `git fsck --no-dangling` to suppress these messages, and still
1618 view real errors.
1619
1620 [[recovering-lost-changes]]
1621 Recovering lost changes
1622 ~~~~~~~~~~~~~~~~~~~~~~~
1623
1624 [[reflogs]]
1625 Reflogs
1626 ^^^^^^^
1627
1628 Say you modify a branch with +linkgit:git-reset[1] \--hard+, and then
1629 realize that the branch was the only reference you had to that point in
1630 history.
1631
1632 Fortunately, git also keeps a log, called a "reflog", of all the
1633 previous values of each branch. So in this case you can still find the
1634 old history using, for example,
1635
1636 -------------------------------------------------
1637 $ git log master@{1}
1638 -------------------------------------------------
1639
1640 This lists the commits reachable from the previous version of the
1641 "master" branch head. This syntax can be used with any git command
1642 that accepts a commit, not just with git log. Some other examples:
1643
1644 -------------------------------------------------
1645 $ git show master@{2} # See where the branch pointed 2,
1646 $ git show master@{3} # 3, ... changes ago.
1647 $ gitk master@{yesterday} # See where it pointed yesterday,
1648 $ gitk master@{"1 week ago"} # ... or last week
1649 $ git log --walk-reflogs master # show reflog entries for master
1650 -------------------------------------------------
1651
1652 A separate reflog is kept for the HEAD, so
1653
1654 -------------------------------------------------
1655 $ git show HEAD@{"1 week ago"}
1656 -------------------------------------------------
1657
1658 will show what HEAD pointed to one week ago, not what the current branch
1659 pointed to one week ago. This allows you to see the history of what
1660 you've checked out.
1661
1662 The reflogs are kept by default for 30 days, after which they may be
1663 pruned. See linkgit:git-reflog[1] and linkgit:git-gc[1] to learn
1664 how to control this pruning, and see the "SPECIFYING REVISIONS"
1665 section of linkgit:gitrevisions[7] for details.
1666
1667 Note that the reflog history is very different from normal git history.
1668 While normal history is shared by every repository that works on the
1669 same project, the reflog history is not shared: it tells you only about
1670 how the branches in your local repository have changed over time.
1671
1672 [[dangling-object-recovery]]
1673 Examining dangling objects
1674 ^^^^^^^^^^^^^^^^^^^^^^^^^^
1675
1676 In some situations the reflog may not be able to save you. For example,
1677 suppose you delete a branch, then realize you need the history it
1678 contained. The reflog is also deleted; however, if you have not yet
1679 pruned the repository, then you may still be able to find the lost
1680 commits in the dangling objects that `git fsck` reports. See
1681 <<dangling-objects>> for the details.
1682
1683 -------------------------------------------------
1684 $ git fsck
1685 dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1686 dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1687 dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1688 ...
1689 -------------------------------------------------
1690
1691 You can examine
1692 one of those dangling commits with, for example,
1693
1694 ------------------------------------------------
1695 $ gitk 7281251ddd --not --all
1696 ------------------------------------------------
1697
1698 which does what it sounds like: it says that you want to see the commit
1699 history that is described by the dangling commit(s), but not the
1700 history that is described by all your existing branches and tags. Thus
1701 you get exactly the history reachable from that commit that is lost.
1702 (And notice that it might not be just one commit: we only report the
1703 "tip of the line" as being dangling, but there might be a whole deep
1704 and complex commit history that was dropped.)
1705
1706 If you decide you want the history back, you can always create a new
1707 reference pointing to it, for example, a new branch:
1708
1709 ------------------------------------------------
1710 $ git branch recovered-branch 7281251ddd
1711 ------------------------------------------------
1712
1713 Other types of dangling objects (blobs and trees) are also possible, and
1714 dangling objects can arise in other situations.
1715
1716
1717 [[sharing-development]]
1718 Sharing development with others
1719 ===============================
1720
1721 [[getting-updates-With-git-pull]]
1722 Getting updates with git pull
1723 -----------------------------
1724
1725 After you clone a repository and commit a few changes of your own, you
1726 may wish to check the original repository for updates and merge them
1727 into your own work.
1728
1729 We have already seen <<Updating-a-repository-With-git-fetch,how to
1730 keep remote-tracking branches up to date>> with linkgit:git-fetch[1],
1731 and how to merge two branches. So you can merge in changes from the
1732 original repository's master branch with:
1733
1734 -------------------------------------------------
1735 $ git fetch
1736 $ git merge origin/master
1737 -------------------------------------------------
1738
1739 However, the linkgit:git-pull[1] command provides a way to do this in
1740 one step:
1741
1742 -------------------------------------------------
1743 $ git pull origin master
1744 -------------------------------------------------
1745
1746 In fact, if you have "master" checked out, then this branch has been
1747 configured by "git clone" to get changes from the HEAD branch of the
1748 origin repository. So often you can
1749 accomplish the above with just a simple
1750
1751 -------------------------------------------------
1752 $ git pull
1753 -------------------------------------------------
1754
1755 This command will fetch changes from the remote branches to your
1756 remote-tracking branches `origin/*`, and merge the default branch into
1757 the current branch.
1758
1759 More generally, a branch that is created from a remote-tracking branch
1760 will pull
1761 by default from that branch. See the descriptions of the
1762 branch.<name>.remote and branch.<name>.merge options in
1763 linkgit:git-config[1], and the discussion of the `--track` option in
1764 linkgit:git-checkout[1], to learn how to control these defaults.
1765
1766 In addition to saving you keystrokes, "git pull" also helps you by
1767 producing a default commit message documenting the branch and
1768 repository that you pulled from.
1769
1770 (But note that no such commit will be created in the case of a
1771 <<fast-forwards,fast-forward>>; instead, your branch will just be
1772 updated to point to the latest commit from the upstream branch.)
1773
1774 The `git pull` command can also be given "." as the "remote" repository,
1775 in which case it just merges in a branch from the current repository; so
1776 the commands
1777
1778 -------------------------------------------------
1779 $ git pull . branch
1780 $ git merge branch
1781 -------------------------------------------------
1782
1783 are roughly equivalent. The former is actually very commonly used.
1784
1785 [[submitting-patches]]
1786 Submitting patches to a project
1787 -------------------------------
1788
1789 If you just have a few changes, the simplest way to submit them may
1790 just be to send them as patches in email:
1791
1792 First, use linkgit:git-format-patch[1]; for example:
1793
1794 -------------------------------------------------
1795 $ git format-patch origin
1796 -------------------------------------------------
1797
1798 will produce a numbered series of files in the current directory, one
1799 for each patch in the current branch but not in origin/HEAD.
1800
1801 `git format-patch` can include an initial "cover letter". You can insert
1802 commentary on individual patches after the three dash line which
1803 `format-patch` places after the commit message but before the patch
1804 itself. If you use `git notes` to track your cover letter material,
1805 `git format-patch --notes` will include the commit's notes in a similar
1806 manner.
1807
1808 You can then import these into your mail client and send them by
1809 hand. However, if you have a lot to send at once, you may prefer to
1810 use the linkgit:git-send-email[1] script to automate the process.
1811 Consult the mailing list for your project first to determine how they
1812 prefer such patches be handled.
1813
1814 [[importing-patches]]
1815 Importing patches to a project
1816 ------------------------------
1817
1818 Git also provides a tool called linkgit:git-am[1] (am stands for
1819 "apply mailbox"), for importing such an emailed series of patches.
1820 Just save all of the patch-containing messages, in order, into a
1821 single mailbox file, say "patches.mbox", then run
1822
1823 -------------------------------------------------
1824 $ git am -3 patches.mbox
1825 -------------------------------------------------
1826
1827 Git will apply each patch in order; if any conflicts are found, it
1828 will stop, and you can fix the conflicts as described in
1829 "<<resolving-a-merge,Resolving a merge>>". (The "-3" option tells
1830 git to perform a merge; if you would prefer it just to abort and
1831 leave your tree and index untouched, you may omit that option.)
1832
1833 Once the index is updated with the results of the conflict
1834 resolution, instead of creating a new commit, just run
1835
1836 -------------------------------------------------
1837 $ git am --resolved
1838 -------------------------------------------------
1839
1840 and git will create the commit for you and continue applying the
1841 remaining patches from the mailbox.
1842
1843 The final result will be a series of commits, one for each patch in
1844 the original mailbox, with authorship and commit log message each
1845 taken from the message containing each patch.
1846
1847 [[public-repositories]]
1848 Public git repositories
1849 -----------------------
1850
1851 Another way to submit changes to a project is to tell the maintainer
1852 of that project to pull the changes from your repository using
1853 linkgit:git-pull[1]. In the section "<<getting-updates-With-git-pull,
1854 Getting updates with `git pull`>>" we described this as a way to get
1855 updates from the "main" repository, but it works just as well in the
1856 other direction.
1857
1858 If you and the maintainer both have accounts on the same machine, then
1859 you can just pull changes from each other's repositories directly;
1860 commands that accept repository URLs as arguments will also accept a
1861 local directory name:
1862
1863 -------------------------------------------------
1864 $ git clone /path/to/repository
1865 $ git pull /path/to/other/repository
1866 -------------------------------------------------
1867
1868 or an ssh URL:
1869
1870 -------------------------------------------------
1871 $ git clone ssh://yourhost/~you/repository
1872 -------------------------------------------------
1873
1874 For projects with few developers, or for synchronizing a few private
1875 repositories, this may be all you need.
1876
1877 However, the more common way to do this is to maintain a separate public
1878 repository (usually on a different host) for others to pull changes
1879 from. This is usually more convenient, and allows you to cleanly
1880 separate private work in progress from publicly visible work.
1881
1882 You will continue to do your day-to-day work in your personal
1883 repository, but periodically "push" changes from your personal
1884 repository into your public repository, allowing other developers to
1885 pull from that repository. So the flow of changes, in a situation
1886 where there is one other developer with a public repository, looks
1887 like this:
1888
1889 you push
1890 your personal repo ------------------> your public repo
1891 ^ |
1892 | |
1893 | you pull | they pull
1894 | |
1895 | |
1896 | they push V
1897 their public repo <------------------- their repo
1898
1899 We explain how to do this in the following sections.
1900
1901 [[setting-up-a-public-repository]]
1902 Setting up a public repository
1903 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1904
1905 Assume your personal repository is in the directory ~/proj. We
1906 first create a new clone of the repository and tell `git daemon` that it
1907 is meant to be public:
1908
1909 -------------------------------------------------
1910 $ git clone --bare ~/proj proj.git
1911 $ touch proj.git/git-daemon-export-ok
1912 -------------------------------------------------
1913
1914 The resulting directory proj.git contains a "bare" git repository--it is
1915 just the contents of the ".git" directory, without any files checked out
1916 around it.
1917
1918 Next, copy proj.git to the server where you plan to host the
1919 public repository. You can use scp, rsync, or whatever is most
1920 convenient.
1921
1922 [[exporting-via-git]]
1923 Exporting a git repository via the git protocol
1924 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1925
1926 This is the preferred method.
1927
1928 If someone else administers the server, they should tell you what
1929 directory to put the repository in, and what git:// URL it will appear
1930 at. You can then skip to the section
1931 "<<pushing-changes-to-a-public-repository,Pushing changes to a public
1932 repository>>", below.
1933
1934 Otherwise, all you need to do is start linkgit:git-daemon[1]; it will
1935 listen on port 9418. By default, it will allow access to any directory
1936 that looks like a git directory and contains the magic file
1937 git-daemon-export-ok. Passing some directory paths as `git daemon`
1938 arguments will further restrict the exports to those paths.
1939
1940 You can also run `git daemon` as an inetd service; see the
1941 linkgit:git-daemon[1] man page for details. (See especially the
1942 examples section.)
1943
1944 [[exporting-via-http]]
1945 Exporting a git repository via HTTP
1946 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1947
1948 The git protocol gives better performance and reliability, but on a
1949 host with a web server set up, HTTP exports may be simpler to set up.
1950
1951 All you need to do is place the newly created bare git repository in
1952 a directory that is exported by the web server, and make some
1953 adjustments to give web clients some extra information they need:
1954
1955 -------------------------------------------------
1956 $ mv proj.git /home/you/public_html/proj.git
1957 $ cd proj.git
1958 $ git --bare update-server-info
1959 $ mv hooks/post-update.sample hooks/post-update
1960 -------------------------------------------------
1961
1962 (For an explanation of the last two lines, see
1963 linkgit:git-update-server-info[1] and linkgit:githooks[5].)
1964
1965 Advertise the URL of proj.git. Anybody else should then be able to
1966 clone or pull from that URL, for example with a command line like:
1967
1968 -------------------------------------------------
1969 $ git clone http://yourserver.com/~you/proj.git
1970 -------------------------------------------------
1971
1972 (See also
1973 link:howto/setup-git-server-over-http.txt[setup-git-server-over-http]
1974 for a slightly more sophisticated setup using WebDAV which also
1975 allows pushing over HTTP.)
1976
1977 [[pushing-changes-to-a-public-repository]]
1978 Pushing changes to a public repository
1979 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1980
1981 Note that the two techniques outlined above (exporting via
1982 <<exporting-via-http,http>> or <<exporting-via-git,git>>) allow other
1983 maintainers to fetch your latest changes, but they do not allow write
1984 access, which you will need to update the public repository with the
1985 latest changes created in your private repository.
1986
1987 The simplest way to do this is using linkgit:git-push[1] and ssh; to
1988 update the remote branch named "master" with the latest state of your
1989 branch named "master", run
1990
1991 -------------------------------------------------
1992 $ git push ssh://yourserver.com/~you/proj.git master:master
1993 -------------------------------------------------
1994
1995 or just
1996
1997 -------------------------------------------------
1998 $ git push ssh://yourserver.com/~you/proj.git master
1999 -------------------------------------------------
2000
2001 As with `git fetch`, `git push` will complain if this does not result in a
2002 <<fast-forwards,fast-forward>>; see the following section for details on
2003 handling this case.
2004
2005 Note that the target of a "push" is normally a
2006 <<def_bare_repository,bare>> repository. You can also push to a
2007 repository that has a checked-out working tree, but the working tree
2008 will not be updated by the push. This may lead to unexpected results if
2009 the branch you push to is the currently checked-out branch!
2010
2011 As with `git fetch`, you may also set up configuration options to
2012 save typing; so, for example:
2013
2014 -------------------------------------------------
2015 $ git remote add public-repo ssh://yourserver.com/~you/proj.git
2016 -------------------------------------------------
2017
2018 adds the following to `.git/config`:
2019
2020 -------------------------------------------------
2021 [remote "public-repo"]
2022 url = yourserver.com:proj.git
2023 fetch = +refs/heads/*:refs/remotes/example/*
2024 -------------------------------------------------
2025
2026 which lets you do the same push with just
2027
2028 -------------------------------------------------
2029 $ git push public-repo master
2030 -------------------------------------------------
2031
2032 See the explanations of the remote.<name>.url, branch.<name>.remote,
2033 and remote.<name>.push options in linkgit:git-config[1] for
2034 details.
2035
2036 [[forcing-push]]
2037 What to do when a push fails
2038 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2039
2040 If a push would not result in a <<fast-forwards,fast-forward>> of the
2041 remote branch, then it will fail with an error like:
2042
2043 -------------------------------------------------
2044 error: remote 'refs/heads/master' is not an ancestor of
2045 local 'refs/heads/master'.
2046 Maybe you are not up-to-date and need to pull first?
2047 error: failed to push to 'ssh://yourserver.com/~you/proj.git'
2048 -------------------------------------------------
2049
2050 This can happen, for example, if you:
2051
2052 - use `git reset --hard` to remove already-published commits, or
2053 - use `git commit --amend` to replace already-published commits
2054 (as in <<fixing-a-mistake-by-rewriting-history>>), or
2055 - use `git rebase` to rebase any already-published commits (as
2056 in <<using-git-rebase>>).
2057
2058 You may force `git push` to perform the update anyway by preceding the
2059 branch name with a plus sign:
2060
2061 -------------------------------------------------
2062 $ git push ssh://yourserver.com/~you/proj.git +master
2063 -------------------------------------------------
2064
2065 Note the addition of the `+` sign. Alternatively, you can use the
2066 `-f` flag to force the remote update, as in:
2067
2068 -------------------------------------------------
2069 $ git push -f ssh://yourserver.com/~you/proj.git master
2070 -------------------------------------------------
2071
2072 Normally whenever a branch head in a public repository is modified, it
2073 is modified to point to a descendant of the commit that it pointed to
2074 before. By forcing a push in this situation, you break that convention.
2075 (See <<problems-With-rewriting-history>>.)
2076
2077 Nevertheless, this is a common practice for people that need a simple
2078 way to publish a work-in-progress patch series, and it is an acceptable
2079 compromise as long as you warn other developers that this is how you
2080 intend to manage the branch.
2081
2082 It's also possible for a push to fail in this way when other people have
2083 the right to push to the same repository. In that case, the correct
2084 solution is to retry the push after first updating your work: either by a
2085 pull, or by a fetch followed by a rebase; see the
2086 <<setting-up-a-shared-repository,next section>> and
2087 linkgit:gitcvs-migration[7] for more.
2088
2089 [[setting-up-a-shared-repository]]
2090 Setting up a shared repository
2091 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2092
2093 Another way to collaborate is by using a model similar to that
2094 commonly used in CVS, where several developers with special rights
2095 all push to and pull from a single shared repository. See
2096 linkgit:gitcvs-migration[7] for instructions on how to
2097 set this up.
2098
2099 However, while there is nothing wrong with git's support for shared
2100 repositories, this mode of operation is not generally recommended,
2101 simply because the mode of collaboration that git supports--by
2102 exchanging patches and pulling from public repositories--has so many
2103 advantages over the central shared repository:
2104
2105 - Git's ability to quickly import and merge patches allows a
2106 single maintainer to process incoming changes even at very
2107 high rates. And when that becomes too much, `git pull` provides
2108 an easy way for that maintainer to delegate this job to other
2109 maintainers while still allowing optional review of incoming
2110 changes.
2111 - Since every developer's repository has the same complete copy
2112 of the project history, no repository is special, and it is
2113 trivial for another developer to take over maintenance of a
2114 project, either by mutual agreement, or because a maintainer
2115 becomes unresponsive or difficult to work with.
2116 - The lack of a central group of "committers" means there is
2117 less need for formal decisions about who is "in" and who is
2118 "out".
2119
2120 [[setting-up-gitweb]]
2121 Allowing web browsing of a repository
2122 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2123
2124 The gitweb cgi script provides users an easy way to browse your
2125 project's files and history without having to install git; see the file
2126 gitweb/INSTALL in the git source tree for instructions on setting it up.
2127
2128 [[sharing-development-examples]]
2129 Examples
2130 --------
2131
2132 [[maintaining-topic-branches]]
2133 Maintaining topic branches for a Linux subsystem maintainer
2134 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2135
2136 This describes how Tony Luck uses git in his role as maintainer of the
2137 IA64 architecture for the Linux kernel.
2138
2139 He uses two public branches:
2140
2141 - A "test" tree into which patches are initially placed so that they
2142 can get some exposure when integrated with other ongoing development.
2143 This tree is available to Andrew for pulling into -mm whenever he
2144 wants.
2145
2146 - A "release" tree into which tested patches are moved for final sanity
2147 checking, and as a vehicle to send them upstream to Linus (by sending
2148 him a "please pull" request.)
2149
2150 He also uses a set of temporary branches ("topic branches"), each
2151 containing a logical grouping of patches.
2152
2153 To set this up, first create your work tree by cloning Linus's public
2154 tree:
2155
2156 -------------------------------------------------
2157 $ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git work
2158 $ cd work
2159 -------------------------------------------------
2160
2161 Linus's tree will be stored in the remote-tracking branch named origin/master,
2162 and can be updated using linkgit:git-fetch[1]; you can track other
2163 public trees using linkgit:git-remote[1] to set up a "remote" and
2164 linkgit:git-fetch[1] to keep them up-to-date; see
2165 <<repositories-and-branches>>.
2166
2167 Now create the branches in which you are going to work; these start out
2168 at the current tip of origin/master branch, and should be set up (using
2169 the --track option to linkgit:git-branch[1]) to merge changes in from
2170 Linus by default.
2171
2172 -------------------------------------------------
2173 $ git branch --track test origin/master
2174 $ git branch --track release origin/master
2175 -------------------------------------------------
2176
2177 These can be easily kept up to date using linkgit:git-pull[1].
2178
2179 -------------------------------------------------
2180 $ git checkout test && git pull
2181 $ git checkout release && git pull
2182 -------------------------------------------------
2183
2184 Important note! If you have any local changes in these branches, then
2185 this merge will create a commit object in the history (with no local
2186 changes git will simply do a "fast-forward" merge). Many people dislike
2187 the "noise" that this creates in the Linux history, so you should avoid
2188 doing this capriciously in the "release" branch, as these noisy commits
2189 will become part of the permanent history when you ask Linus to pull
2190 from the release branch.
2191
2192 A few configuration variables (see linkgit:git-config[1]) can
2193 make it easy to push both branches to your public tree. (See
2194 <<setting-up-a-public-repository>>.)
2195
2196 -------------------------------------------------
2197 $ cat >> .git/config <<EOF
2198 [remote "mytree"]
2199 url = master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6.git
2200 push = release
2201 push = test
2202 EOF
2203 -------------------------------------------------
2204
2205 Then you can push both the test and release trees using
2206 linkgit:git-push[1]:
2207
2208 -------------------------------------------------
2209 $ git push mytree
2210 -------------------------------------------------
2211
2212 or push just one of the test and release branches using:
2213
2214 -------------------------------------------------
2215 $ git push mytree test
2216 -------------------------------------------------
2217
2218 or
2219
2220 -------------------------------------------------
2221 $ git push mytree release
2222 -------------------------------------------------
2223
2224 Now to apply some patches from the community. Think of a short
2225 snappy name for a branch to hold this patch (or related group of
2226 patches), and create a new branch from a recent stable tag of
2227 Linus's branch. Picking a stable base for your branch will:
2228 1) help you: by avoiding inclusion of unrelated and perhaps lightly
2229 tested changes
2230 2) help future bug hunters that use "git bisect" to find problems
2231
2232 -------------------------------------------------
2233 $ git checkout -b speed-up-spinlocks v2.6.35
2234 -------------------------------------------------
2235
2236 Now you apply the patch(es), run some tests, and commit the change(s). If
2237 the patch is a multi-part series, then you should apply each as a separate
2238 commit to this branch.
2239
2240 -------------------------------------------------
2241 $ ... patch ... test ... commit [ ... patch ... test ... commit ]*
2242 -------------------------------------------------
2243
2244 When you are happy with the state of this change, you can pull it into the
2245 "test" branch in preparation to make it public:
2246
2247 -------------------------------------------------
2248 $ git checkout test && git pull . speed-up-spinlocks
2249 -------------------------------------------------
2250
2251 It is unlikely that you would have any conflicts here ... but you might if you
2252 spent a while on this step and had also pulled new versions from upstream.
2253
2254 Some time later when enough time has passed and testing done, you can pull the
2255 same branch into the "release" tree ready to go upstream. This is where you
2256 see the value of keeping each patch (or patch series) in its own branch. It
2257 means that the patches can be moved into the "release" tree in any order.
2258
2259 -------------------------------------------------
2260 $ git checkout release && git pull . speed-up-spinlocks
2261 -------------------------------------------------
2262
2263 After a while, you will have a number of branches, and despite the
2264 well chosen names you picked for each of them, you may forget what
2265 they are for, or what status they are in. To get a reminder of what
2266 changes are in a specific branch, use:
2267
2268 -------------------------------------------------
2269 $ git log linux..branchname | git shortlog
2270 -------------------------------------------------
2271
2272 To see whether it has already been merged into the test or release branches,
2273 use:
2274
2275 -------------------------------------------------
2276 $ git log test..branchname
2277 -------------------------------------------------
2278
2279 or
2280
2281 -------------------------------------------------
2282 $ git log release..branchname
2283 -------------------------------------------------
2284
2285 (If this branch has not yet been merged, you will see some log entries.
2286 If it has been merged, then there will be no output.)
2287
2288 Once a patch completes the great cycle (moving from test to release,
2289 then pulled by Linus, and finally coming back into your local
2290 "origin/master" branch), the branch for this change is no longer needed.
2291 You detect this when the output from:
2292
2293 -------------------------------------------------
2294 $ git log origin..branchname
2295 -------------------------------------------------
2296
2297 is empty. At this point the branch can be deleted:
2298
2299 -------------------------------------------------
2300 $ git branch -d branchname
2301 -------------------------------------------------
2302
2303 Some changes are so trivial that it is not necessary to create a separate
2304 branch and then merge into each of the test and release branches. For
2305 these changes, just apply directly to the "release" branch, and then
2306 merge that into the "test" branch.
2307
2308 To create diffstat and shortlog summaries of changes to include in a "please
2309 pull" request to Linus you can use:
2310
2311 -------------------------------------------------
2312 $ git diff --stat origin..release
2313 -------------------------------------------------
2314
2315 and
2316
2317 -------------------------------------------------
2318 $ git log -p origin..release | git shortlog
2319 -------------------------------------------------
2320
2321 Here are some of the scripts that simplify all this even further.
2322
2323 -------------------------------------------------
2324 ==== update script ====
2325 # Update a branch in my GIT tree. If the branch to be updated
2326 # is origin, then pull from kernel.org. Otherwise merge
2327 # origin/master branch into test|release branch
2328
2329 case "$1" in
2330 test|release)
2331 git checkout $1 && git pull . origin
2332 ;;
2333 origin)
2334 before=$(git rev-parse refs/remotes/origin/master)
2335 git fetch origin
2336 after=$(git rev-parse refs/remotes/origin/master)
2337 if [ $before != $after ]
2338 then
2339 git log $before..$after | git shortlog
2340 fi
2341 ;;
2342 *)
2343 echo "Usage: $0 origin|test|release" 1>&2
2344 exit 1
2345 ;;
2346 esac
2347 -------------------------------------------------
2348
2349 -------------------------------------------------
2350 ==== merge script ====
2351 # Merge a branch into either the test or release branch
2352
2353 pname=$0
2354
2355 usage()
2356 {
2357 echo "Usage: $pname branch test|release" 1>&2
2358 exit 1
2359 }
2360
2361 git show-ref -q --verify -- refs/heads/"$1" || {
2362 echo "Can't see branch <$1>" 1>&2
2363 usage
2364 }
2365
2366 case "$2" in
2367 test|release)
2368 if [ $(git log $2..$1 | wc -c) -eq 0 ]
2369 then
2370 echo $1 already merged into $2 1>&2
2371 exit 1
2372 fi
2373 git checkout $2 && git pull . $1
2374 ;;
2375 *)
2376 usage
2377 ;;
2378 esac
2379 -------------------------------------------------
2380
2381 -------------------------------------------------
2382 ==== status script ====
2383 # report on status of my ia64 GIT tree
2384
2385 gb=$(tput setab 2)
2386 rb=$(tput setab 1)
2387 restore=$(tput setab 9)
2388
2389 if [ `git rev-list test..release | wc -c` -gt 0 ]
2390 then
2391 echo $rb Warning: commits in release that are not in test $restore
2392 git log test..release
2393 fi
2394
2395 for branch in `git show-ref --heads | sed 's|^.*/||'`
2396 do
2397 if [ $branch = test -o $branch = release ]
2398 then
2399 continue
2400 fi
2401
2402 echo -n $gb ======= $branch ====== $restore " "
2403 status=
2404 for ref in test release origin/master
2405 do
2406 if [ `git rev-list $ref..$branch | wc -c` -gt 0 ]
2407 then
2408 status=$status${ref:0:1}
2409 fi
2410 done
2411 case $status in
2412 trl)
2413 echo $rb Need to pull into test $restore
2414 ;;
2415 rl)
2416 echo "In test"
2417 ;;
2418 l)
2419 echo "Waiting for linus"
2420 ;;
2421 "")
2422 echo $rb All done $restore
2423 ;;
2424 *)
2425 echo $rb "<$status>" $restore
2426 ;;
2427 esac
2428 git log origin/master..$branch | git shortlog
2429 done
2430 -------------------------------------------------
2431
2432
2433 [[cleaning-up-history]]
2434 Rewriting history and maintaining patch series
2435 ==============================================
2436
2437 Normally commits are only added to a project, never taken away or
2438 replaced. Git is designed with this assumption, and violating it will
2439 cause git's merge machinery (for example) to do the wrong thing.
2440
2441 However, there is a situation in which it can be useful to violate this
2442 assumption.
2443
2444 [[patch-series]]
2445 Creating the perfect patch series
2446 ---------------------------------
2447
2448 Suppose you are a contributor to a large project, and you want to add a
2449 complicated feature, and to present it to the other developers in a way
2450 that makes it easy for them to read your changes, verify that they are
2451 correct, and understand why you made each change.
2452
2453 If you present all of your changes as a single patch (or commit), they
2454 may find that it is too much to digest all at once.
2455
2456 If you present them with the entire history of your work, complete with
2457 mistakes, corrections, and dead ends, they may be overwhelmed.
2458
2459 So the ideal is usually to produce a series of patches such that:
2460
2461 1. Each patch can be applied in order.
2462
2463 2. Each patch includes a single logical change, together with a
2464 message explaining the change.
2465
2466 3. No patch introduces a regression: after applying any initial
2467 part of the series, the resulting project still compiles and
2468 works, and has no bugs that it didn't have before.
2469
2470 4. The complete series produces the same end result as your own
2471 (probably much messier!) development process did.
2472
2473 We will introduce some tools that can help you do this, explain how to
2474 use them, and then explain some of the problems that can arise because
2475 you are rewriting history.
2476
2477 [[using-git-rebase]]
2478 Keeping a patch series up to date using git rebase
2479 --------------------------------------------------
2480
2481 Suppose that you create a branch "mywork" on a remote-tracking branch
2482 "origin", and create some commits on top of it:
2483
2484 -------------------------------------------------
2485 $ git checkout -b mywork origin
2486 $ vi file.txt
2487 $ git commit
2488 $ vi otherfile.txt
2489 $ git commit
2490 ...
2491 -------------------------------------------------
2492
2493 You have performed no merges into mywork, so it is just a simple linear
2494 sequence of patches on top of "origin":
2495
2496 ................................................
2497 o--o--O <-- origin
2498 \
2499 a--b--c <-- mywork
2500 ................................................
2501
2502 Some more interesting work has been done in the upstream project, and
2503 "origin" has advanced:
2504
2505 ................................................
2506 o--o--O--o--o--o <-- origin
2507 \
2508 a--b--c <-- mywork
2509 ................................................
2510
2511 At this point, you could use "pull" to merge your changes back in;
2512 the result would create a new merge commit, like this:
2513
2514 ................................................
2515 o--o--O--o--o--o <-- origin
2516 \ \
2517 a--b--c--m <-- mywork
2518 ................................................
2519
2520 However, if you prefer to keep the history in mywork a simple series of
2521 commits without any merges, you may instead choose to use
2522 linkgit:git-rebase[1]:
2523
2524 -------------------------------------------------
2525 $ git checkout mywork
2526 $ git rebase origin
2527 -------------------------------------------------
2528
2529 This will remove each of your commits from mywork, temporarily saving
2530 them as patches (in a directory named ".git/rebase-apply"), update mywork to
2531 point at the latest version of origin, then apply each of the saved
2532 patches to the new mywork. The result will look like:
2533
2534
2535 ................................................
2536 o--o--O--o--o--o <-- origin
2537 \
2538 a'--b'--c' <-- mywork
2539 ................................................
2540
2541 In the process, it may discover conflicts. In that case it will stop
2542 and allow you to fix the conflicts; after fixing conflicts, use `git add`
2543 to update the index with those contents, and then, instead of
2544 running `git commit`, just run
2545
2546 -------------------------------------------------
2547 $ git rebase --continue
2548 -------------------------------------------------
2549
2550 and git will continue applying the rest of the patches.
2551
2552 At any point you may use the `--abort` option to abort this process and
2553 return mywork to the state it had before you started the rebase:
2554
2555 -------------------------------------------------
2556 $ git rebase --abort
2557 -------------------------------------------------
2558
2559 If you need to reorder or edit a number of commits in a branch, it may
2560 be easier to use `git rebase -i`, which allows you to reorder and
2561 squash commits, as well as marking them for individual editing during
2562 the rebase. See <<interactive-rebase>> for details, and
2563 <<reordering-patch-series>> for alternatives.
2564
2565 [[rewriting-one-commit]]
2566 Rewriting a single commit
2567 -------------------------
2568
2569 We saw in <<fixing-a-mistake-by-rewriting-history>> that you can replace the
2570 most recent commit using
2571
2572 -------------------------------------------------
2573 $ git commit --amend
2574 -------------------------------------------------
2575
2576 which will replace the old commit by a new commit incorporating your
2577 changes, giving you a chance to edit the old commit message first.
2578 This is useful for fixing typos in your last commit, or for adjusting
2579 the patch contents of a poorly staged commit.
2580
2581 If you need to amend commits from deeper in your history, you can
2582 use <<interactive-rebase,interactive rebase's `edit` instruction>>.
2583
2584 [[reordering-patch-series]]
2585 Reordering or selecting from a patch series
2586 -------------------------------------------
2587
2588 Sometimes you want to edit a commit deeper in your history. One
2589 approach is to use `git format-patch` to create a series of patches
2590 and then reset the state to before the patches:
2591
2592 -------------------------------------------------
2593 $ git format-patch origin
2594 $ git reset --hard origin
2595 -------------------------------------------------
2596
2597 Then modify, reorder, or eliminate patches as needed before applying
2598 them again with linkgit:git-am[1]:
2599
2600 -------------------------------------------------
2601 $ git am *.patch
2602 -------------------------------------------------
2603
2604 [[interactive-rebase]]
2605 Using interactive rebases
2606 -------------------------
2607
2608 You can also edit a patch series with an interactive rebase. This is
2609 the same as <<reordering-patch-series,reordering a patch series using
2610 `format-patch`>>, so use whichever interface you like best.
2611
2612 Rebase your current HEAD on the last commit you want to retain as-is.
2613 For example, if you want to reorder the last 5 commits, use:
2614
2615 -------------------------------------------------
2616 $ git rebase -i HEAD~5
2617 -------------------------------------------------
2618
2619 This will open your editor with a list of steps to be taken to perform
2620 your rebase.
2621
2622 -------------------------------------------------
2623 pick deadbee The oneline of this commit
2624 pick fa1afe1 The oneline of the next commit
2625 ...
2626
2627 # Rebase c0ffeee..deadbee onto c0ffeee
2628 #
2629 # Commands:
2630 # p, pick = use commit
2631 # r, reword = use commit, but edit the commit message
2632 # e, edit = use commit, but stop for amending
2633 # s, squash = use commit, but meld into previous commit
2634 # f, fixup = like "squash", but discard this commit's log message
2635 # x, exec = run command (the rest of the line) using shell
2636 #
2637 # These lines can be re-ordered; they are executed from top to bottom.
2638 #
2639 # If you remove a line here THAT COMMIT WILL BE LOST.
2640 #
2641 # However, if you remove everything, the rebase will be aborted.
2642 #
2643 # Note that empty commits are commented out
2644 -------------------------------------------------
2645
2646 As explained in the comments, you can reorder commits, squash them
2647 together, edit commit messages, etc. by editing the list. Once you
2648 are satisfied, save the list and close your editor, and the rebase
2649 will begin.
2650
2651 The rebase will stop where `pick` has been replaced with `edit` or
2652 when a step in the list fails to mechanically resolve conflicts and
2653 needs your help. When you are done editing and/or resolving conflicts
2654 you can continue with `git rebase --continue`. If you decide that
2655 things are getting too hairy, you can always bail out with `git rebase
2656 --abort`. Even after the rebase is complete, you can still recover
2657 the original branch by using the <<reflogs,reflog>>.
2658
2659 For a more detailed discussion of the procedure and additional tips,
2660 see the "INTERACTIVE MODE" section of linkgit:git-rebase[1].
2661
2662 [[patch-series-tools]]
2663 Other tools
2664 -----------
2665
2666 There are numerous other tools, such as StGit, which exist for the
2667 purpose of maintaining a patch series. These are outside of the scope of
2668 this manual.
2669
2670 [[problems-With-rewriting-history]]
2671 Problems with rewriting history
2672 -------------------------------
2673
2674 The primary problem with rewriting the history of a branch has to do
2675 with merging. Suppose somebody fetches your branch and merges it into
2676 their branch, with a result something like this:
2677
2678 ................................................
2679 o--o--O--o--o--o <-- origin
2680 \ \
2681 t--t--t--m <-- their branch:
2682 ................................................
2683
2684 Then suppose you modify the last three commits:
2685
2686 ................................................
2687 o--o--o <-- new head of origin
2688 /
2689 o--o--O--o--o--o <-- old head of origin
2690 ................................................
2691
2692 If we examined all this history together in one repository, it will
2693 look like:
2694
2695 ................................................
2696 o--o--o <-- new head of origin
2697 /
2698 o--o--O--o--o--o <-- old head of origin
2699 \ \
2700 t--t--t--m <-- their branch:
2701 ................................................
2702
2703 Git has no way of knowing that the new head is an updated version of
2704 the old head; it treats this situation exactly the same as it would if
2705 two developers had independently done the work on the old and new heads
2706 in parallel. At this point, if someone attempts to merge the new head
2707 in to their branch, git will attempt to merge together the two (old and
2708 new) lines of development, instead of trying to replace the old by the
2709 new. The results are likely to be unexpected.
2710
2711 You may still choose to publish branches whose history is rewritten,
2712 and it may be useful for others to be able to fetch those branches in
2713 order to examine or test them, but they should not attempt to pull such
2714 branches into their own work.
2715
2716 For true distributed development that supports proper merging,
2717 published branches should never be rewritten.
2718
2719 [[bisect-merges]]
2720 Why bisecting merge commits can be harder than bisecting linear history
2721 -----------------------------------------------------------------------
2722
2723 The linkgit:git-bisect[1] command correctly handles history that
2724 includes merge commits. However, when the commit that it finds is a
2725 merge commit, the user may need to work harder than usual to figure out
2726 why that commit introduced a problem.
2727
2728 Imagine this history:
2729
2730 ................................................
2731 ---Z---o---X---...---o---A---C---D
2732 \ /
2733 o---o---Y---...---o---B
2734 ................................................
2735
2736 Suppose that on the upper line of development, the meaning of one
2737 of the functions that exists at Z is changed at commit X. The
2738 commits from Z leading to A change both the function's
2739 implementation and all calling sites that exist at Z, as well
2740 as new calling sites they add, to be consistent. There is no
2741 bug at A.
2742
2743 Suppose that in the meantime on the lower line of development somebody
2744 adds a new calling site for that function at commit Y. The
2745 commits from Z leading to B all assume the old semantics of that
2746 function and the callers and the callee are consistent with each
2747 other. There is no bug at B, either.
2748
2749 Suppose further that the two development lines merge cleanly at C,
2750 so no conflict resolution is required.
2751
2752 Nevertheless, the code at C is broken, because the callers added
2753 on the lower line of development have not been converted to the new
2754 semantics introduced on the upper line of development. So if all
2755 you know is that D is bad, that Z is good, and that
2756 linkgit:git-bisect[1] identifies C as the culprit, how will you
2757 figure out that the problem is due to this change in semantics?
2758
2759 When the result of a `git bisect` is a non-merge commit, you should
2760 normally be able to discover the problem by examining just that commit.
2761 Developers can make this easy by breaking their changes into small
2762 self-contained commits. That won't help in the case above, however,
2763 because the problem isn't obvious from examination of any single
2764 commit; instead, a global view of the development is required. To
2765 make matters worse, the change in semantics in the problematic
2766 function may be just one small part of the changes in the upper
2767 line of development.
2768
2769 On the other hand, if instead of merging at C you had rebased the
2770 history between Z to B on top of A, you would have gotten this
2771 linear history:
2772
2773 ................................................................
2774 ---Z---o---X--...---o---A---o---o---Y*--...---o---B*--D*
2775 ................................................................
2776
2777 Bisecting between Z and D* would hit a single culprit commit Y*,
2778 and understanding why Y* was broken would probably be easier.
2779
2780 Partly for this reason, many experienced git users, even when
2781 working on an otherwise merge-heavy project, keep the history
2782 linear by rebasing against the latest upstream version before
2783 publishing.
2784
2785 [[advanced-branch-management]]
2786 Advanced branch management
2787 ==========================
2788
2789 [[fetching-individual-branches]]
2790 Fetching individual branches
2791 ----------------------------
2792
2793 Instead of using linkgit:git-remote[1], you can also choose just
2794 to update one branch at a time, and to store it locally under an
2795 arbitrary name:
2796
2797 -------------------------------------------------
2798 $ git fetch origin todo:my-todo-work
2799 -------------------------------------------------
2800
2801 The first argument, "origin", just tells git to fetch from the
2802 repository you originally cloned from. The second argument tells git
2803 to fetch the branch named "todo" from the remote repository, and to
2804 store it locally under the name refs/heads/my-todo-work.
2805
2806 You can also fetch branches from other repositories; so
2807
2808 -------------------------------------------------
2809 $ git fetch git://example.com/proj.git master:example-master
2810 -------------------------------------------------
2811
2812 will create a new branch named "example-master" and store in it the
2813 branch named "master" from the repository at the given URL. If you
2814 already have a branch named example-master, it will attempt to
2815 <<fast-forwards,fast-forward>> to the commit given by example.com's
2816 master branch. In more detail:
2817
2818 [[fetch-fast-forwards]]
2819 git fetch and fast-forwards
2820 ---------------------------
2821
2822 In the previous example, when updating an existing branch, "git fetch"
2823 checks to make sure that the most recent commit on the remote
2824 branch is a descendant of the most recent commit on your copy of the
2825 branch before updating your copy of the branch to point at the new
2826 commit. Git calls this process a <<fast-forwards,fast-forward>>.
2827
2828 A fast-forward looks something like this:
2829
2830 ................................................
2831 o--o--o--o <-- old head of the branch
2832 \
2833 o--o--o <-- new head of the branch
2834 ................................................
2835
2836
2837 In some cases it is possible that the new head will *not* actually be
2838 a descendant of the old head. For example, the developer may have
2839 realized she made a serious mistake, and decided to backtrack,
2840 resulting in a situation like:
2841
2842 ................................................
2843 o--o--o--o--a--b <-- old head of the branch
2844 \
2845 o--o--o <-- new head of the branch
2846 ................................................
2847
2848 In this case, "git fetch" will fail, and print out a warning.
2849
2850 In that case, you can still force git to update to the new head, as
2851 described in the following section. However, note that in the
2852 situation above this may mean losing the commits labeled "a" and "b",
2853 unless you've already created a reference of your own pointing to
2854 them.
2855
2856 [[forcing-fetch]]
2857 Forcing git fetch to do non-fast-forward updates
2858 ------------------------------------------------
2859
2860 If git fetch fails because the new head of a branch is not a
2861 descendant of the old head, you may force the update with:
2862
2863 -------------------------------------------------
2864 $ git fetch git://example.com/proj.git +master:refs/remotes/example/master
2865 -------------------------------------------------
2866
2867 Note the addition of the "+" sign. Alternatively, you can use the "-f"
2868 flag to force updates of all the fetched branches, as in:
2869
2870 -------------------------------------------------
2871 $ git fetch -f origin
2872 -------------------------------------------------
2873
2874 Be aware that commits that the old version of example/master pointed at
2875 may be lost, as we saw in the previous section.
2876
2877 [[remote-branch-configuration]]
2878 Configuring remote-tracking branches
2879 ------------------------------------
2880
2881 We saw above that "origin" is just a shortcut to refer to the
2882 repository that you originally cloned from. This information is
2883 stored in git configuration variables, which you can see using
2884 linkgit:git-config[1]:
2885
2886 -------------------------------------------------
2887 $ git config -l
2888 core.repositoryformatversion=0
2889 core.filemode=true
2890 core.logallrefupdates=true
2891 remote.origin.url=git://git.kernel.org/pub/scm/git/git.git
2892 remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*
2893 branch.master.remote=origin
2894 branch.master.merge=refs/heads/master
2895 -------------------------------------------------
2896
2897 If there are other repositories that you also use frequently, you can
2898 create similar configuration options to save typing; for example,
2899
2900 -------------------------------------------------
2901 $ git remote add example git://example.com/proj.git
2902 -------------------------------------------------
2903
2904 adds the following to `.git/config`:
2905
2906 -------------------------------------------------
2907 [remote "example"]
2908 url = git://example.com/proj.git
2909 fetch = +refs/heads/*:refs/remotes/example/*
2910 -------------------------------------------------
2911
2912 Also note that the above configuration can be performed by directly
2913 editing the file `.git/config` instead of using linkgit:git-remote[1].
2914
2915 After configuring the remote, the following three commands will do the
2916 same thing:
2917
2918 -------------------------------------------------
2919 $ git fetch git://example.com/proj.git +refs/heads/*:refs/remotes/example/*
2920 $ git fetch example +refs/heads/*:refs/remotes/example/*
2921 $ git fetch example
2922 -------------------------------------------------
2923
2924 See linkgit:git-config[1] for more details on the configuration
2925 options mentioned above and linkgit:git-fetch[1] for more details on
2926 the refspec syntax.
2927
2928
2929 [[git-concepts]]
2930 Git concepts
2931 ============
2932
2933 Git is built on a small number of simple but powerful ideas. While it
2934 is possible to get things done without understanding them, you will find
2935 git much more intuitive if you do.
2936
2937 We start with the most important, the <<def_object_database,object
2938 database>> and the <<def_index,index>>.
2939
2940 [[the-object-database]]
2941 The Object Database
2942 -------------------
2943
2944
2945 We already saw in <<understanding-commits>> that all commits are stored
2946 under a 40-digit "object name". In fact, all the information needed to
2947 represent the history of a project is stored in objects with such names.
2948 In each case the name is calculated by taking the SHA-1 hash of the
2949 contents of the object. The SHA-1 hash is a cryptographic hash function.
2950 What that means to us is that it is impossible to find two different
2951 objects with the same name. This has a number of advantages; among
2952 others:
2953
2954 - Git can quickly determine whether two objects are identical or not,
2955 just by comparing names.
2956 - Since object names are computed the same way in every repository, the
2957 same content stored in two repositories will always be stored under
2958 the same name.
2959 - Git can detect errors when it reads an object, by checking that the
2960 object's name is still the SHA-1 hash of its contents.
2961
2962 (See <<object-details>> for the details of the object formatting and
2963 SHA-1 calculation.)
2964
2965 There are four different types of objects: "blob", "tree", "commit", and
2966 "tag".
2967
2968 - A <<def_blob_object,"blob" object>> is used to store file data.
2969 - A <<def_tree_object,"tree" object>> ties one or more
2970 "blob" objects into a directory structure. In addition, a tree object
2971 can refer to other tree objects, thus creating a directory hierarchy.
2972 - A <<def_commit_object,"commit" object>> ties such directory hierarchies
2973 together into a <<def_DAG,directed acyclic graph>> of revisions--each
2974 commit contains the object name of exactly one tree designating the
2975 directory hierarchy at the time of the commit. In addition, a commit
2976 refers to "parent" commit objects that describe the history of how we
2977 arrived at that directory hierarchy.
2978 - A <<def_tag_object,"tag" object>> symbolically identifies and can be
2979 used to sign other objects. It contains the object name and type of
2980 another object, a symbolic name (of course!) and, optionally, a
2981 signature.
2982
2983 The object types in some more detail:
2984
2985 [[commit-object]]
2986 Commit Object
2987 ~~~~~~~~~~~~~
2988
2989 The "commit" object links a physical state of a tree with a description
2990 of how we got there and why. Use the --pretty=raw option to
2991 linkgit:git-show[1] or linkgit:git-log[1] to examine your favorite
2992 commit:
2993
2994 ------------------------------------------------
2995 $ git show -s --pretty=raw 2be7fcb476
2996 commit 2be7fcb4764f2dbcee52635b91fedb1b3dcf7ab4
2997 tree fb3a8bdd0ceddd019615af4d57a53f43d8cee2bf
2998 parent 257a84d9d02e90447b149af58b271c19405edb6a
2999 author Dave Watson <dwatson@mimvista.com> 1187576872 -0400
3000 committer Junio C Hamano <gitster@pobox.com> 1187591163 -0700
3001
3002 Fix misspelling of 'suppress' in docs
3003
3004 Signed-off-by: Junio C Hamano <gitster@pobox.com>
3005 ------------------------------------------------
3006
3007 As you can see, a commit is defined by:
3008
3009 - a tree: The SHA-1 name of a tree object (as defined below), representing
3010 the contents of a directory at a certain point in time.
3011 - parent(s): The SHA-1 name(s) of some number of commits which represent the
3012 immediately previous step(s) in the history of the project. The
3013 example above has one parent; merge commits may have more than
3014 one. A commit with no parents is called a "root" commit, and
3015 represents the initial revision of a project. Each project must have
3016 at least one root. A project can also have multiple roots, though
3017 that isn't common (or necessarily a good idea).
3018 - an author: The name of the person responsible for this change, together
3019 with its date.
3020 - a committer: The name of the person who actually created the commit,
3021 with the date it was done. This may be different from the author, for
3022 example, if the author was someone who wrote a patch and emailed it
3023 to the person who used it to create the commit.
3024 - a comment describing this commit.
3025
3026 Note that a commit does not itself contain any information about what
3027 actually changed; all changes are calculated by comparing the contents
3028 of the tree referred to by this commit with the trees associated with
3029 its parents. In particular, git does not attempt to record file renames
3030 explicitly, though it can identify cases where the existence of the same
3031 file data at changing paths suggests a rename. (See, for example, the
3032 -M option to linkgit:git-diff[1]).
3033
3034 A commit is usually created by linkgit:git-commit[1], which creates a
3035 commit whose parent is normally the current HEAD, and whose tree is
3036 taken from the content currently stored in the index.
3037
3038 [[tree-object]]
3039 Tree Object
3040 ~~~~~~~~~~~
3041
3042 The ever-versatile linkgit:git-show[1] command can also be used to
3043 examine tree objects, but linkgit:git-ls-tree[1] will give you more
3044 details:
3045
3046 ------------------------------------------------
3047 $ git ls-tree fb3a8bdd0ce
3048 100644 blob 63c918c667fa005ff12ad89437f2fdc80926e21c .gitignore
3049 100644 blob 5529b198e8d14decbe4ad99db3f7fb632de0439d .mailmap
3050 100644 blob 6ff87c4664981e4397625791c8ea3bbb5f2279a3 COPYING
3051 040000 tree 2fb783e477100ce076f6bf57e4a6f026013dc745 Documentation
3052 100755 blob 3c0032cec592a765692234f1cba47dfdcc3a9200 GIT-VERSION-GEN
3053 100644 blob 289b046a443c0647624607d471289b2c7dcd470b INSTALL
3054 100644 blob 4eb463797adc693dc168b926b6932ff53f17d0b1 Makefile
3055 100644 blob 548142c327a6790ff8821d67c2ee1eff7a656b52 README
3056 ...
3057 ------------------------------------------------
3058
3059 As you can see, a tree object contains a list of entries, each with a
3060 mode, object type, SHA-1 name, and name, sorted by name. It represents
3061 the contents of a single directory tree.
3062
3063 The object type may be a blob, representing the contents of a file, or
3064 another tree, representing the contents of a subdirectory. Since trees
3065 and blobs, like all other objects, are named by the SHA-1 hash of their
3066 contents, two trees have the same SHA-1 name if and only if their
3067 contents (including, recursively, the contents of all subdirectories)
3068 are identical. This allows git to quickly determine the differences
3069 between two related tree objects, since it can ignore any entries with
3070 identical object names.
3071
3072 (Note: in the presence of submodules, trees may also have commits as
3073 entries. See <<submodules>> for documentation.)
3074
3075 Note that the files all have mode 644 or 755: git actually only pays
3076 attention to the executable bit.
3077
3078 [[blob-object]]
3079 Blob Object
3080 ~~~~~~~~~~~
3081
3082 You can use linkgit:git-show[1] to examine the contents of a blob; take,
3083 for example, the blob in the entry for "COPYING" from the tree above:
3084
3085 ------------------------------------------------
3086 $ git show 6ff87c4664
3087
3088 Note that the only valid version of the GPL as far as this project
3089 is concerned is _this_ particular version of the license (ie v2, not
3090 v2.2 or v3.x or whatever), unless explicitly otherwise stated.
3091 ...
3092 ------------------------------------------------
3093
3094 A "blob" object is nothing but a binary blob of data. It doesn't refer
3095 to anything else or have attributes of any kind.
3096
3097 Since the blob is entirely defined by its data, if two files in a
3098 directory tree (or in multiple different versions of the repository)
3099 have the same contents, they will share the same blob object. The object
3100 is totally independent of its location in the directory tree, and
3101 renaming a file does not change the object that file is associated with.
3102
3103 Note that any tree or blob object can be examined using
3104 linkgit:git-show[1] with the <revision>:<path> syntax. This can
3105 sometimes be useful for browsing the contents of a tree that is not
3106 currently checked out.
3107
3108 [[trust]]
3109 Trust
3110 ~~~~~
3111
3112 If you receive the SHA-1 name of a blob from one source, and its contents
3113 from another (possibly untrusted) source, you can still trust that those
3114 contents are correct as long as the SHA-1 name agrees. This is because
3115 the SHA-1 is designed so that it is infeasible to find different contents
3116 that produce the same hash.
3117
3118 Similarly, you need only trust the SHA-1 name of a top-level tree object
3119 to trust the contents of the entire directory that it refers to, and if
3120 you receive the SHA-1 name of a commit from a trusted source, then you
3121 can easily verify the entire history of commits reachable through
3122 parents of that commit, and all of those contents of the trees referred
3123 to by those commits.
3124
3125 So to introduce some real trust in the system, the only thing you need
3126 to do is to digitally sign just 'one' special note, which includes the
3127 name of a top-level commit. Your digital signature shows others
3128 that you trust that commit, and the immutability of the history of
3129 commits tells others that they can trust the whole history.
3130
3131 In other words, you can easily validate a whole archive by just
3132 sending out a single email that tells the people the name (SHA-1 hash)
3133 of the top commit, and digitally sign that email using something
3134 like GPG/PGP.
3135
3136 To assist in this, git also provides the tag object...
3137
3138 [[tag-object]]
3139 Tag Object
3140 ~~~~~~~~~~
3141
3142 A tag object contains an object, object type, tag name, the name of the
3143 person ("tagger") who created the tag, and a message, which may contain
3144 a signature, as can be seen using linkgit:git-cat-file[1]:
3145
3146 ------------------------------------------------
3147 $ git cat-file tag v1.5.0
3148 object 437b1b20df4b356c9342dac8d38849f24ef44f27
3149 type commit
3150 tag v1.5.0
3151 tagger Junio C Hamano <junkio@cox.net> 1171411200 +0000
3152
3153 GIT 1.5.0
3154 -----BEGIN PGP SIGNATURE-----
3155 Version: GnuPG v1.4.6 (GNU/Linux)
3156
3157 iD8DBQBF0lGqwMbZpPMRm5oRAuRiAJ9ohBLd7s2kqjkKlq1qqC57SbnmzQCdG4ui
3158 nLE/L9aUXdWeTFPron96DLA=
3159 =2E+0
3160 -----END PGP SIGNATURE-----
3161 ------------------------------------------------
3162
3163 See the linkgit:git-tag[1] command to learn how to create and verify tag
3164 objects. (Note that linkgit:git-tag[1] can also be used to create
3165 "lightweight tags", which are not tag objects at all, but just simple
3166 references whose names begin with "refs/tags/").
3167
3168 [[pack-files]]
3169 How git stores objects efficiently: pack files
3170 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3171
3172 Newly created objects are initially created in a file named after the
3173 object's SHA-1 hash (stored in .git/objects).
3174
3175 Unfortunately this system becomes inefficient once a project has a
3176 lot of objects. Try this on an old project:
3177
3178 ------------------------------------------------
3179 $ git count-objects
3180 6930 objects, 47620 kilobytes
3181 ------------------------------------------------
3182
3183 The first number is the number of objects which are kept in
3184 individual files. The second is the amount of space taken up by
3185 those "loose" objects.
3186
3187 You can save space and make git faster by moving these loose objects in
3188 to a "pack file", which stores a group of objects in an efficient
3189 compressed format; the details of how pack files are formatted can be
3190 found in link:technical/pack-format.txt[technical/pack-format.txt].
3191
3192 To put the loose objects into a pack, just run git repack:
3193
3194 ------------------------------------------------
3195 $ git repack
3196 Generating pack...
3197 Done counting 6020 objects.
3198 Deltifying 6020 objects.
3199 100% (6020/6020) done
3200 Writing 6020 objects.
3201 100% (6020/6020) done
3202 Total 6020, written 6020 (delta 4070), reused 0 (delta 0)
3203 Pack pack-3e54ad29d5b2e05838c75df582c65257b8d08e1c created.
3204 ------------------------------------------------
3205
3206 You can then run
3207
3208 ------------------------------------------------
3209 $ git prune
3210 ------------------------------------------------
3211
3212 to remove any of the "loose" objects that are now contained in the
3213 pack. This will also remove any unreferenced objects (which may be
3214 created when, for example, you use "git reset" to remove a commit).
3215 You can verify that the loose objects are gone by looking at the
3216 .git/objects directory or by running
3217
3218 ------------------------------------------------
3219 $ git count-objects
3220 0 objects, 0 kilobytes
3221 ------------------------------------------------
3222
3223 Although the object files are gone, any commands that refer to those
3224 objects will work exactly as they did before.
3225
3226 The linkgit:git-gc[1] command performs packing, pruning, and more for
3227 you, so is normally the only high-level command you need.
3228
3229 [[dangling-objects]]
3230 Dangling objects
3231 ~~~~~~~~~~~~~~~~
3232
3233 The linkgit:git-fsck[1] command will sometimes complain about dangling
3234 objects. They are not a problem.
3235
3236 The most common cause of dangling objects is that you've rebased a
3237 branch, or you have pulled from somebody else who rebased a branch--see
3238 <<cleaning-up-history>>. In that case, the old head of the original
3239 branch still exists, as does everything it pointed to. The branch
3240 pointer itself just doesn't, since you replaced it with another one.
3241
3242 There are also other situations that cause dangling objects. For
3243 example, a "dangling blob" may arise because you did a "git add" of a
3244 file, but then, before you actually committed it and made it part of the
3245 bigger picture, you changed something else in that file and committed
3246 that *updated* thing--the old state that you added originally ends up
3247 not being pointed to by any commit or tree, so it's now a dangling blob
3248 object.
3249
3250 Similarly, when the "recursive" merge strategy runs, and finds that
3251 there are criss-cross merges and thus more than one merge base (which is
3252 fairly unusual, but it does happen), it will generate one temporary
3253 midway tree (or possibly even more, if you had lots of criss-crossing
3254 merges and more than two merge bases) as a temporary internal merge
3255 base, and again, those are real objects, but the end result will not end
3256 up pointing to them, so they end up "dangling" in your repository.
3257
3258 Generally, dangling objects aren't anything to worry about. They can
3259 even be very useful: if you screw something up, the dangling objects can
3260 be how you recover your old tree (say, you did a rebase, and realized
3261 that you really didn't want to--you can look at what dangling objects
3262 you have, and decide to reset your head to some old dangling state).
3263
3264 For commits, you can just use:
3265
3266 ------------------------------------------------
3267 $ gitk <dangling-commit-sha-goes-here> --not --all
3268 ------------------------------------------------
3269
3270 This asks for all the history reachable from the given commit but not
3271 from any branch, tag, or other reference. If you decide it's something
3272 you want, you can always create a new reference to it, e.g.,
3273
3274 ------------------------------------------------
3275 $ git branch recovered-branch <dangling-commit-sha-goes-here>
3276 ------------------------------------------------
3277
3278 For blobs and trees, you can't do the same, but you can still examine
3279 them. You can just do
3280
3281 ------------------------------------------------
3282 $ git show <dangling-blob/tree-sha-goes-here>
3283 ------------------------------------------------
3284
3285 to show what the contents of the blob were (or, for a tree, basically
3286 what the "ls" for that directory was), and that may give you some idea
3287 of what the operation was that left that dangling object.
3288
3289 Usually, dangling blobs and trees aren't very interesting. They're
3290 almost always the result of either being a half-way mergebase (the blob
3291 will often even have the conflict markers from a merge in it, if you
3292 have had conflicting merges that you fixed up by hand), or simply
3293 because you interrupted a "git fetch" with ^C or something like that,
3294 leaving _some_ of the new objects in the object database, but just
3295 dangling and useless.
3296
3297 Anyway, once you are sure that you're not interested in any dangling
3298 state, you can just prune all unreachable objects:
3299
3300 ------------------------------------------------
3301 $ git prune
3302 ------------------------------------------------
3303
3304 and they'll be gone. But you should only run "git prune" on a quiescent
3305 repository--it's kind of like doing a filesystem fsck recovery: you
3306 don't want to do that while the filesystem is mounted.
3307
3308 (The same is true of "git fsck" itself, btw, but since
3309 `git fsck` never actually *changes* the repository, it just reports
3310 on what it found, `git fsck` itself is never 'dangerous' to run.
3311 Running it while somebody is actually changing the repository can cause
3312 confusing and scary messages, but it won't actually do anything bad. In
3313 contrast, running "git prune" while somebody is actively changing the
3314 repository is a *BAD* idea).
3315
3316 [[recovering-from-repository-corruption]]
3317 Recovering from repository corruption
3318 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3319
3320 By design, git treats data trusted to it with caution. However, even in
3321 the absence of bugs in git itself, it is still possible that hardware or
3322 operating system errors could corrupt data.
3323
3324 The first defense against such problems is backups. You can back up a
3325 git directory using clone, or just using cp, tar, or any other backup
3326 mechanism.
3327
3328 As a last resort, you can search for the corrupted objects and attempt
3329 to replace them by hand. Back up your repository before attempting this
3330 in case you corrupt things even more in the process.
3331
3332 We'll assume that the problem is a single missing or corrupted blob,
3333 which is sometimes a solvable problem. (Recovering missing trees and
3334 especially commits is *much* harder).
3335
3336 Before starting, verify that there is corruption, and figure out where
3337 it is with linkgit:git-fsck[1]; this may be time-consuming.
3338
3339 Assume the output looks like this:
3340
3341 ------------------------------------------------
3342 $ git fsck --full --no-dangling
3343 broken link from tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff8
3344 to blob 4b9458b3786228369c63936db65827de3cc06200
3345 missing blob 4b9458b3786228369c63936db65827de3cc06200
3346 ------------------------------------------------
3347
3348 Now you know that blob 4b9458b3 is missing, and that the tree 2d9263c6
3349 points to it. If you could find just one copy of that missing blob
3350 object, possibly in some other repository, you could move it into
3351 .git/objects/4b/9458b3... and be done. Suppose you can't. You can
3352 still examine the tree that pointed to it with linkgit:git-ls-tree[1],
3353 which might output something like:
3354
3355 ------------------------------------------------
3356 $ git ls-tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff8
3357 100644 blob 8d14531846b95bfa3564b58ccfb7913a034323b8 .gitignore
3358 100644 blob ebf9bf84da0aab5ed944264a5db2a65fe3a3e883 .mailmap
3359 100644 blob ca442d313d86dc67e0a2e5d584b465bd382cbf5c COPYING
3360 ...
3361 100644 blob 4b9458b3786228369c63936db65827de3cc06200 myfile
3362 ...
3363 ------------------------------------------------
3364
3365 So now you know that the missing blob was the data for a file named
3366 "myfile". And chances are you can also identify the directory--let's
3367 say it's in "somedirectory". If you're lucky the missing copy might be
3368 the same as the copy you have checked out in your working tree at
3369 "somedirectory/myfile"; you can test whether that's right with
3370 linkgit:git-hash-object[1]:
3371
3372 ------------------------------------------------
3373 $ git hash-object -w somedirectory/myfile
3374 ------------------------------------------------
3375
3376 which will create and store a blob object with the contents of
3377 somedirectory/myfile, and output the SHA-1 of that object. if you're
3378 extremely lucky it might be 4b9458b3786228369c63936db65827de3cc06200, in
3379 which case you've guessed right, and the corruption is fixed!
3380
3381 Otherwise, you need more information. How do you tell which version of
3382 the file has been lost?
3383
3384 The easiest way to do this is with:
3385
3386 ------------------------------------------------
3387 $ git log --raw --all --full-history -- somedirectory/myfile
3388 ------------------------------------------------
3389
3390 Because you're asking for raw output, you'll now get something like
3391
3392 ------------------------------------------------
3393 commit abc
3394 Author:
3395 Date:
3396 ...
3397 :100644 100644 4b9458b... newsha... M somedirectory/myfile
3398
3399
3400 commit xyz
3401 Author:
3402 Date:
3403
3404 ...
3405 :100644 100644 oldsha... 4b9458b... M somedirectory/myfile
3406 ------------------------------------------------
3407
3408 This tells you that the immediately following version of the file was
3409 "newsha", and that the immediately preceding version was "oldsha".
3410 You also know the commit messages that went with the change from oldsha
3411 to 4b9458b and with the change from 4b9458b to newsha.
3412
3413 If you've been committing small enough changes, you may now have a good
3414 shot at reconstructing the contents of the in-between state 4b9458b.
3415
3416 If you can do that, you can now recreate the missing object with
3417
3418 ------------------------------------------------
3419 $ git hash-object -w <recreated-file>
3420 ------------------------------------------------
3421
3422 and your repository is good again!
3423
3424 (Btw, you could have ignored the fsck, and started with doing a
3425
3426 ------------------------------------------------
3427 $ git log --raw --all
3428 ------------------------------------------------
3429
3430 and just looked for the sha of the missing object (4b9458b..) in that
3431 whole thing. It's up to you--Git does *have* a lot of information, it is
3432 just missing one particular blob version.
3433
3434 [[the-index]]
3435 The index
3436 -----------
3437
3438 The index is a binary file (generally kept in .git/index) containing a
3439 sorted list of path names, each with permissions and the SHA-1 of a blob
3440 object; linkgit:git-ls-files[1] can show you the contents of the index:
3441
3442 -------------------------------------------------
3443 $ git ls-files --stage
3444 100644 63c918c667fa005ff12ad89437f2fdc80926e21c 0 .gitignore
3445 100644 5529b198e8d14decbe4ad99db3f7fb632de0439d 0 .mailmap
3446 100644 6ff87c4664981e4397625791c8ea3bbb5f2279a3 0 COPYING
3447 100644 a37b2152bd26be2c2289e1f57a292534a51a93c7 0 Documentation/.gitignore
3448 100644 fbefe9a45b00a54b58d94d06eca48b03d40a50e0 0 Documentation/Makefile
3449 ...
3450 100644 2511aef8d89ab52be5ec6a5e46236b4b6bcd07ea 0 xdiff/xtypes.h
3451 100644 2ade97b2574a9f77e7ae4002a4e07a6a38e46d07 0 xdiff/xutils.c
3452 100644 d5de8292e05e7c36c4b68857c1cf9855e3d2f70a 0 xdiff/xutils.h
3453 -------------------------------------------------
3454
3455 Note that in older documentation you may see the index called the
3456 "current directory cache" or just the "cache". It has three important
3457 properties:
3458
3459 1. The index contains all the information necessary to generate a single
3460 (uniquely determined) tree object.
3461 +
3462 For example, running linkgit:git-commit[1] generates this tree object
3463 from the index, stores it in the object database, and uses it as the
3464 tree object associated with the new commit.
3465
3466 2. The index enables fast comparisons between the tree object it defines
3467 and the working tree.
3468 +
3469 It does this by storing some additional data for each entry (such as
3470 the last modified time). This data is not displayed above, and is not
3471 stored in the created tree object, but it can be used to determine
3472 quickly which files in the working directory differ from what was
3473 stored in the index, and thus save git from having to read all of the
3474 data from such files to look for changes.
3475
3476 3. It can efficiently represent information about merge conflicts
3477 between different tree objects, allowing each pathname to be
3478 associated with sufficient information about the trees involved that
3479 you can create a three-way merge between them.
3480 +
3481 We saw in <<conflict-resolution>> that during a merge the index can
3482 store multiple versions of a single file (called "stages"). The third
3483 column in the linkgit:git-ls-files[1] output above is the stage
3484 number, and will take on values other than 0 for files with merge
3485 conflicts.
3486
3487 The index is thus a sort of temporary staging area, which is filled with
3488 a tree which you are in the process of working on.
3489
3490 If you blow the index away entirely, you generally haven't lost any
3491 information as long as you have the name of the tree that it described.
3492
3493 [[submodules]]
3494 Submodules
3495 ==========
3496
3497 Large projects are often composed of smaller, self-contained modules. For
3498 example, an embedded Linux distribution's source tree would include every
3499 piece of software in the distribution with some local modifications; a movie
3500 player might need to build against a specific, known-working version of a
3501 decompression library; several independent programs might all share the same
3502 build scripts.
3503
3504 With centralized revision control systems this is often accomplished by
3505 including every module in one single repository. Developers can check out
3506 all modules or only the modules they need to work with. They can even modify
3507 files across several modules in a single commit while moving things around
3508 or updating APIs and translations.
3509
3510 Git does not allow partial checkouts, so duplicating this approach in Git
3511 would force developers to keep a local copy of modules they are not
3512 interested in touching. Commits in an enormous checkout would be slower
3513 than you'd expect as Git would have to scan every directory for changes.
3514 If modules have a lot of local history, clones would take forever.
3515
3516 On the plus side, distributed revision control systems can much better
3517 integrate with external sources. In a centralized model, a single arbitrary
3518 snapshot of the external project is exported from its own revision control
3519 and then imported into the local revision control on a vendor branch. All
3520 the history is hidden. With distributed revision control you can clone the
3521 entire external history and much more easily follow development and re-merge
3522 local changes.
3523
3524 Git's submodule support allows a repository to contain, as a subdirectory, a
3525 checkout of an external project. Submodules maintain their own identity;
3526 the submodule support just stores the submodule repository location and
3527 commit ID, so other developers who clone the containing project
3528 ("superproject") can easily clone all the submodules at the same revision.
3529 Partial checkouts of the superproject are possible: you can tell Git to
3530 clone none, some or all of the submodules.
3531
3532 The linkgit:git-submodule[1] command is available since Git 1.5.3. Users
3533 with Git 1.5.2 can look up the submodule commits in the repository and
3534 manually check them out; earlier versions won't recognize the submodules at
3535 all.
3536
3537 To see how submodule support works, create (for example) four example
3538 repositories that can be used later as a submodule:
3539
3540 -------------------------------------------------
3541 $ mkdir ~/git
3542 $ cd ~/git
3543 $ for i in a b c d
3544 do
3545 mkdir $i
3546 cd $i
3547 git init
3548 echo "module $i" > $i.txt
3549 git add $i.txt
3550 git commit -m "Initial commit, submodule $i"
3551 cd ..
3552 done
3553 -------------------------------------------------
3554
3555 Now create the superproject and add all the submodules:
3556
3557 -------------------------------------------------
3558 $ mkdir super
3559 $ cd super
3560 $ git init
3561 $ for i in a b c d
3562 do
3563 git submodule add ~/git/$i $i
3564 done
3565 -------------------------------------------------
3566
3567 NOTE: Do not use local URLs here if you plan to publish your superproject!
3568
3569 See what files `git submodule` created:
3570
3571 -------------------------------------------------
3572 $ ls -a
3573 . .. .git .gitmodules a b c d
3574 -------------------------------------------------
3575
3576 The `git submodule add <repo> <path>` command does a couple of things:
3577
3578 - It clones the submodule from <repo> to the given <path> under the
3579 current directory and by default checks out the master branch.
3580 - It adds the submodule's clone path to the linkgit:gitmodules[5] file and
3581 adds this file to the index, ready to be committed.
3582 - It adds the submodule's current commit ID to the index, ready to be
3583 committed.
3584
3585 Commit the superproject:
3586
3587 -------------------------------------------------
3588 $ git commit -m "Add submodules a, b, c and d."
3589 -------------------------------------------------
3590
3591 Now clone the superproject:
3592
3593 -------------------------------------------------
3594 $ cd ..
3595 $ git clone super cloned
3596 $ cd cloned
3597 -------------------------------------------------
3598
3599 The submodule directories are there, but they're empty:
3600
3601 -------------------------------------------------
3602 $ ls -a a
3603 . ..
3604 $ git submodule status
3605 -d266b9873ad50488163457f025db7cdd9683d88b a
3606 -e81d457da15309b4fef4249aba9b50187999670d b
3607 -c1536a972b9affea0f16e0680ba87332dc059146 c
3608 -d96249ff5d57de5de093e6baff9e0aafa5276a74 d
3609 -------------------------------------------------
3610
3611 NOTE: The commit object names shown above would be different for you, but they
3612 should match the HEAD commit object names of your repositories. You can check
3613 it by running `git ls-remote ../a`.
3614
3615 Pulling down the submodules is a two-step process. First run `git submodule
3616 init` to add the submodule repository URLs to `.git/config`:
3617
3618 -------------------------------------------------
3619 $ git submodule init
3620 -------------------------------------------------
3621
3622 Now use `git submodule update` to clone the repositories and check out the
3623 commits specified in the superproject:
3624
3625 -------------------------------------------------
3626 $ git submodule update
3627 $ cd a
3628 $ ls -a
3629 . .. .git a.txt
3630 -------------------------------------------------
3631
3632 One major difference between `git submodule update` and `git submodule add` is
3633 that `git submodule update` checks out a specific commit, rather than the tip
3634 of a branch. It's like checking out a tag: the head is detached, so you're not
3635 working on a branch.
3636
3637 -------------------------------------------------
3638 $ git branch
3639 * (no branch)
3640 master
3641 -------------------------------------------------
3642
3643 If you want to make a change within a submodule and you have a detached head,
3644 then you should create or checkout a branch, make your changes, publish the
3645 change within the submodule, and then update the superproject to reference the
3646 new commit:
3647
3648 -------------------------------------------------
3649 $ git checkout master
3650 -------------------------------------------------
3651
3652 or
3653
3654 -------------------------------------------------
3655 $ git checkout -b fix-up
3656 -------------------------------------------------
3657
3658 then
3659
3660 -------------------------------------------------
3661 $ echo "adding a line again" >> a.txt
3662 $ git commit -a -m "Updated the submodule from within the superproject."
3663 $ git push
3664 $ cd ..
3665 $ git diff
3666 diff --git a/a b/a
3667 index d266b98..261dfac 160000
3668 --- a/a
3669 +++ b/a
3670 @@ -1 +1 @@
3671 -Subproject commit d266b9873ad50488163457f025db7cdd9683d88b
3672 +Subproject commit 261dfac35cb99d380eb966e102c1197139f7fa24
3673 $ git add a
3674 $ git commit -m "Updated submodule a."
3675 $ git push
3676 -------------------------------------------------
3677
3678 You have to run `git submodule update` after `git pull` if you want to update
3679 submodules, too.
3680
3681 Pitfalls with submodules
3682 ------------------------
3683
3684 Always publish the submodule change before publishing the change to the
3685 superproject that references it. If you forget to publish the submodule change,
3686 others won't be able to clone the repository:
3687
3688 -------------------------------------------------
3689 $ cd ~/git/super/a
3690 $ echo i added another line to this file >> a.txt
3691 $ git commit -a -m "doing it wrong this time"
3692 $ cd ..
3693 $ git add a
3694 $ git commit -m "Updated submodule a again."
3695 $ git push
3696 $ cd ~/git/cloned
3697 $ git pull
3698 $ git submodule update
3699 error: pathspec '261dfac35cb99d380eb966e102c1197139f7fa24' did not match any file(s) known to git.
3700 Did you forget to 'git add'?
3701 Unable to checkout '261dfac35cb99d380eb966e102c1197139f7fa24' in submodule path 'a'
3702 -------------------------------------------------
3703
3704 In older git versions it could be easily forgotten to commit new or modified
3705 files in a submodule, which silently leads to similar problems as not pushing
3706 the submodule changes. Starting with git 1.7.0 both "git status" and "git diff"
3707 in the superproject show submodules as modified when they contain new or
3708 modified files to protect against accidentally committing such a state. "git
3709 diff" will also add a "-dirty" to the work tree side when generating patch
3710 output or used with the --submodule option:
3711
3712 -------------------------------------------------
3713 $ git diff
3714 diff --git a/sub b/sub
3715 --- a/sub
3716 +++ b/sub
3717 @@ -1 +1 @@
3718 -Subproject commit 3f356705649b5d566d97ff843cf193359229a453
3719 +Subproject commit 3f356705649b5d566d97ff843cf193359229a453-dirty
3720 $ git diff --submodule
3721 Submodule sub 3f35670..3f35670-dirty:
3722 -------------------------------------------------
3723
3724 You also should not rewind branches in a submodule beyond commits that were
3725 ever recorded in any superproject.
3726
3727 It's not safe to run `git submodule update` if you've made and committed
3728 changes within a submodule without checking out a branch first. They will be
3729 silently overwritten:
3730
3731 -------------------------------------------------
3732 $ cat a.txt
3733 module a
3734 $ echo line added from private2 >> a.txt
3735 $ git commit -a -m "line added inside private2"
3736 $ cd ..
3737 $ git submodule update
3738 Submodule path 'a': checked out 'd266b9873ad50488163457f025db7cdd9683d88b'
3739 $ cd a
3740 $ cat a.txt
3741 module a
3742 -------------------------------------------------
3743
3744 NOTE: The changes are still visible in the submodule's reflog.
3745
3746 This is not the case if you did not commit your changes.
3747
3748 [[low-level-operations]]
3749 Low-level git operations
3750 ========================
3751
3752 Many of the higher-level commands were originally implemented as shell
3753 scripts using a smaller core of low-level git commands. These can still
3754 be useful when doing unusual things with git, or just as a way to
3755 understand its inner workings.
3756
3757 [[object-manipulation]]
3758 Object access and manipulation
3759 ------------------------------
3760
3761 The linkgit:git-cat-file[1] command can show the contents of any object,
3762 though the higher-level linkgit:git-show[1] is usually more useful.
3763
3764 The linkgit:git-commit-tree[1] command allows constructing commits with
3765 arbitrary parents and trees.
3766
3767 A tree can be created with linkgit:git-write-tree[1] and its data can be
3768 accessed by linkgit:git-ls-tree[1]. Two trees can be compared with
3769 linkgit:git-diff-tree[1].
3770
3771 A tag is created with linkgit:git-mktag[1], and the signature can be
3772 verified by linkgit:git-verify-tag[1], though it is normally simpler to
3773 use linkgit:git-tag[1] for both.
3774
3775 [[the-workflow]]
3776 The Workflow
3777 ------------
3778
3779 High-level operations such as linkgit:git-commit[1],
3780 linkgit:git-checkout[1] and linkgit:git-reset[1] work by moving data
3781 between the working tree, the index, and the object database. Git
3782 provides low-level operations which perform each of these steps
3783 individually.
3784
3785 Generally, all "git" operations work on the index file. Some operations
3786 work *purely* on the index file (showing the current state of the
3787 index), but most operations move data between the index file and either
3788 the database or the working directory. Thus there are four main
3789 combinations:
3790
3791 [[working-directory-to-index]]
3792 working directory -> index
3793 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3794
3795 The linkgit:git-update-index[1] command updates the index with
3796 information from the working directory. You generally update the
3797 index information by just specifying the filename you want to update,
3798 like so:
3799
3800 -------------------------------------------------
3801 $ git update-index filename
3802 -------------------------------------------------
3803
3804 but to avoid common mistakes with filename globbing etc, the command
3805 will not normally add totally new entries or remove old entries,
3806 i.e. it will normally just update existing cache entries.
3807
3808 To tell git that yes, you really do realize that certain files no
3809 longer exist, or that new files should be added, you
3810 should use the `--remove` and `--add` flags respectively.
3811
3812 NOTE! A `--remove` flag does 'not' mean that subsequent filenames will
3813 necessarily be removed: if the files still exist in your directory
3814 structure, the index will be updated with their new status, not
3815 removed. The only thing `--remove` means is that update-index will be
3816 considering a removed file to be a valid thing, and if the file really
3817 does not exist any more, it will update the index accordingly.
3818
3819 As a special case, you can also do `git update-index --refresh`, which
3820 will refresh the "stat" information of each index to match the current
3821 stat information. It will 'not' update the object status itself, and
3822 it will only update the fields that are used to quickly test whether
3823 an object still matches its old backing store object.
3824
3825 The previously introduced linkgit:git-add[1] is just a wrapper for
3826 linkgit:git-update-index[1].
3827
3828 [[index-to-object-database]]
3829 index -> object database
3830 ~~~~~~~~~~~~~~~~~~~~~~~~
3831
3832 You write your current index file to a "tree" object with the program
3833
3834 -------------------------------------------------
3835 $ git write-tree
3836 -------------------------------------------------
3837
3838 that doesn't come with any options--it will just write out the
3839 current index into the set of tree objects that describe that state,
3840 and it will return the name of the resulting top-level tree. You can
3841 use that tree to re-generate the index at any time by going in the
3842 other direction:
3843
3844 [[object-database-to-index]]
3845 object database -> index
3846 ~~~~~~~~~~~~~~~~~~~~~~~~
3847
3848 You read a "tree" file from the object database, and use that to
3849 populate (and overwrite--don't do this if your index contains any
3850 unsaved state that you might want to restore later!) your current
3851 index. Normal operation is just
3852
3853 -------------------------------------------------
3854 $ git read-tree <SHA-1 of tree>
3855 -------------------------------------------------
3856
3857 and your index file will now be equivalent to the tree that you saved
3858 earlier. However, that is only your 'index' file: your working
3859 directory contents have not been modified.
3860
3861 [[index-to-working-directory]]
3862 index -> working directory
3863 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3864
3865 You update your working directory from the index by "checking out"
3866 files. This is not a very common operation, since normally you'd just
3867 keep your files updated, and rather than write to your working
3868 directory, you'd tell the index files about the changes in your
3869 working directory (i.e. `git update-index`).
3870
3871 However, if you decide to jump to a new version, or check out somebody
3872 else's version, or just restore a previous tree, you'd populate your
3873 index file with read-tree, and then you need to check out the result
3874 with
3875
3876 -------------------------------------------------
3877 $ git checkout-index filename
3878 -------------------------------------------------
3879
3880 or, if you want to check out all of the index, use `-a`.
3881
3882 NOTE! `git checkout-index` normally refuses to overwrite old files, so
3883 if you have an old version of the tree already checked out, you will
3884 need to use the "-f" flag ('before' the "-a" flag or the filename) to
3885 'force' the checkout.
3886
3887
3888 Finally, there are a few odds and ends which are not purely moving
3889 from one representation to the other:
3890
3891 [[tying-it-all-together]]
3892 Tying it all together
3893 ~~~~~~~~~~~~~~~~~~~~~
3894
3895 To commit a tree you have instantiated with "git write-tree", you'd
3896 create a "commit" object that refers to that tree and the history
3897 behind it--most notably the "parent" commits that preceded it in
3898 history.
3899
3900 Normally a "commit" has one parent: the previous state of the tree
3901 before a certain change was made. However, sometimes it can have two
3902 or more parent commits, in which case we call it a "merge", due to the
3903 fact that such a commit brings together ("merges") two or more
3904 previous states represented by other commits.
3905
3906 In other words, while a "tree" represents a particular directory state
3907 of a working directory, a "commit" represents that state in "time",
3908 and explains how we got there.
3909
3910 You create a commit object by giving it the tree that describes the
3911 state at the time of the commit, and a list of parents:
3912
3913 -------------------------------------------------
3914 $ git commit-tree <tree> -p <parent> [(-p <parent2>)...]
3915 -------------------------------------------------
3916
3917 and then giving the reason for the commit on stdin (either through
3918 redirection from a pipe or file, or by just typing it at the tty).
3919
3920 `git commit-tree` will return the name of the object that represents
3921 that commit, and you should save it away for later use. Normally,
3922 you'd commit a new `HEAD` state, and while git doesn't care where you
3923 save the note about that state, in practice we tend to just write the
3924 result to the file pointed at by `.git/HEAD`, so that we can always see
3925 what the last committed state was.
3926
3927 Here is an ASCII art by Jon Loeliger that illustrates how
3928 various pieces fit together.
3929
3930 ------------
3931
3932 commit-tree
3933 commit obj
3934 +----+
3935 | |
3936 | |
3937 V V
3938 +-----------+
3939 | Object DB |
3940 | Backing |
3941 | Store |
3942 +-----------+
3943 ^
3944 write-tree | |
3945 tree obj | |
3946 | | read-tree
3947 | | tree obj
3948 V
3949 +-----------+
3950 | Index |
3951 | "cache" |
3952 +-----------+
3953 update-index ^
3954 blob obj | |
3955 | |
3956 checkout-index -u | | checkout-index
3957 stat | | blob obj
3958 V
3959 +-----------+
3960 | Working |
3961 | Directory |
3962 +-----------+
3963
3964 ------------
3965
3966
3967 [[examining-the-data]]
3968 Examining the data
3969 ------------------
3970
3971 You can examine the data represented in the object database and the
3972 index with various helper tools. For every object, you can use
3973 linkgit:git-cat-file[1] to examine details about the
3974 object:
3975
3976 -------------------------------------------------
3977 $ git cat-file -t <objectname>
3978 -------------------------------------------------
3979
3980 shows the type of the object, and once you have the type (which is
3981 usually implicit in where you find the object), you can use
3982
3983 -------------------------------------------------
3984 $ git cat-file blob|tree|commit|tag <objectname>
3985 -------------------------------------------------
3986
3987 to show its contents. NOTE! Trees have binary content, and as a result
3988 there is a special helper for showing that content, called
3989 `git ls-tree`, which turns the binary content into a more easily
3990 readable form.
3991
3992 It's especially instructive to look at "commit" objects, since those
3993 tend to be small and fairly self-explanatory. In particular, if you
3994 follow the convention of having the top commit name in `.git/HEAD`,
3995 you can do
3996
3997 -------------------------------------------------
3998 $ git cat-file commit HEAD
3999 -------------------------------------------------
4000
4001 to see what the top commit was.
4002
4003 [[merging-multiple-trees]]
4004 Merging multiple trees
4005 ----------------------
4006
4007 Git helps you do a three-way merge, which you can expand to n-way by
4008 repeating the merge procedure arbitrary times until you finally
4009 "commit" the state. The normal situation is that you'd only do one
4010 three-way merge (two parents), and commit it, but if you like to, you
4011 can do multiple parents in one go.
4012
4013 To do a three-way merge, you need the two sets of "commit" objects
4014 that you want to merge, use those to find the closest common parent (a
4015 third "commit" object), and then use those commit objects to find the
4016 state of the directory ("tree" object) at these points.
4017
4018 To get the "base" for the merge, you first look up the common parent
4019 of two commits with
4020
4021 -------------------------------------------------
4022 $ git merge-base <commit1> <commit2>
4023 -------------------------------------------------
4024
4025 which will return you the commit they are both based on. You should
4026 now look up the "tree" objects of those commits, which you can easily
4027 do with (for example)
4028
4029 -------------------------------------------------
4030 $ git cat-file commit <commitname> | head -1
4031 -------------------------------------------------
4032
4033 since the tree object information is always the first line in a commit
4034 object.
4035
4036 Once you know the three trees you are going to merge (the one "original"
4037 tree, aka the common tree, and the two "result" trees, aka the branches
4038 you want to merge), you do a "merge" read into the index. This will
4039 complain if it has to throw away your old index contents, so you should
4040 make sure that you've committed those--in fact you would normally
4041 always do a merge against your last commit (which should thus match what
4042 you have in your current index anyway).
4043
4044 To do the merge, do
4045
4046 -------------------------------------------------
4047 $ git read-tree -m -u <origtree> <yourtree> <targettree>
4048 -------------------------------------------------
4049
4050 which will do all trivial merge operations for you directly in the
4051 index file, and you can just write the result out with
4052 `git write-tree`.
4053
4054
4055 [[merging-multiple-trees-2]]
4056 Merging multiple trees, continued
4057 ---------------------------------
4058
4059 Sadly, many merges aren't trivial. If there are files that have
4060 been added, moved or removed, or if both branches have modified the
4061 same file, you will be left with an index tree that contains "merge
4062 entries" in it. Such an index tree can 'NOT' be written out to a tree
4063 object, and you will have to resolve any such merge clashes using
4064 other tools before you can write out the result.
4065
4066 You can examine such index state with `git ls-files --unmerged`
4067 command. An example:
4068
4069 ------------------------------------------------
4070 $ git read-tree -m $orig HEAD $target
4071 $ git ls-files --unmerged
4072 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello.c
4073 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello.c
4074 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello.c
4075 ------------------------------------------------
4076
4077 Each line of the `git ls-files --unmerged` output begins with
4078 the blob mode bits, blob SHA-1, 'stage number', and the
4079 filename. The 'stage number' is git's way to say which tree it
4080 came from: stage 1 corresponds to the `$orig` tree, stage 2 to
4081 the `HEAD` tree, and stage 3 to the `$target` tree.
4082
4083 Earlier we said that trivial merges are done inside
4084 `git read-tree -m`. For example, if the file did not change
4085 from `$orig` to `HEAD` nor `$target`, or if the file changed
4086 from `$orig` to `HEAD` and `$orig` to `$target` the same way,
4087 obviously the final outcome is what is in `HEAD`. What the
4088 above example shows is that file `hello.c` was changed from
4089 `$orig` to `HEAD` and `$orig` to `$target` in a different way.
4090 You could resolve this by running your favorite 3-way merge
4091 program, e.g. `diff3`, `merge`, or git's own merge-file, on
4092 the blob objects from these three stages yourself, like this:
4093
4094 ------------------------------------------------
4095 $ git cat-file blob 263414f... >hello.c~1
4096 $ git cat-file blob 06fa6a2... >hello.c~2
4097 $ git cat-file blob cc44c73... >hello.c~3
4098 $ git merge-file hello.c~2 hello.c~1 hello.c~3
4099 ------------------------------------------------
4100
4101 This would leave the merge result in `hello.c~2` file, along
4102 with conflict markers if there are conflicts. After verifying
4103 the merge result makes sense, you can tell git what the final
4104 merge result for this file is by:
4105
4106 -------------------------------------------------
4107 $ mv -f hello.c~2 hello.c
4108 $ git update-index hello.c
4109 -------------------------------------------------
4110
4111 When a path is in the "unmerged" state, running `git update-index` for
4112 that path tells git to mark the path resolved.
4113
4114 The above is the description of a git merge at the lowest level,
4115 to help you understand what conceptually happens under the hood.
4116 In practice, nobody, not even git itself, runs `git cat-file` three times
4117 for this. There is a `git merge-index` program that extracts the
4118 stages to temporary files and calls a "merge" script on it:
4119
4120 -------------------------------------------------
4121 $ git merge-index git-merge-one-file hello.c
4122 -------------------------------------------------
4123
4124 and that is what higher level `git merge -s resolve` is implemented with.
4125
4126 [[hacking-git]]
4127 Hacking git
4128 ===========
4129
4130 This chapter covers internal details of the git implementation which
4131 probably only git developers need to understand.
4132
4133 [[object-details]]
4134 Object storage format
4135 ---------------------
4136
4137 All objects have a statically determined "type" which identifies the
4138 format of the object (i.e. how it is used, and how it can refer to other
4139 objects). There are currently four different object types: "blob",
4140 "tree", "commit", and "tag".
4141
4142 Regardless of object type, all objects share the following
4143 characteristics: they are all deflated with zlib, and have a header
4144 that not only specifies their type, but also provides size information
4145 about the data in the object. It's worth noting that the SHA-1 hash
4146 that is used to name the object is the hash of the original data
4147 plus this header, so `sha1sum` 'file' does not match the object name
4148 for 'file'.
4149 (Historical note: in the dawn of the age of git the hash
4150 was the SHA-1 of the 'compressed' object.)
4151
4152 As a result, the general consistency of an object can always be tested
4153 independently of the contents or the type of the object: all objects can
4154 be validated by verifying that (a) their hashes match the content of the
4155 file and (b) the object successfully inflates to a stream of bytes that
4156 forms a sequence of <ascii type without space> {plus} <space> {plus} <ascii decimal
4157 size> {plus} <byte\0> {plus} <binary object data>.
4158
4159 The structured objects can further have their structure and
4160 connectivity to other objects verified. This is generally done with
4161 the `git fsck` program, which generates a full dependency graph
4162 of all objects, and verifies their internal consistency (in addition
4163 to just verifying their superficial consistency through the hash).
4164
4165 [[birdview-on-the-source-code]]
4166 A birds-eye view of Git's source code
4167 -------------------------------------
4168
4169 It is not always easy for new developers to find their way through Git's
4170 source code. This section gives you a little guidance to show where to
4171 start.
4172
4173 A good place to start is with the contents of the initial commit, with:
4174
4175 ----------------------------------------------------
4176 $ git checkout e83c5163
4177 ----------------------------------------------------
4178
4179 The initial revision lays the foundation for almost everything git has
4180 today, but is small enough to read in one sitting.
4181
4182 Note that terminology has changed since that revision. For example, the
4183 README in that revision uses the word "changeset" to describe what we
4184 now call a <<def_commit_object,commit>>.
4185
4186 Also, we do not call it "cache" any more, but rather "index"; however, the
4187 file is still called `cache.h`. Remark: Not much reason to change it now,
4188 especially since there is no good single name for it anyway, because it is
4189 basically _the_ header file which is included by _all_ of Git's C sources.
4190
4191 If you grasp the ideas in that initial commit, you should check out a
4192 more recent version and skim `cache.h`, `object.h` and `commit.h`.
4193
4194 In the early days, Git (in the tradition of UNIX) was a bunch of programs
4195 which were extremely simple, and which you used in scripts, piping the
4196 output of one into another. This turned out to be good for initial
4197 development, since it was easier to test new things. However, recently
4198 many of these parts have become builtins, and some of the core has been
4199 "libified", i.e. put into libgit.a for performance, portability reasons,
4200 and to avoid code duplication.
4201
4202 By now, you know what the index is (and find the corresponding data
4203 structures in `cache.h`), and that there are just a couple of object types
4204 (blobs, trees, commits and tags) which inherit their common structure from
4205 `struct object`, which is their first member (and thus, you can cast e.g.
4206 `(struct object *)commit` to achieve the _same_ as `&commit->object`, i.e.
4207 get at the object name and flags).
4208
4209 Now is a good point to take a break to let this information sink in.
4210
4211 Next step: get familiar with the object naming. Read <<naming-commits>>.
4212 There are quite a few ways to name an object (and not only revisions!).
4213 All of these are handled in `sha1_name.c`. Just have a quick look at
4214 the function `get_sha1()`. A lot of the special handling is done by
4215 functions like `get_sha1_basic()` or the likes.
4216
4217 This is just to get you into the groove for the most libified part of Git:
4218 the revision walker.
4219
4220 Basically, the initial version of `git log` was a shell script:
4221
4222 ----------------------------------------------------------------
4223 $ git-rev-list --pretty $(git-rev-parse --default HEAD "$@") | \
4224 LESS=-S ${PAGER:-less}
4225 ----------------------------------------------------------------
4226
4227 What does this mean?
4228
4229 `git rev-list` is the original version of the revision walker, which
4230 _always_ printed a list of revisions to stdout. It is still functional,
4231 and needs to, since most new Git commands start out as scripts using
4232 `git rev-list`.
4233
4234 `git rev-parse` is not as important any more; it was only used to filter out
4235 options that were relevant for the different plumbing commands that were
4236 called by the script.
4237
4238 Most of what `git rev-list` did is contained in `revision.c` and
4239 `revision.h`. It wraps the options in a struct named `rev_info`, which
4240 controls how and what revisions are walked, and more.
4241
4242 The original job of `git rev-parse` is now taken by the function
4243 `setup_revisions()`, which parses the revisions and the common command line
4244 options for the revision walker. This information is stored in the struct
4245 `rev_info` for later consumption. You can do your own command line option
4246 parsing after calling `setup_revisions()`. After that, you have to call
4247 `prepare_revision_walk()` for initialization, and then you can get the
4248 commits one by one with the function `get_revision()`.
4249
4250 If you are interested in more details of the revision walking process,
4251 just have a look at the first implementation of `cmd_log()`; call
4252 `git show v1.3.0~155^2~4` and scroll down to that function (note that you
4253 no longer need to call `setup_pager()` directly).
4254
4255 Nowadays, `git log` is a builtin, which means that it is _contained_ in the
4256 command `git`. The source side of a builtin is
4257
4258 - a function called `cmd_<bla>`, typically defined in `builtin-<bla>.c`,
4259 and declared in `builtin.h`,
4260
4261 - an entry in the `commands[]` array in `git.c`, and
4262
4263 - an entry in `BUILTIN_OBJECTS` in the `Makefile`.
4264
4265 Sometimes, more than one builtin is contained in one source file. For
4266 example, `cmd_whatchanged()` and `cmd_log()` both reside in `builtin-log.c`,
4267 since they share quite a bit of code. In that case, the commands which are
4268 _not_ named like the `.c` file in which they live have to be listed in
4269 `BUILT_INS` in the `Makefile`.
4270
4271 `git log` looks more complicated in C than it does in the original script,
4272 but that allows for a much greater flexibility and performance.
4273
4274 Here again it is a good point to take a pause.
4275
4276 Lesson three is: study the code. Really, it is the best way to learn about
4277 the organization of Git (after you know the basic concepts).
4278
4279 So, think about something which you are interested in, say, "how can I
4280 access a blob just knowing the object name of it?". The first step is to
4281 find a Git command with which you can do it. In this example, it is either
4282 `git show` or `git cat-file`.
4283
4284 For the sake of clarity, let's stay with `git cat-file`, because it
4285
4286 - is plumbing, and
4287
4288 - was around even in the initial commit (it literally went only through
4289 some 20 revisions as `cat-file.c`, was renamed to `builtin-cat-file.c`
4290 when made a builtin, and then saw less than 10 versions).
4291
4292 So, look into `builtin-cat-file.c`, search for `cmd_cat_file()` and look what
4293 it does.
4294
4295 ------------------------------------------------------------------
4296 git_config(git_default_config);
4297 if (argc != 3)
4298 usage("git cat-file [-t|-s|-e|-p|<type>] <sha1>");
4299 if (get_sha1(argv[2], sha1))
4300 die("Not a valid object name %s", argv[2]);
4301 ------------------------------------------------------------------
4302
4303 Let's skip over the obvious details; the only really interesting part
4304 here is the call to `get_sha1()`. It tries to interpret `argv[2]` as an
4305 object name, and if it refers to an object which is present in the current
4306 repository, it writes the resulting SHA-1 into the variable `sha1`.
4307
4308 Two things are interesting here:
4309
4310 - `get_sha1()` returns 0 on _success_. This might surprise some new
4311 Git hackers, but there is a long tradition in UNIX to return different
4312 negative numbers in case of different errors--and 0 on success.
4313
4314 - the variable `sha1` in the function signature of `get_sha1()` is `unsigned
4315 char *`, but is actually expected to be a pointer to `unsigned
4316 char[20]`. This variable will contain the 160-bit SHA-1 of the given
4317 commit. Note that whenever a SHA-1 is passed as `unsigned char *`, it
4318 is the binary representation, as opposed to the ASCII representation in
4319 hex characters, which is passed as `char *`.
4320
4321 You will see both of these things throughout the code.
4322
4323 Now, for the meat:
4324
4325 -----------------------------------------------------------------------------
4326 case 0:
4327 buf = read_object_with_reference(sha1, argv[1], &size, NULL);
4328 -----------------------------------------------------------------------------
4329
4330 This is how you read a blob (actually, not only a blob, but any type of
4331 object). To know how the function `read_object_with_reference()` actually
4332 works, find the source code for it (something like `git grep
4333 read_object_with | grep ":[a-z]"` in the git repository), and read
4334 the source.
4335
4336 To find out how the result can be used, just read on in `cmd_cat_file()`:
4337
4338 -----------------------------------
4339 write_or_die(1, buf, size);
4340 -----------------------------------
4341
4342 Sometimes, you do not know where to look for a feature. In many such cases,
4343 it helps to search through the output of `git log`, and then `git show` the
4344 corresponding commit.
4345
4346 Example: If you know that there was some test case for `git bundle`, but
4347 do not remember where it was (yes, you _could_ `git grep bundle t/`, but that
4348 does not illustrate the point!):
4349
4350 ------------------------
4351 $ git log --no-merges t/
4352 ------------------------
4353
4354 In the pager (`less`), just search for "bundle", go a few lines back,
4355 and see that it is in commit 18449ab0... Now just copy this object name,
4356 and paste it into the command line
4357
4358 -------------------
4359 $ git show 18449ab0
4360 -------------------
4361
4362 Voila.
4363
4364 Another example: Find out what to do in order to make some script a
4365 builtin:
4366
4367 -------------------------------------------------
4368 $ git log --no-merges --diff-filter=A builtin-*.c
4369 -------------------------------------------------
4370
4371 You see, Git is actually the best tool to find out about the source of Git
4372 itself!
4373
4374 [[glossary]]
4375 Git Glossary
4376 ============
4377
4378 include::glossary-content.txt[]
4379
4380 [[git-quick-start]]
4381 Appendix A: Git Quick Reference
4382 ===============================
4383
4384 This is a quick summary of the major commands; the previous chapters
4385 explain how these work in more detail.
4386
4387 [[quick-creating-a-new-repository]]
4388 Creating a new repository
4389 -------------------------
4390
4391 From a tarball:
4392
4393 -----------------------------------------------
4394 $ tar xzf project.tar.gz
4395 $ cd project
4396 $ git init
4397 Initialized empty Git repository in .git/
4398 $ git add .
4399 $ git commit
4400 -----------------------------------------------
4401
4402 From a remote repository:
4403
4404 -----------------------------------------------
4405 $ git clone git://example.com/pub/project.git
4406 $ cd project
4407 -----------------------------------------------
4408
4409 [[managing-branches]]
4410 Managing branches
4411 -----------------
4412
4413 -----------------------------------------------
4414 $ git branch # list all local branches in this repo
4415 $ git checkout test # switch working directory to branch "test"
4416 $ git branch new # create branch "new" starting at current HEAD
4417 $ git branch -d new # delete branch "new"
4418 -----------------------------------------------
4419
4420 Instead of basing a new branch on current HEAD (the default), use:
4421
4422 -----------------------------------------------
4423 $ git branch new test # branch named "test"
4424 $ git branch new v2.6.15 # tag named v2.6.15
4425 $ git branch new HEAD^ # commit before the most recent
4426 $ git branch new HEAD^^ # commit before that
4427 $ git branch new test~10 # ten commits before tip of branch "test"
4428 -----------------------------------------------
4429
4430 Create and switch to a new branch at the same time:
4431
4432 -----------------------------------------------
4433 $ git checkout -b new v2.6.15
4434 -----------------------------------------------
4435
4436 Update and examine branches from the repository you cloned from:
4437
4438 -----------------------------------------------
4439 $ git fetch # update
4440 $ git branch -r # list
4441 origin/master
4442 origin/next
4443 ...
4444 $ git checkout -b masterwork origin/master
4445 -----------------------------------------------
4446
4447 Fetch a branch from a different repository, and give it a new
4448 name in your repository:
4449
4450 -----------------------------------------------
4451 $ git fetch git://example.com/project.git theirbranch:mybranch
4452 $ git fetch git://example.com/project.git v2.6.15:mybranch
4453 -----------------------------------------------
4454
4455 Keep a list of repositories you work with regularly:
4456
4457 -----------------------------------------------
4458 $ git remote add example git://example.com/project.git
4459 $ git remote # list remote repositories
4460 example
4461 origin
4462 $ git remote show example # get details
4463 * remote example
4464 URL: git://example.com/project.git
4465 Tracked remote branches
4466 master
4467 next
4468 ...
4469 $ git fetch example # update branches from example
4470 $ git branch -r # list all remote branches
4471 -----------------------------------------------
4472
4473
4474 [[exploring-history]]
4475 Exploring history
4476 -----------------
4477
4478 -----------------------------------------------
4479 $ gitk # visualize and browse history
4480 $ git log # list all commits
4481 $ git log src/ # ...modifying src/
4482 $ git log v2.6.15..v2.6.16 # ...in v2.6.16, not in v2.6.15
4483 $ git log master..test # ...in branch test, not in branch master
4484 $ git log test..master # ...in branch master, but not in test
4485 $ git log test...master # ...in one branch, not in both
4486 $ git log -S'foo()' # ...where difference contain "foo()"
4487 $ git log --since="2 weeks ago"
4488 $ git log -p # show patches as well
4489 $ git show # most recent commit
4490 $ git diff v2.6.15..v2.6.16 # diff between two tagged versions
4491 $ git diff v2.6.15..HEAD # diff with current head
4492 $ git grep "foo()" # search working directory for "foo()"
4493 $ git grep v2.6.15 "foo()" # search old tree for "foo()"
4494 $ git show v2.6.15:a.txt # look at old version of a.txt
4495 -----------------------------------------------
4496
4497 Search for regressions:
4498
4499 -----------------------------------------------
4500 $ git bisect start
4501 $ git bisect bad # current version is bad
4502 $ git bisect good v2.6.13-rc2 # last known good revision
4503 Bisecting: 675 revisions left to test after this
4504 # test here, then:
4505 $ git bisect good # if this revision is good, or
4506 $ git bisect bad # if this revision is bad.
4507 # repeat until done.
4508 -----------------------------------------------
4509
4510 [[making-changes]]
4511 Making changes
4512 --------------
4513
4514 Make sure git knows who to blame:
4515
4516 ------------------------------------------------
4517 $ cat >>~/.gitconfig <<\EOF
4518 [user]
4519 name = Your Name Comes Here
4520 email = you@yourdomain.example.com
4521 EOF
4522 ------------------------------------------------
4523
4524 Select file contents to include in the next commit, then make the
4525 commit:
4526
4527 -----------------------------------------------
4528 $ git add a.txt # updated file
4529 $ git add b.txt # new file
4530 $ git rm c.txt # old file
4531 $ git commit
4532 -----------------------------------------------
4533
4534 Or, prepare and create the commit in one step:
4535
4536 -----------------------------------------------
4537 $ git commit d.txt # use latest content only of d.txt
4538 $ git commit -a # use latest content of all tracked files
4539 -----------------------------------------------
4540
4541 [[merging]]
4542 Merging
4543 -------
4544
4545 -----------------------------------------------
4546 $ git merge test # merge branch "test" into the current branch
4547 $ git pull git://example.com/project.git master
4548 # fetch and merge in remote branch
4549 $ git pull . test # equivalent to git merge test
4550 -----------------------------------------------
4551
4552 [[sharing-your-changes]]
4553 Sharing your changes
4554 --------------------
4555
4556 Importing or exporting patches:
4557
4558 -----------------------------------------------
4559 $ git format-patch origin..HEAD # format a patch for each commit
4560 # in HEAD but not in origin
4561 $ git am mbox # import patches from the mailbox "mbox"
4562 -----------------------------------------------
4563
4564 Fetch a branch in a different git repository, then merge into the
4565 current branch:
4566
4567 -----------------------------------------------
4568 $ git pull git://example.com/project.git theirbranch
4569 -----------------------------------------------
4570
4571 Store the fetched branch into a local branch before merging into the
4572 current branch:
4573
4574 -----------------------------------------------
4575 $ git pull git://example.com/project.git theirbranch:mybranch
4576 -----------------------------------------------
4577
4578 After creating commits on a local branch, update the remote
4579 branch with your commits:
4580
4581 -----------------------------------------------
4582 $ git push ssh://example.com/project.git mybranch:theirbranch
4583 -----------------------------------------------
4584
4585 When remote and local branch are both named "test":
4586
4587 -----------------------------------------------
4588 $ git push ssh://example.com/project.git test
4589 -----------------------------------------------
4590
4591 Shortcut version for a frequently used remote repository:
4592
4593 -----------------------------------------------
4594 $ git remote add example ssh://example.com/project.git
4595 $ git push example test
4596 -----------------------------------------------
4597
4598 [[repository-maintenance]]
4599 Repository maintenance
4600 ----------------------
4601
4602 Check for corruption:
4603
4604 -----------------------------------------------
4605 $ git fsck
4606 -----------------------------------------------
4607
4608 Recompress, remove unused cruft:
4609
4610 -----------------------------------------------
4611 $ git gc
4612 -----------------------------------------------
4613
4614
4615 [[todo]]
4616 Appendix B: Notes and todo list for this manual
4617 ===============================================
4618
4619 This is a work in progress.
4620
4621 The basic requirements:
4622
4623 - It must be readable in order, from beginning to end, by someone
4624 intelligent with a basic grasp of the UNIX command line, but without
4625 any special knowledge of git. If necessary, any other prerequisites
4626 should be specifically mentioned as they arise.
4627 - Whenever possible, section headings should clearly describe the task
4628 they explain how to do, in language that requires no more knowledge
4629 than necessary: for example, "importing patches into a project" rather
4630 than "the `git am` command"
4631
4632 Think about how to create a clear chapter dependency graph that will
4633 allow people to get to important topics without necessarily reading
4634 everything in between.
4635
4636 Scan Documentation/ for other stuff left out; in particular:
4637
4638 - howto's
4639 - some of technical/?
4640 - hooks
4641 - list of commands in linkgit:git[1]
4642
4643 Scan email archives for other stuff left out
4644
4645 Scan man pages to see if any assume more background than this manual
4646 provides.
4647
4648 Simplify beginning by suggesting disconnected head instead of
4649 temporary branch creation?
4650
4651 Add more good examples. Entire sections of just cookbook examples
4652 might be a good idea; maybe make an "advanced examples" section a
4653 standard end-of-chapter section?
4654
4655 Include cross-references to the glossary, where appropriate.
4656
4657 Document shallow clones? See draft 1.5.0 release notes for some
4658 documentation.
4659
4660 Add a section on working with other version control systems, including
4661 CVS, Subversion, and just imports of series of release tarballs.
4662
4663 More details on gitweb?
4664
4665 Write a chapter on using plumbing and writing scripts.
4666
4667 Alternates, clone -reference, etc.
4668
4669 More on recovery from repository corruption. See:
4670 http://marc.theaimsgroup.com/?l=git&m=117263864820799&w=2
4671 http://marc.theaimsgroup.com/?l=git&m=117147855503798&w=2