456457b047af71a2abfb48ed46a2ddd963ef1ee2
[git/git.git] / Documentation / user-manual.txt
1 Git User Manual
2 ===============
3
4 Git is a fast distributed revision control system.
5
6 This manual is designed to be readable by someone with basic UNIX
7 command-line skills, but no previous knowledge of Git.
8
9 <<repositories-and-branches>> and <<exploring-git-history>> explain how
10 to fetch and study a project using git--read these chapters to learn how
11 to build and test a particular version of a software project, search for
12 regressions, and so on.
13
14 People needing to do actual development will also want to read
15 <<Developing-With-git>> and <<sharing-development>>.
16
17 Further chapters cover more specialized topics.
18
19 Comprehensive reference documentation is available through the man
20 pages, or linkgit:git-help[1] command. For example, for the command
21 `git clone <repo>`, you can either use:
22
23 ------------------------------------------------
24 $ man git-clone
25 ------------------------------------------------
26
27 or:
28
29 ------------------------------------------------
30 $ git help clone
31 ------------------------------------------------
32
33 With the latter, you can use the manual viewer of your choice; see
34 linkgit:git-help[1] for more information.
35
36 See also <<git-quick-start>> for a brief overview of Git commands,
37 without any explanation.
38
39 Finally, see <<todo>> for ways that you can help make this manual more
40 complete.
41
42
43 [[repositories-and-branches]]
44 Repositories and Branches
45 =========================
46
47 [[how-to-get-a-git-repository]]
48 How to get a Git repository
49 ---------------------------
50
51 It will be useful to have a Git repository to experiment with as you
52 read this manual.
53
54 The best way to get one is by using the linkgit:git-clone[1] command to
55 download a copy of an existing repository. If you don't already have a
56 project in mind, here are some interesting examples:
57
58 ------------------------------------------------
59 # Git itself (approx. 40MB download):
60 $ git clone git://git.kernel.org/pub/scm/git/git.git
61 # the Linux kernel (approx. 640MB download):
62 $ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
63 ------------------------------------------------
64
65 The initial clone may be time-consuming for a large project, but you
66 will only need to clone once.
67
68 The clone command creates a new directory named after the project
69 (`git` or `linux` in the examples above). After you cd into this
70 directory, you will see that it contains a copy of the project files,
71 called the <<def_working_tree,working tree>>, together with a special
72 top-level directory named `.git`, which contains all the information
73 about the history of the project.
74
75 [[how-to-check-out]]
76 How to check out a different version of a project
77 -------------------------------------------------
78
79 Git is best thought of as a tool for storing the history of a collection
80 of files. It stores the history as a compressed collection of
81 interrelated snapshots of the project's contents. In Git each such
82 version is called a <<def_commit,commit>>.
83
84 Those snapshots aren't necessarily all arranged in a single line from
85 oldest to newest; instead, work may simultaneously proceed along
86 parallel lines of development, called <<def_branch,branches>>, which may
87 merge and diverge.
88
89 A single Git repository can track development on multiple branches. It
90 does this by keeping a list of <<def_head,heads>> which reference the
91 latest commit on each branch; the linkgit:git-branch[1] command shows
92 you the list of branch heads:
93
94 ------------------------------------------------
95 $ git branch
96 * master
97 ------------------------------------------------
98
99 A freshly cloned repository contains a single branch head, by default
100 named "master", with the working directory initialized to the state of
101 the project referred to by that branch head.
102
103 Most projects also use <<def_tag,tags>>. Tags, like heads, are
104 references into the project's history, and can be listed using the
105 linkgit:git-tag[1] command:
106
107 ------------------------------------------------
108 $ git tag -l
109 v2.6.11
110 v2.6.11-tree
111 v2.6.12
112 v2.6.12-rc2
113 v2.6.12-rc3
114 v2.6.12-rc4
115 v2.6.12-rc5
116 v2.6.12-rc6
117 v2.6.13
118 ...
119 ------------------------------------------------
120
121 Tags are expected to always point at the same version of a project,
122 while heads are expected to advance as development progresses.
123
124 Create a new branch head pointing to one of these versions and check it
125 out using linkgit:git-checkout[1]:
126
127 ------------------------------------------------
128 $ git checkout -b new v2.6.13
129 ------------------------------------------------
130
131 The working directory then reflects the contents that the project had
132 when it was tagged v2.6.13, and linkgit:git-branch[1] shows two
133 branches, with an asterisk marking the currently checked-out branch:
134
135 ------------------------------------------------
136 $ git branch
137 master
138 * new
139 ------------------------------------------------
140
141 If you decide that you'd rather see version 2.6.17, you can modify
142 the current branch to point at v2.6.17 instead, with
143
144 ------------------------------------------------
145 $ git reset --hard v2.6.17
146 ------------------------------------------------
147
148 Note that if the current branch head was your only reference to a
149 particular point in history, then resetting that branch may leave you
150 with no way to find the history it used to point to; so use this command
151 carefully.
152
153 [[understanding-commits]]
154 Understanding History: Commits
155 ------------------------------
156
157 Every change in the history of a project is represented by a commit.
158 The linkgit:git-show[1] command shows the most recent commit on the
159 current branch:
160
161 ------------------------------------------------
162 $ git show
163 commit 17cf781661e6d38f737f15f53ab552f1e95960d7
164 Author: Linus Torvalds <torvalds@ppc970.osdl.org.(none)>
165 Date: Tue Apr 19 14:11:06 2005 -0700
166
167 Remove duplicate getenv(DB_ENVIRONMENT) call
168
169 Noted by Tony Luck.
170
171 diff --git a/init-db.c b/init-db.c
172 index 65898fa..b002dc6 100644
173 --- a/init-db.c
174 +++ b/init-db.c
175 @@ -7,7 +7,7 @@
176
177 int main(int argc, char **argv)
178 {
179 - char *sha1_dir = getenv(DB_ENVIRONMENT), *path;
180 + char *sha1_dir, *path;
181 int len, i;
182
183 if (mkdir(".git", 0755) < 0) {
184 ------------------------------------------------
185
186 As you can see, a commit shows who made the latest change, what they
187 did, and why.
188
189 Every commit has a 40-hexdigit id, sometimes called the "object name" or the
190 "SHA-1 id", shown on the first line of the `git show` output. You can usually
191 refer to a commit by a shorter name, such as a tag or a branch name, but this
192 longer name can also be useful. Most importantly, it is a globally unique
193 name for this commit: so if you tell somebody else the object name (for
194 example in email), then you are guaranteed that name will refer to the same
195 commit in their repository that it does in yours (assuming their repository
196 has that commit at all). Since the object name is computed as a hash over the
197 contents of the commit, you are guaranteed that the commit can never change
198 without its name also changing.
199
200 In fact, in <<git-concepts>> we shall see that everything stored in Git
201 history, including file data and directory contents, is stored in an object
202 with a name that is a hash of its contents.
203
204 [[understanding-reachability]]
205 Understanding history: commits, parents, and reachability
206 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
207
208 Every commit (except the very first commit in a project) also has a
209 parent commit which shows what happened before this commit.
210 Following the chain of parents will eventually take you back to the
211 beginning of the project.
212
213 However, the commits do not form a simple list; Git allows lines of
214 development to diverge and then reconverge, and the point where two
215 lines of development reconverge is called a "merge". The commit
216 representing a merge can therefore have more than one parent, with
217 each parent representing the most recent commit on one of the lines
218 of development leading to that point.
219
220 The best way to see how this works is using the linkgit:gitk[1]
221 command; running gitk now on a Git repository and looking for merge
222 commits will help understand how Git organizes history.
223
224 In the following, we say that commit X is "reachable" from commit Y
225 if commit X is an ancestor of commit Y. Equivalently, you could say
226 that Y is a descendant of X, or that there is a chain of parents
227 leading from commit Y to commit X.
228
229 [[history-diagrams]]
230 Understanding history: History diagrams
231 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
232
233 We will sometimes represent Git history using diagrams like the one
234 below. Commits are shown as "o", and the links between them with
235 lines drawn with - / and \. Time goes left to right:
236
237
238 ................................................
239 o--o--o <-- Branch A
240 /
241 o--o--o <-- master
242 \
243 o--o--o <-- Branch B
244 ................................................
245
246 If we need to talk about a particular commit, the character "o" may
247 be replaced with another letter or number.
248
249 [[what-is-a-branch]]
250 Understanding history: What is a branch?
251 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
252
253 When we need to be precise, we will use the word "branch" to mean a line
254 of development, and "branch head" (or just "head") to mean a reference
255 to the most recent commit on a branch. In the example above, the branch
256 head named "A" is a pointer to one particular commit, but we refer to
257 the line of three commits leading up to that point as all being part of
258 "branch A".
259
260 However, when no confusion will result, we often just use the term
261 "branch" both for branches and for branch heads.
262
263 [[manipulating-branches]]
264 Manipulating branches
265 ---------------------
266
267 Creating, deleting, and modifying branches is quick and easy; here's
268 a summary of the commands:
269
270 `git branch`::
271 list all branches.
272 `git branch <branch>`::
273 create a new branch named `<branch>`, referencing the same
274 point in history as the current branch.
275 `git branch <branch> <start-point>`::
276 create a new branch named `<branch>`, referencing
277 `<start-point>`, which may be specified any way you like,
278 including using a branch name or a tag name.
279 `git branch -d <branch>`::
280 delete the branch `<branch>`; if the branch is not fully
281 merged in its upstream branch or contained in the current branch,
282 this command will fail with a warning.
283 `git branch -D <branch>`::
284 delete the branch `<branch>` irrespective of its merged status.
285 `git checkout <branch>`::
286 make the current branch `<branch>`, updating the working
287 directory to reflect the version referenced by `<branch>`.
288 `git checkout -b <new> <start-point>`::
289 create a new branch `<new>` referencing `<start-point>`, and
290 check it out.
291
292 The special symbol "HEAD" can always be used to refer to the current
293 branch. In fact, Git uses a file named `HEAD` in the `.git` directory
294 to remember which branch is current:
295
296 ------------------------------------------------
297 $ cat .git/HEAD
298 ref: refs/heads/master
299 ------------------------------------------------
300
301 [[detached-head]]
302 Examining an old version without creating a new branch
303 ------------------------------------------------------
304
305 The `git checkout` command normally expects a branch head, but will also
306 accept an arbitrary commit; for example, you can check out the commit
307 referenced by a tag:
308
309 ------------------------------------------------
310 $ git checkout v2.6.17
311 Note: checking out 'v2.6.17'.
312
313 You are in 'detached HEAD' state. You can look around, make experimental
314 changes and commit them, and you can discard any commits you make in this
315 state without impacting any branches by performing another checkout.
316
317 If you want to create a new branch to retain commits you create, you may
318 do so (now or later) by using -b with the checkout command again. Example:
319
320 git checkout -b new_branch_name
321
322 HEAD is now at 427abfa... Linux v2.6.17
323 ------------------------------------------------
324
325 The HEAD then refers to the SHA-1 of the commit instead of to a branch,
326 and git branch shows that you are no longer on a branch:
327
328 ------------------------------------------------
329 $ cat .git/HEAD
330 427abfa28afedffadfca9dd8b067eb6d36bac53f
331 $ git branch
332 * (detached from v2.6.17)
333 master
334 ------------------------------------------------
335
336 In this case we say that the HEAD is "detached".
337
338 This is an easy way to check out a particular version without having to
339 make up a name for the new branch. You can still create a new branch
340 (or tag) for this version later if you decide to.
341
342 [[examining-remote-branches]]
343 Examining branches from a remote repository
344 -------------------------------------------
345
346 The "master" branch that was created at the time you cloned is a copy
347 of the HEAD in the repository that you cloned from. That repository
348 may also have had other branches, though, and your local repository
349 keeps branches which track each of those remote branches, called
350 remote-tracking branches, which you
351 can view using the `-r` option to linkgit:git-branch[1]:
352
353 ------------------------------------------------
354 $ git branch -r
355 origin/HEAD
356 origin/html
357 origin/maint
358 origin/man
359 origin/master
360 origin/next
361 origin/pu
362 origin/todo
363 ------------------------------------------------
364
365 In this example, "origin" is called a remote repository, or "remote"
366 for short. The branches of this repository are called "remote
367 branches" from our point of view. The remote-tracking branches listed
368 above were created based on the remote branches at clone time and will
369 be updated by `git fetch` (hence `git pull`) and `git push`. See
370 <<Updating-a-repository-With-git-fetch>> for details.
371
372 You might want to build on one of these remote-tracking branches
373 on a branch of your own, just as you would for a tag:
374
375 ------------------------------------------------
376 $ git checkout -b my-todo-copy origin/todo
377 ------------------------------------------------
378
379 You can also check out `origin/todo` directly to examine it or
380 write a one-off patch. See <<detached-head,detached head>>.
381
382 Note that the name "origin" is just the name that Git uses by default
383 to refer to the repository that you cloned from.
384
385 [[how-git-stores-references]]
386 Naming branches, tags, and other references
387 -------------------------------------------
388
389 Branches, remote-tracking branches, and tags are all references to
390 commits. All references are named with a slash-separated path name
391 starting with `refs`; the names we've been using so far are actually
392 shorthand:
393
394 - The branch `test` is short for `refs/heads/test`.
395 - The tag `v2.6.18` is short for `refs/tags/v2.6.18`.
396 - `origin/master` is short for `refs/remotes/origin/master`.
397
398 The full name is occasionally useful if, for example, there ever
399 exists a tag and a branch with the same name.
400
401 (Newly created refs are actually stored in the `.git/refs` directory,
402 under the path given by their name. However, for efficiency reasons
403 they may also be packed together in a single file; see
404 linkgit:git-pack-refs[1]).
405
406 As another useful shortcut, the "HEAD" of a repository can be referred
407 to just using the name of that repository. So, for example, "origin"
408 is usually a shortcut for the HEAD branch in the repository "origin".
409
410 For the complete list of paths which Git checks for references, and
411 the order it uses to decide which to choose when there are multiple
412 references with the same shorthand name, see the "SPECIFYING
413 REVISIONS" section of linkgit:gitrevisions[7].
414
415 [[Updating-a-repository-With-git-fetch]]
416 Updating a repository with git fetch
417 ------------------------------------
418
419 After you clone a repository and commit a few changes of your own, you
420 may wish to check the original repository for updates.
421
422 The `git-fetch` command, with no arguments, will update all of the
423 remote-tracking branches to the latest version found in the original
424 repository. It will not touch any of your own branches--not even the
425 "master" branch that was created for you on clone.
426
427 [[fetching-branches]]
428 Fetching branches from other repositories
429 -----------------------------------------
430
431 You can also track branches from repositories other than the one you
432 cloned from, using linkgit:git-remote[1]:
433
434 -------------------------------------------------
435 $ git remote add staging git://git.kernel.org/.../gregkh/staging.git
436 $ git fetch staging
437 ...
438 From git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
439 * [new branch] master -> staging/master
440 * [new branch] staging-linus -> staging/staging-linus
441 * [new branch] staging-next -> staging/staging-next
442 -------------------------------------------------
443
444 New remote-tracking branches will be stored under the shorthand name
445 that you gave `git remote add`, in this case `staging`:
446
447 -------------------------------------------------
448 $ git branch -r
449 origin/HEAD -> origin/master
450 origin/master
451 staging/master
452 staging/staging-linus
453 staging/staging-next
454 -------------------------------------------------
455
456 If you run `git fetch <remote>` later, the remote-tracking branches
457 for the named `<remote>` will be updated.
458
459 If you examine the file `.git/config`, you will see that Git has added
460 a new stanza:
461
462 -------------------------------------------------
463 $ cat .git/config
464 ...
465 [remote "staging"]
466 url = git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging.git
467 fetch = +refs/heads/*:refs/remotes/staging/*
468 ...
469 -------------------------------------------------
470
471 This is what causes Git to track the remote's branches; you may modify
472 or delete these configuration options by editing `.git/config` with a
473 text editor. (See the "CONFIGURATION FILE" section of
474 linkgit:git-config[1] for details.)
475
476 [[exploring-git-history]]
477 Exploring Git history
478 =====================
479
480 Git is best thought of as a tool for storing the history of a
481 collection of files. It does this by storing compressed snapshots of
482 the contents of a file hierarchy, together with "commits" which show
483 the relationships between these snapshots.
484
485 Git provides extremely flexible and fast tools for exploring the
486 history of a project.
487
488 We start with one specialized tool that is useful for finding the
489 commit that introduced a bug into a project.
490
491 [[using-bisect]]
492 How to use bisect to find a regression
493 --------------------------------------
494
495 Suppose version 2.6.18 of your project worked, but the version at
496 "master" crashes. Sometimes the best way to find the cause of such a
497 regression is to perform a brute-force search through the project's
498 history to find the particular commit that caused the problem. The
499 linkgit:git-bisect[1] command can help you do this:
500
501 -------------------------------------------------
502 $ git bisect start
503 $ git bisect good v2.6.18
504 $ git bisect bad master
505 Bisecting: 3537 revisions left to test after this
506 [65934a9a028b88e83e2b0f8b36618fe503349f8e] BLOCK: Make USB storage depend on SCSI rather than selecting it [try #6]
507 -------------------------------------------------
508
509 If you run `git branch` at this point, you'll see that Git has
510 temporarily moved you in "(no branch)". HEAD is now detached from any
511 branch and points directly to a commit (with commit id 65934...) that
512 is reachable from "master" but not from v2.6.18. Compile and test it,
513 and see whether it crashes. Assume it does crash. Then:
514
515 -------------------------------------------------
516 $ git bisect bad
517 Bisecting: 1769 revisions left to test after this
518 [7eff82c8b1511017ae605f0c99ac275a7e21b867] i2c-core: Drop useless bitmaskings
519 -------------------------------------------------
520
521 checks out an older version. Continue like this, telling Git at each
522 stage whether the version it gives you is good or bad, and notice
523 that the number of revisions left to test is cut approximately in
524 half each time.
525
526 After about 13 tests (in this case), it will output the commit id of
527 the guilty commit. You can then examine the commit with
528 linkgit:git-show[1], find out who wrote it, and mail them your bug
529 report with the commit id. Finally, run
530
531 -------------------------------------------------
532 $ git bisect reset
533 -------------------------------------------------
534
535 to return you to the branch you were on before.
536
537 Note that the version which `git bisect` checks out for you at each
538 point is just a suggestion, and you're free to try a different
539 version if you think it would be a good idea. For example,
540 occasionally you may land on a commit that broke something unrelated;
541 run
542
543 -------------------------------------------------
544 $ git bisect visualize
545 -------------------------------------------------
546
547 which will run gitk and label the commit it chose with a marker that
548 says "bisect". Choose a safe-looking commit nearby, note its commit
549 id, and check it out with:
550
551 -------------------------------------------------
552 $ git reset --hard fb47ddb2db...
553 -------------------------------------------------
554
555 then test, run `bisect good` or `bisect bad` as appropriate, and
556 continue.
557
558 Instead of `git bisect visualize` and then `git reset --hard
559 fb47ddb2db...`, you might just want to tell Git that you want to skip
560 the current commit:
561
562 -------------------------------------------------
563 $ git bisect skip
564 -------------------------------------------------
565
566 In this case, though, Git may not eventually be able to tell the first
567 bad one between some first skipped commits and a later bad commit.
568
569 There are also ways to automate the bisecting process if you have a
570 test script that can tell a good from a bad commit. See
571 linkgit:git-bisect[1] for more information about this and other `git
572 bisect` features.
573
574 [[naming-commits]]
575 Naming commits
576 --------------
577
578 We have seen several ways of naming commits already:
579
580 - 40-hexdigit object name
581 - branch name: refers to the commit at the head of the given
582 branch
583 - tag name: refers to the commit pointed to by the given tag
584 (we've seen branches and tags are special cases of
585 <<how-git-stores-references,references>>).
586 - HEAD: refers to the head of the current branch
587
588 There are many more; see the "SPECIFYING REVISIONS" section of the
589 linkgit:gitrevisions[7] man page for the complete list of ways to
590 name revisions. Some examples:
591
592 -------------------------------------------------
593 $ git show fb47ddb2 # the first few characters of the object name
594 # are usually enough to specify it uniquely
595 $ git show HEAD^ # the parent of the HEAD commit
596 $ git show HEAD^^ # the grandparent
597 $ git show HEAD~4 # the great-great-grandparent
598 -------------------------------------------------
599
600 Recall that merge commits may have more than one parent; by default,
601 `^` and `~` follow the first parent listed in the commit, but you can
602 also choose:
603
604 -------------------------------------------------
605 $ git show HEAD^1 # show the first parent of HEAD
606 $ git show HEAD^2 # show the second parent of HEAD
607 -------------------------------------------------
608
609 In addition to HEAD, there are several other special names for
610 commits:
611
612 Merges (to be discussed later), as well as operations such as
613 `git reset`, which change the currently checked-out commit, generally
614 set ORIG_HEAD to the value HEAD had before the current operation.
615
616 The `git fetch` operation always stores the head of the last fetched
617 branch in FETCH_HEAD. For example, if you run `git fetch` without
618 specifying a local branch as the target of the operation
619
620 -------------------------------------------------
621 $ git fetch git://example.com/proj.git theirbranch
622 -------------------------------------------------
623
624 the fetched commits will still be available from FETCH_HEAD.
625
626 When we discuss merges we'll also see the special name MERGE_HEAD,
627 which refers to the other branch that we're merging in to the current
628 branch.
629
630 The linkgit:git-rev-parse[1] command is a low-level command that is
631 occasionally useful for translating some name for a commit to the object
632 name for that commit:
633
634 -------------------------------------------------
635 $ git rev-parse origin
636 e05db0fd4f31dde7005f075a84f96b360d05984b
637 -------------------------------------------------
638
639 [[creating-tags]]
640 Creating tags
641 -------------
642
643 We can also create a tag to refer to a particular commit; after
644 running
645
646 -------------------------------------------------
647 $ git tag stable-1 1b2e1d63ff
648 -------------------------------------------------
649
650 You can use `stable-1` to refer to the commit 1b2e1d63ff.
651
652 This creates a "lightweight" tag. If you would also like to include a
653 comment with the tag, and possibly sign it cryptographically, then you
654 should create a tag object instead; see the linkgit:git-tag[1] man page
655 for details.
656
657 [[browsing-revisions]]
658 Browsing revisions
659 ------------------
660
661 The linkgit:git-log[1] command can show lists of commits. On its
662 own, it shows all commits reachable from the parent commit; but you
663 can also make more specific requests:
664
665 -------------------------------------------------
666 $ git log v2.5.. # commits since (not reachable from) v2.5
667 $ git log test..master # commits reachable from master but not test
668 $ git log master..test # ...reachable from test but not master
669 $ git log master...test # ...reachable from either test or master,
670 # but not both
671 $ git log --since="2 weeks ago" # commits from the last 2 weeks
672 $ git log Makefile # commits which modify Makefile
673 $ git log fs/ # ... which modify any file under fs/
674 $ git log -S'foo()' # commits which add or remove any file data
675 # matching the string 'foo()'
676 -------------------------------------------------
677
678 And of course you can combine all of these; the following finds
679 commits since v2.5 which touch the `Makefile` or any file under `fs`:
680
681 -------------------------------------------------
682 $ git log v2.5.. Makefile fs/
683 -------------------------------------------------
684
685 You can also ask git log to show patches:
686
687 -------------------------------------------------
688 $ git log -p
689 -------------------------------------------------
690
691 See the `--pretty` option in the linkgit:git-log[1] man page for more
692 display options.
693
694 Note that git log starts with the most recent commit and works
695 backwards through the parents; however, since Git history can contain
696 multiple independent lines of development, the particular order that
697 commits are listed in may be somewhat arbitrary.
698
699 [[generating-diffs]]
700 Generating diffs
701 ----------------
702
703 You can generate diffs between any two versions using
704 linkgit:git-diff[1]:
705
706 -------------------------------------------------
707 $ git diff master..test
708 -------------------------------------------------
709
710 That will produce the diff between the tips of the two branches. If
711 you'd prefer to find the diff from their common ancestor to test, you
712 can use three dots instead of two:
713
714 -------------------------------------------------
715 $ git diff master...test
716 -------------------------------------------------
717
718 Sometimes what you want instead is a set of patches; for this you can
719 use linkgit:git-format-patch[1]:
720
721 -------------------------------------------------
722 $ git format-patch master..test
723 -------------------------------------------------
724
725 will generate a file with a patch for each commit reachable from test
726 but not from master.
727
728 [[viewing-old-file-versions]]
729 Viewing old file versions
730 -------------------------
731
732 You can always view an old version of a file by just checking out the
733 correct revision first. But sometimes it is more convenient to be
734 able to view an old version of a single file without checking
735 anything out; this command does that:
736
737 -------------------------------------------------
738 $ git show v2.5:fs/locks.c
739 -------------------------------------------------
740
741 Before the colon may be anything that names a commit, and after it
742 may be any path to a file tracked by Git.
743
744 [[history-examples]]
745 Examples
746 --------
747
748 [[counting-commits-on-a-branch]]
749 Counting the number of commits on a branch
750 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
751
752 Suppose you want to know how many commits you've made on `mybranch`
753 since it diverged from `origin`:
754
755 -------------------------------------------------
756 $ git log --pretty=oneline origin..mybranch | wc -l
757 -------------------------------------------------
758
759 Alternatively, you may often see this sort of thing done with the
760 lower-level command linkgit:git-rev-list[1], which just lists the SHA-1's
761 of all the given commits:
762
763 -------------------------------------------------
764 $ git rev-list origin..mybranch | wc -l
765 -------------------------------------------------
766
767 [[checking-for-equal-branches]]
768 Check whether two branches point at the same history
769 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
770
771 Suppose you want to check whether two branches point at the same point
772 in history.
773
774 -------------------------------------------------
775 $ git diff origin..master
776 -------------------------------------------------
777
778 will tell you whether the contents of the project are the same at the
779 two branches; in theory, however, it's possible that the same project
780 contents could have been arrived at by two different historical
781 routes. You could compare the object names:
782
783 -------------------------------------------------
784 $ git rev-list origin
785 e05db0fd4f31dde7005f075a84f96b360d05984b
786 $ git rev-list master
787 e05db0fd4f31dde7005f075a84f96b360d05984b
788 -------------------------------------------------
789
790 Or you could recall that the `...` operator selects all commits
791 reachable from either one reference or the other but not
792 both; so
793
794 -------------------------------------------------
795 $ git log origin...master
796 -------------------------------------------------
797
798 will return no commits when the two branches are equal.
799
800 [[finding-tagged-descendants]]
801 Find first tagged version including a given fix
802 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
803
804 Suppose you know that the commit e05db0fd fixed a certain problem.
805 You'd like to find the earliest tagged release that contains that
806 fix.
807
808 Of course, there may be more than one answer--if the history branched
809 after commit e05db0fd, then there could be multiple "earliest" tagged
810 releases.
811
812 You could just visually inspect the commits since e05db0fd:
813
814 -------------------------------------------------
815 $ gitk e05db0fd..
816 -------------------------------------------------
817
818 or you can use linkgit:git-name-rev[1], which will give the commit a
819 name based on any tag it finds pointing to one of the commit's
820 descendants:
821
822 -------------------------------------------------
823 $ git name-rev --tags e05db0fd
824 e05db0fd tags/v1.5.0-rc1^0~23
825 -------------------------------------------------
826
827 The linkgit:git-describe[1] command does the opposite, naming the
828 revision using a tag on which the given commit is based:
829
830 -------------------------------------------------
831 $ git describe e05db0fd
832 v1.5.0-rc0-260-ge05db0f
833 -------------------------------------------------
834
835 but that may sometimes help you guess which tags might come after the
836 given commit.
837
838 If you just want to verify whether a given tagged version contains a
839 given commit, you could use linkgit:git-merge-base[1]:
840
841 -------------------------------------------------
842 $ git merge-base e05db0fd v1.5.0-rc1
843 e05db0fd4f31dde7005f075a84f96b360d05984b
844 -------------------------------------------------
845
846 The merge-base command finds a common ancestor of the given commits,
847 and always returns one or the other in the case where one is a
848 descendant of the other; so the above output shows that e05db0fd
849 actually is an ancestor of v1.5.0-rc1.
850
851 Alternatively, note that
852
853 -------------------------------------------------
854 $ git log v1.5.0-rc1..e05db0fd
855 -------------------------------------------------
856
857 will produce empty output if and only if v1.5.0-rc1 includes e05db0fd,
858 because it outputs only commits that are not reachable from v1.5.0-rc1.
859
860 As yet another alternative, the linkgit:git-show-branch[1] command lists
861 the commits reachable from its arguments with a display on the left-hand
862 side that indicates which arguments that commit is reachable from.
863 So, if you run something like
864
865 -------------------------------------------------
866 $ git show-branch e05db0fd v1.5.0-rc0 v1.5.0-rc1 v1.5.0-rc2
867 ! [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
868 available
869 ! [v1.5.0-rc0] GIT v1.5.0 preview
870 ! [v1.5.0-rc1] GIT v1.5.0-rc1
871 ! [v1.5.0-rc2] GIT v1.5.0-rc2
872 ...
873 -------------------------------------------------
874
875 then a line like
876
877 -------------------------------------------------
878 + ++ [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
879 available
880 -------------------------------------------------
881
882 shows that e05db0fd is reachable from itself, from v1.5.0-rc1,
883 and from v1.5.0-rc2, and not from v1.5.0-rc0.
884
885 [[showing-commits-unique-to-a-branch]]
886 Showing commits unique to a given branch
887 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
888
889 Suppose you would like to see all the commits reachable from the branch
890 head named `master` but not from any other head in your repository.
891
892 We can list all the heads in this repository with
893 linkgit:git-show-ref[1]:
894
895 -------------------------------------------------
896 $ git show-ref --heads
897 bf62196b5e363d73353a9dcf094c59595f3153b7 refs/heads/core-tutorial
898 db768d5504c1bb46f63ee9d6e1772bd047e05bf9 refs/heads/maint
899 a07157ac624b2524a059a3414e99f6f44bebc1e7 refs/heads/master
900 24dbc180ea14dc1aebe09f14c8ecf32010690627 refs/heads/tutorial-2
901 1e87486ae06626c2f31eaa63d26fc0fd646c8af2 refs/heads/tutorial-fixes
902 -------------------------------------------------
903
904 We can get just the branch-head names, and remove `master`, with
905 the help of the standard utilities cut and grep:
906
907 -------------------------------------------------
908 $ git show-ref --heads | cut -d' ' -f2 | grep -v '^refs/heads/master'
909 refs/heads/core-tutorial
910 refs/heads/maint
911 refs/heads/tutorial-2
912 refs/heads/tutorial-fixes
913 -------------------------------------------------
914
915 And then we can ask to see all the commits reachable from master
916 but not from these other heads:
917
918 -------------------------------------------------
919 $ gitk master --not $( git show-ref --heads | cut -d' ' -f2 |
920 grep -v '^refs/heads/master' )
921 -------------------------------------------------
922
923 Obviously, endless variations are possible; for example, to see all
924 commits reachable from some head but not from any tag in the repository:
925
926 -------------------------------------------------
927 $ gitk $( git show-ref --heads ) --not $( git show-ref --tags )
928 -------------------------------------------------
929
930 (See linkgit:gitrevisions[7] for explanations of commit-selecting
931 syntax such as `--not`.)
932
933 [[making-a-release]]
934 Creating a changelog and tarball for a software release
935 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
936
937 The linkgit:git-archive[1] command can create a tar or zip archive from
938 any version of a project; for example:
939
940 -------------------------------------------------
941 $ git archive -o latest.tar.gz --prefix=project/ HEAD
942 -------------------------------------------------
943
944 will use HEAD to produce a gzipped tar archive in which each filename
945 is preceded by `project/`. The output file format is inferred from
946 the output file extension if possible, see linkgit:git-archive[1] for
947 details.
948
949 Versions of Git older than 1.7.7 don't know about the `tar.gz` format,
950 you'll need to use gzip explicitly:
951
952 -------------------------------------------------
953 $ git archive --format=tar --prefix=project/ HEAD | gzip >latest.tar.gz
954 -------------------------------------------------
955
956 If you're releasing a new version of a software project, you may want
957 to simultaneously make a changelog to include in the release
958 announcement.
959
960 Linus Torvalds, for example, makes new kernel releases by tagging them,
961 then running:
962
963 -------------------------------------------------
964 $ release-script 2.6.12 2.6.13-rc6 2.6.13-rc7
965 -------------------------------------------------
966
967 where release-script is a shell script that looks like:
968
969 -------------------------------------------------
970 #!/bin/sh
971 stable="$1"
972 last="$2"
973 new="$3"
974 echo "# git tag v$new"
975 echo "git archive --prefix=linux-$new/ v$new | gzip -9 > ../linux-$new.tar.gz"
976 echo "git diff v$stable v$new | gzip -9 > ../patch-$new.gz"
977 echo "git log --no-merges v$new ^v$last > ../ChangeLog-$new"
978 echo "git shortlog --no-merges v$new ^v$last > ../ShortLog"
979 echo "git diff --stat --summary -M v$last v$new > ../diffstat-$new"
980 -------------------------------------------------
981
982 and then he just cut-and-pastes the output commands after verifying that
983 they look OK.
984
985 [[Finding-commits-With-given-Content]]
986 Finding commits referencing a file with given content
987 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
988
989 Somebody hands you a copy of a file, and asks which commits modified a
990 file such that it contained the given content either before or after the
991 commit. You can find out with this:
992
993 -------------------------------------------------
994 $ git log --raw --abbrev=40 --pretty=oneline |
995 grep -B 1 `git hash-object filename`
996 -------------------------------------------------
997
998 Figuring out why this works is left as an exercise to the (advanced)
999 student. The linkgit:git-log[1], linkgit:git-diff-tree[1], and
1000 linkgit:git-hash-object[1] man pages may prove helpful.
1001
1002 [[Developing-With-git]]
1003 Developing with Git
1004 ===================
1005
1006 [[telling-git-your-name]]
1007 Telling Git your name
1008 ---------------------
1009
1010 Before creating any commits, you should introduce yourself to Git.
1011 The easiest way to do so is to use linkgit:git-config[1]:
1012
1013 ------------------------------------------------
1014 $ git config --global user.name 'Your Name Comes Here'
1015 $ git config --global user.email 'you@yourdomain.example.com'
1016 ------------------------------------------------
1017
1018 Which will add the following to a file named `.gitconfig` in your
1019 home directory:
1020
1021 ------------------------------------------------
1022 [user]
1023 name = Your Name Comes Here
1024 email = you@yourdomain.example.com
1025 ------------------------------------------------
1026
1027 See the "CONFIGURATION FILE" section of linkgit:git-config[1] for
1028 details on the configuration file. The file is plain text, so you can
1029 also edit it with your favorite editor.
1030
1031
1032 [[creating-a-new-repository]]
1033 Creating a new repository
1034 -------------------------
1035
1036 Creating a new repository from scratch is very easy:
1037
1038 -------------------------------------------------
1039 $ mkdir project
1040 $ cd project
1041 $ git init
1042 -------------------------------------------------
1043
1044 If you have some initial content (say, a tarball):
1045
1046 -------------------------------------------------
1047 $ tar xzvf project.tar.gz
1048 $ cd project
1049 $ git init
1050 $ git add . # include everything below ./ in the first commit:
1051 $ git commit
1052 -------------------------------------------------
1053
1054 [[how-to-make-a-commit]]
1055 How to make a commit
1056 --------------------
1057
1058 Creating a new commit takes three steps:
1059
1060 1. Making some changes to the working directory using your
1061 favorite editor.
1062 2. Telling Git about your changes.
1063 3. Creating the commit using the content you told Git about
1064 in step 2.
1065
1066 In practice, you can interleave and repeat steps 1 and 2 as many
1067 times as you want: in order to keep track of what you want committed
1068 at step 3, Git maintains a snapshot of the tree's contents in a
1069 special staging area called "the index."
1070
1071 At the beginning, the content of the index will be identical to
1072 that of the HEAD. The command `git diff --cached`, which shows
1073 the difference between the HEAD and the index, should therefore
1074 produce no output at that point.
1075
1076 Modifying the index is easy:
1077
1078 To update the index with the contents of a new or modified file, use
1079
1080 -------------------------------------------------
1081 $ git add path/to/file
1082 -------------------------------------------------
1083
1084 To remove a file from the index and from the working tree, use
1085
1086 -------------------------------------------------
1087 $ git rm path/to/file
1088 -------------------------------------------------
1089
1090 After each step you can verify that
1091
1092 -------------------------------------------------
1093 $ git diff --cached
1094 -------------------------------------------------
1095
1096 always shows the difference between the HEAD and the index file--this
1097 is what you'd commit if you created the commit now--and that
1098
1099 -------------------------------------------------
1100 $ git diff
1101 -------------------------------------------------
1102
1103 shows the difference between the working tree and the index file.
1104
1105 Note that `git add` always adds just the current contents of a file
1106 to the index; further changes to the same file will be ignored unless
1107 you run `git add` on the file again.
1108
1109 When you're ready, just run
1110
1111 -------------------------------------------------
1112 $ git commit
1113 -------------------------------------------------
1114
1115 and Git will prompt you for a commit message and then create the new
1116 commit. Check to make sure it looks like what you expected with
1117
1118 -------------------------------------------------
1119 $ git show
1120 -------------------------------------------------
1121
1122 As a special shortcut,
1123
1124 -------------------------------------------------
1125 $ git commit -a
1126 -------------------------------------------------
1127
1128 will update the index with any files that you've modified or removed
1129 and create a commit, all in one step.
1130
1131 A number of commands are useful for keeping track of what you're
1132 about to commit:
1133
1134 -------------------------------------------------
1135 $ git diff --cached # difference between HEAD and the index; what
1136 # would be committed if you ran "commit" now.
1137 $ git diff # difference between the index file and your
1138 # working directory; changes that would not
1139 # be included if you ran "commit" now.
1140 $ git diff HEAD # difference between HEAD and working tree; what
1141 # would be committed if you ran "commit -a" now.
1142 $ git status # a brief per-file summary of the above.
1143 -------------------------------------------------
1144
1145 You can also use linkgit:git-gui[1] to create commits, view changes in
1146 the index and the working tree files, and individually select diff hunks
1147 for inclusion in the index (by right-clicking on the diff hunk and
1148 choosing "Stage Hunk For Commit").
1149
1150 [[creating-good-commit-messages]]
1151 Creating good commit messages
1152 -----------------------------
1153
1154 Though not required, it's a good idea to begin the commit message
1155 with a single short (less than 50 character) line summarizing the
1156 change, followed by a blank line and then a more thorough
1157 description. The text up to the first blank line in a commit
1158 message is treated as the commit title, and that title is used
1159 throughout Git. For example, linkgit:git-format-patch[1] turns a
1160 commit into email, and it uses the title on the Subject line and the
1161 rest of the commit in the body.
1162
1163
1164 [[ignoring-files]]
1165 Ignoring files
1166 --------------
1167
1168 A project will often generate files that you do 'not' want to track with Git.
1169 This typically includes files generated by a build process or temporary
1170 backup files made by your editor. Of course, 'not' tracking files with Git
1171 is just a matter of 'not' calling `git add` on them. But it quickly becomes
1172 annoying to have these untracked files lying around; e.g. they make
1173 `git add .` practically useless, and they keep showing up in the output of
1174 `git status`.
1175
1176 You can tell Git to ignore certain files by creating a file called
1177 `.gitignore` in the top level of your working directory, with contents
1178 such as:
1179
1180 -------------------------------------------------
1181 # Lines starting with '#' are considered comments.
1182 # Ignore any file named foo.txt.
1183 foo.txt
1184 # Ignore (generated) html files,
1185 *.html
1186 # except foo.html which is maintained by hand.
1187 !foo.html
1188 # Ignore objects and archives.
1189 *.[oa]
1190 -------------------------------------------------
1191
1192 See linkgit:gitignore[5] for a detailed explanation of the syntax. You can
1193 also place .gitignore files in other directories in your working tree, and they
1194 will apply to those directories and their subdirectories. The `.gitignore`
1195 files can be added to your repository like any other files (just run `git add
1196 .gitignore` and `git commit`, as usual), which is convenient when the exclude
1197 patterns (such as patterns matching build output files) would also make sense
1198 for other users who clone your repository.
1199
1200 If you wish the exclude patterns to affect only certain repositories
1201 (instead of every repository for a given project), you may instead put
1202 them in a file in your repository named `.git/info/exclude`, or in any
1203 file specified by the `core.excludesfile` configuration variable.
1204 Some Git commands can also take exclude patterns directly on the
1205 command line. See linkgit:gitignore[5] for the details.
1206
1207 [[how-to-merge]]
1208 How to merge
1209 ------------
1210
1211 You can rejoin two diverging branches of development using
1212 linkgit:git-merge[1]:
1213
1214 -------------------------------------------------
1215 $ git merge branchname
1216 -------------------------------------------------
1217
1218 merges the development in the branch `branchname` into the current
1219 branch.
1220
1221 A merge is made by combining the changes made in `branchname` and the
1222 changes made up to the latest commit in your current branch since
1223 their histories forked. The work tree is overwritten by the result of
1224 the merge when this combining is done cleanly, or overwritten by a
1225 half-merged results when this combining results in conflicts.
1226 Therefore, if you have uncommitted changes touching the same files as
1227 the ones impacted by the merge, Git will refuse to proceed. Most of
1228 the time, you will want to commit your changes before you can merge,
1229 and if you don't, then linkgit:git-stash[1] can take these changes
1230 away while you're doing the merge, and reapply them afterwards.
1231
1232 If the changes are independent enough, Git will automatically complete
1233 the merge and commit the result (or reuse an existing commit in case
1234 of <<fast-forwards,fast-forward>>, see below). On the other hand,
1235 if there are conflicts--for example, if the same file is
1236 modified in two different ways in the remote branch and the local
1237 branch--then you are warned; the output may look something like this:
1238
1239 -------------------------------------------------
1240 $ git merge next
1241 100% (4/4) done
1242 Auto-merged file.txt
1243 CONFLICT (content): Merge conflict in file.txt
1244 Automatic merge failed; fix conflicts and then commit the result.
1245 -------------------------------------------------
1246
1247 Conflict markers are left in the problematic files, and after
1248 you resolve the conflicts manually, you can update the index
1249 with the contents and run Git commit, as you normally would when
1250 creating a new file.
1251
1252 If you examine the resulting commit using gitk, you will see that it
1253 has two parents, one pointing to the top of the current branch, and
1254 one to the top of the other branch.
1255
1256 [[resolving-a-merge]]
1257 Resolving a merge
1258 -----------------
1259
1260 When a merge isn't resolved automatically, Git leaves the index and
1261 the working tree in a special state that gives you all the
1262 information you need to help resolve the merge.
1263
1264 Files with conflicts are marked specially in the index, so until you
1265 resolve the problem and update the index, linkgit:git-commit[1] will
1266 fail:
1267
1268 -------------------------------------------------
1269 $ git commit
1270 file.txt: needs merge
1271 -------------------------------------------------
1272
1273 Also, linkgit:git-status[1] will list those files as "unmerged", and the
1274 files with conflicts will have conflict markers added, like this:
1275
1276 -------------------------------------------------
1277 <<<<<<< HEAD:file.txt
1278 Hello world
1279 =======
1280 Goodbye
1281 >>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1282 -------------------------------------------------
1283
1284 All you need to do is edit the files to resolve the conflicts, and then
1285
1286 -------------------------------------------------
1287 $ git add file.txt
1288 $ git commit
1289 -------------------------------------------------
1290
1291 Note that the commit message will already be filled in for you with
1292 some information about the merge. Normally you can just use this
1293 default message unchanged, but you may add additional commentary of
1294 your own if desired.
1295
1296 The above is all you need to know to resolve a simple merge. But Git
1297 also provides more information to help resolve conflicts:
1298
1299 [[conflict-resolution]]
1300 Getting conflict-resolution help during a merge
1301 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1302
1303 All of the changes that Git was able to merge automatically are
1304 already added to the index file, so linkgit:git-diff[1] shows only
1305 the conflicts. It uses an unusual syntax:
1306
1307 -------------------------------------------------
1308 $ git diff
1309 diff --cc file.txt
1310 index 802992c,2b60207..0000000
1311 --- a/file.txt
1312 +++ b/file.txt
1313 @@@ -1,1 -1,1 +1,5 @@@
1314 ++<<<<<<< HEAD:file.txt
1315 +Hello world
1316 ++=======
1317 + Goodbye
1318 ++>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1319 -------------------------------------------------
1320
1321 Recall that the commit which will be committed after we resolve this
1322 conflict will have two parents instead of the usual one: one parent
1323 will be HEAD, the tip of the current branch; the other will be the
1324 tip of the other branch, which is stored temporarily in MERGE_HEAD.
1325
1326 During the merge, the index holds three versions of each file. Each of
1327 these three "file stages" represents a different version of the file:
1328
1329 -------------------------------------------------
1330 $ git show :1:file.txt # the file in a common ancestor of both branches
1331 $ git show :2:file.txt # the version from HEAD.
1332 $ git show :3:file.txt # the version from MERGE_HEAD.
1333 -------------------------------------------------
1334
1335 When you ask linkgit:git-diff[1] to show the conflicts, it runs a
1336 three-way diff between the conflicted merge results in the work tree with
1337 stages 2 and 3 to show only hunks whose contents come from both sides,
1338 mixed (in other words, when a hunk's merge results come only from stage 2,
1339 that part is not conflicting and is not shown. Same for stage 3).
1340
1341 The diff above shows the differences between the working-tree version of
1342 file.txt and the stage 2 and stage 3 versions. So instead of preceding
1343 each line by a single `+` or `-`, it now uses two columns: the first
1344 column is used for differences between the first parent and the working
1345 directory copy, and the second for differences between the second parent
1346 and the working directory copy. (See the "COMBINED DIFF FORMAT" section
1347 of linkgit:git-diff-files[1] for a details of the format.)
1348
1349 After resolving the conflict in the obvious way (but before updating the
1350 index), the diff will look like:
1351
1352 -------------------------------------------------
1353 $ git diff
1354 diff --cc file.txt
1355 index 802992c,2b60207..0000000
1356 --- a/file.txt
1357 +++ b/file.txt
1358 @@@ -1,1 -1,1 +1,1 @@@
1359 - Hello world
1360 -Goodbye
1361 ++Goodbye world
1362 -------------------------------------------------
1363
1364 This shows that our resolved version deleted "Hello world" from the
1365 first parent, deleted "Goodbye" from the second parent, and added
1366 "Goodbye world", which was previously absent from both.
1367
1368 Some special diff options allow diffing the working directory against
1369 any of these stages:
1370
1371 -------------------------------------------------
1372 $ git diff -1 file.txt # diff against stage 1
1373 $ git diff --base file.txt # same as the above
1374 $ git diff -2 file.txt # diff against stage 2
1375 $ git diff --ours file.txt # same as the above
1376 $ git diff -3 file.txt # diff against stage 3
1377 $ git diff --theirs file.txt # same as the above.
1378 -------------------------------------------------
1379
1380 The linkgit:git-log[1] and linkgit:gitk[1] commands also provide special help
1381 for merges:
1382
1383 -------------------------------------------------
1384 $ git log --merge
1385 $ gitk --merge
1386 -------------------------------------------------
1387
1388 These will display all commits which exist only on HEAD or on
1389 MERGE_HEAD, and which touch an unmerged file.
1390
1391 You may also use linkgit:git-mergetool[1], which lets you merge the
1392 unmerged files using external tools such as Emacs or kdiff3.
1393
1394 Each time you resolve the conflicts in a file and update the index:
1395
1396 -------------------------------------------------
1397 $ git add file.txt
1398 -------------------------------------------------
1399
1400 the different stages of that file will be "collapsed", after which
1401 `git diff` will (by default) no longer show diffs for that file.
1402
1403 [[undoing-a-merge]]
1404 Undoing a merge
1405 ---------------
1406
1407 If you get stuck and decide to just give up and throw the whole mess
1408 away, you can always return to the pre-merge state with
1409
1410 -------------------------------------------------
1411 $ git reset --hard HEAD
1412 -------------------------------------------------
1413
1414 Or, if you've already committed the merge that you want to throw away,
1415
1416 -------------------------------------------------
1417 $ git reset --hard ORIG_HEAD
1418 -------------------------------------------------
1419
1420 However, this last command can be dangerous in some cases--never
1421 throw away a commit you have already committed if that commit may
1422 itself have been merged into another branch, as doing so may confuse
1423 further merges.
1424
1425 [[fast-forwards]]
1426 Fast-forward merges
1427 -------------------
1428
1429 There is one special case not mentioned above, which is treated
1430 differently. Normally, a merge results in a merge commit, with two
1431 parents, one pointing at each of the two lines of development that
1432 were merged.
1433
1434 However, if the current branch is a descendant of the other--so every
1435 commit present in the one is already contained in the other--then Git
1436 just performs a "fast-forward"; the head of the current branch is moved
1437 forward to point at the head of the merged-in branch, without any new
1438 commits being created.
1439
1440 [[fixing-mistakes]]
1441 Fixing mistakes
1442 ---------------
1443
1444 If you've messed up the working tree, but haven't yet committed your
1445 mistake, you can return the entire working tree to the last committed
1446 state with
1447
1448 -------------------------------------------------
1449 $ git reset --hard HEAD
1450 -------------------------------------------------
1451
1452 If you make a commit that you later wish you hadn't, there are two
1453 fundamentally different ways to fix the problem:
1454
1455 1. You can create a new commit that undoes whatever was done
1456 by the old commit. This is the correct thing if your
1457 mistake has already been made public.
1458
1459 2. You can go back and modify the old commit. You should
1460 never do this if you have already made the history public;
1461 Git does not normally expect the "history" of a project to
1462 change, and cannot correctly perform repeated merges from
1463 a branch that has had its history changed.
1464
1465 [[reverting-a-commit]]
1466 Fixing a mistake with a new commit
1467 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1468
1469 Creating a new commit that reverts an earlier change is very easy;
1470 just pass the linkgit:git-revert[1] command a reference to the bad
1471 commit; for example, to revert the most recent commit:
1472
1473 -------------------------------------------------
1474 $ git revert HEAD
1475 -------------------------------------------------
1476
1477 This will create a new commit which undoes the change in HEAD. You
1478 will be given a chance to edit the commit message for the new commit.
1479
1480 You can also revert an earlier change, for example, the next-to-last:
1481
1482 -------------------------------------------------
1483 $ git revert HEAD^
1484 -------------------------------------------------
1485
1486 In this case Git will attempt to undo the old change while leaving
1487 intact any changes made since then. If more recent changes overlap
1488 with the changes to be reverted, then you will be asked to fix
1489 conflicts manually, just as in the case of <<resolving-a-merge,
1490 resolving a merge>>.
1491
1492 [[fixing-a-mistake-by-rewriting-history]]
1493 Fixing a mistake by rewriting history
1494 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1495
1496 If the problematic commit is the most recent commit, and you have not
1497 yet made that commit public, then you may just
1498 <<undoing-a-merge,destroy it using `git reset`>>.
1499
1500 Alternatively, you
1501 can edit the working directory and update the index to fix your
1502 mistake, just as if you were going to <<how-to-make-a-commit,create a
1503 new commit>>, then run
1504
1505 -------------------------------------------------
1506 $ git commit --amend
1507 -------------------------------------------------
1508
1509 which will replace the old commit by a new commit incorporating your
1510 changes, giving you a chance to edit the old commit message first.
1511
1512 Again, you should never do this to a commit that may already have
1513 been merged into another branch; use linkgit:git-revert[1] instead in
1514 that case.
1515
1516 It is also possible to replace commits further back in the history, but
1517 this is an advanced topic to be left for
1518 <<cleaning-up-history,another chapter>>.
1519
1520 [[checkout-of-path]]
1521 Checking out an old version of a file
1522 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1523
1524 In the process of undoing a previous bad change, you may find it
1525 useful to check out an older version of a particular file using
1526 linkgit:git-checkout[1]. We've used `git checkout` before to switch
1527 branches, but it has quite different behavior if it is given a path
1528 name: the command
1529
1530 -------------------------------------------------
1531 $ git checkout HEAD^ path/to/file
1532 -------------------------------------------------
1533
1534 replaces path/to/file by the contents it had in the commit HEAD^, and
1535 also updates the index to match. It does not change branches.
1536
1537 If you just want to look at an old version of the file, without
1538 modifying the working directory, you can do that with
1539 linkgit:git-show[1]:
1540
1541 -------------------------------------------------
1542 $ git show HEAD^:path/to/file
1543 -------------------------------------------------
1544
1545 which will display the given version of the file.
1546
1547 [[interrupted-work]]
1548 Temporarily setting aside work in progress
1549 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1550
1551 While you are in the middle of working on something complicated, you
1552 find an unrelated but obvious and trivial bug. You would like to fix it
1553 before continuing. You can use linkgit:git-stash[1] to save the current
1554 state of your work, and after fixing the bug (or, optionally after doing
1555 so on a different branch and then coming back), unstash the
1556 work-in-progress changes.
1557
1558 ------------------------------------------------
1559 $ git stash save "work in progress for foo feature"
1560 ------------------------------------------------
1561
1562 This command will save your changes away to the `stash`, and
1563 reset your working tree and the index to match the tip of your
1564 current branch. Then you can make your fix as usual.
1565
1566 ------------------------------------------------
1567 ... edit and test ...
1568 $ git commit -a -m "blorpl: typofix"
1569 ------------------------------------------------
1570
1571 After that, you can go back to what you were working on with
1572 `git stash pop`:
1573
1574 ------------------------------------------------
1575 $ git stash pop
1576 ------------------------------------------------
1577
1578
1579 [[ensuring-good-performance]]
1580 Ensuring good performance
1581 -------------------------
1582
1583 On large repositories, Git depends on compression to keep the history
1584 information from taking up too much space on disk or in memory. Some
1585 Git commands may automatically run linkgit:git-gc[1], so you don't
1586 have to worry about running it manually. However, compressing a large
1587 repository may take a while, so you may want to call `gc` explicitly
1588 to avoid automatic compression kicking in when it is not convenient.
1589
1590
1591 [[ensuring-reliability]]
1592 Ensuring reliability
1593 --------------------
1594
1595 [[checking-for-corruption]]
1596 Checking the repository for corruption
1597 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1598
1599 The linkgit:git-fsck[1] command runs a number of self-consistency checks
1600 on the repository, and reports on any problems. This may take some
1601 time.
1602
1603 -------------------------------------------------
1604 $ git fsck
1605 dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1606 dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1607 dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1608 dangling blob 218761f9d90712d37a9c5e36f406f92202db07eb
1609 dangling commit bf093535a34a4d35731aa2bd90fe6b176302f14f
1610 dangling commit 8e4bec7f2ddaa268bef999853c25755452100f8e
1611 dangling tree d50bb86186bf27b681d25af89d3b5b68382e4085
1612 dangling tree b24c2473f1fd3d91352a624795be026d64c8841f
1613 ...
1614 -------------------------------------------------
1615
1616 You will see informational messages on dangling objects. They are objects
1617 that still exist in the repository but are no longer referenced by any of
1618 your branches, and can (and will) be removed after a while with `gc`.
1619 You can run `git fsck --no-dangling` to suppress these messages, and still
1620 view real errors.
1621
1622 [[recovering-lost-changes]]
1623 Recovering lost changes
1624 ~~~~~~~~~~~~~~~~~~~~~~~
1625
1626 [[reflogs]]
1627 Reflogs
1628 ^^^^^^^
1629
1630 Say you modify a branch with <<fixing-mistakes,`git reset --hard`>>,
1631 and then realize that the branch was the only reference you had to
1632 that point in history.
1633
1634 Fortunately, Git also keeps a log, called a "reflog", of all the
1635 previous values of each branch. So in this case you can still find the
1636 old history using, for example,
1637
1638 -------------------------------------------------
1639 $ git log master@{1}
1640 -------------------------------------------------
1641
1642 This lists the commits reachable from the previous version of the
1643 `master` branch head. This syntax can be used with any Git command
1644 that accepts a commit, not just with `git log`. Some other examples:
1645
1646 -------------------------------------------------
1647 $ git show master@{2} # See where the branch pointed 2,
1648 $ git show master@{3} # 3, ... changes ago.
1649 $ gitk master@{yesterday} # See where it pointed yesterday,
1650 $ gitk master@{"1 week ago"} # ... or last week
1651 $ git log --walk-reflogs master # show reflog entries for master
1652 -------------------------------------------------
1653
1654 A separate reflog is kept for the HEAD, so
1655
1656 -------------------------------------------------
1657 $ git show HEAD@{"1 week ago"}
1658 -------------------------------------------------
1659
1660 will show what HEAD pointed to one week ago, not what the current branch
1661 pointed to one week ago. This allows you to see the history of what
1662 you've checked out.
1663
1664 The reflogs are kept by default for 30 days, after which they may be
1665 pruned. See linkgit:git-reflog[1] and linkgit:git-gc[1] to learn
1666 how to control this pruning, and see the "SPECIFYING REVISIONS"
1667 section of linkgit:gitrevisions[7] for details.
1668
1669 Note that the reflog history is very different from normal Git history.
1670 While normal history is shared by every repository that works on the
1671 same project, the reflog history is not shared: it tells you only about
1672 how the branches in your local repository have changed over time.
1673
1674 [[dangling-object-recovery]]
1675 Examining dangling objects
1676 ^^^^^^^^^^^^^^^^^^^^^^^^^^
1677
1678 In some situations the reflog may not be able to save you. For example,
1679 suppose you delete a branch, then realize you need the history it
1680 contained. The reflog is also deleted; however, if you have not yet
1681 pruned the repository, then you may still be able to find the lost
1682 commits in the dangling objects that `git fsck` reports. See
1683 <<dangling-objects>> for the details.
1684
1685 -------------------------------------------------
1686 $ git fsck
1687 dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1688 dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1689 dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1690 ...
1691 -------------------------------------------------
1692
1693 You can examine
1694 one of those dangling commits with, for example,
1695
1696 ------------------------------------------------
1697 $ gitk 7281251ddd --not --all
1698 ------------------------------------------------
1699
1700 which does what it sounds like: it says that you want to see the commit
1701 history that is described by the dangling commit(s), but not the
1702 history that is described by all your existing branches and tags. Thus
1703 you get exactly the history reachable from that commit that is lost.
1704 (And notice that it might not be just one commit: we only report the
1705 "tip of the line" as being dangling, but there might be a whole deep
1706 and complex commit history that was dropped.)
1707
1708 If you decide you want the history back, you can always create a new
1709 reference pointing to it, for example, a new branch:
1710
1711 ------------------------------------------------
1712 $ git branch recovered-branch 7281251ddd
1713 ------------------------------------------------
1714
1715 Other types of dangling objects (blobs and trees) are also possible, and
1716 dangling objects can arise in other situations.
1717
1718
1719 [[sharing-development]]
1720 Sharing development with others
1721 ===============================
1722
1723 [[getting-updates-With-git-pull]]
1724 Getting updates with git pull
1725 -----------------------------
1726
1727 After you clone a repository and commit a few changes of your own, you
1728 may wish to check the original repository for updates and merge them
1729 into your own work.
1730
1731 We have already seen <<Updating-a-repository-With-git-fetch,how to
1732 keep remote-tracking branches up to date>> with linkgit:git-fetch[1],
1733 and how to merge two branches. So you can merge in changes from the
1734 original repository's master branch with:
1735
1736 -------------------------------------------------
1737 $ git fetch
1738 $ git merge origin/master
1739 -------------------------------------------------
1740
1741 However, the linkgit:git-pull[1] command provides a way to do this in
1742 one step:
1743
1744 -------------------------------------------------
1745 $ git pull origin master
1746 -------------------------------------------------
1747
1748 In fact, if you have `master` checked out, then this branch has been
1749 configured by `git clone` to get changes from the HEAD branch of the
1750 origin repository. So often you can
1751 accomplish the above with just a simple
1752
1753 -------------------------------------------------
1754 $ git pull
1755 -------------------------------------------------
1756
1757 This command will fetch changes from the remote branches to your
1758 remote-tracking branches `origin/*`, and merge the default branch into
1759 the current branch.
1760
1761 More generally, a branch that is created from a remote-tracking branch
1762 will pull
1763 by default from that branch. See the descriptions of the
1764 `branch.<name>.remote` and `branch.<name>.merge` options in
1765 linkgit:git-config[1], and the discussion of the `--track` option in
1766 linkgit:git-checkout[1], to learn how to control these defaults.
1767
1768 In addition to saving you keystrokes, `git pull` also helps you by
1769 producing a default commit message documenting the branch and
1770 repository that you pulled from.
1771
1772 (But note that no such commit will be created in the case of a
1773 <<fast-forwards,fast-forward>>; instead, your branch will just be
1774 updated to point to the latest commit from the upstream branch.)
1775
1776 The `git pull` command can also be given `.` as the "remote" repository,
1777 in which case it just merges in a branch from the current repository; so
1778 the commands
1779
1780 -------------------------------------------------
1781 $ git pull . branch
1782 $ git merge branch
1783 -------------------------------------------------
1784
1785 are roughly equivalent.
1786
1787 [[submitting-patches]]
1788 Submitting patches to a project
1789 -------------------------------
1790
1791 If you just have a few changes, the simplest way to submit them may
1792 just be to send them as patches in email:
1793
1794 First, use linkgit:git-format-patch[1]; for example:
1795
1796 -------------------------------------------------
1797 $ git format-patch origin
1798 -------------------------------------------------
1799
1800 will produce a numbered series of files in the current directory, one
1801 for each patch in the current branch but not in `origin/HEAD`.
1802
1803 `git format-patch` can include an initial "cover letter". You can insert
1804 commentary on individual patches after the three dash line which
1805 `format-patch` places after the commit message but before the patch
1806 itself. If you use `git notes` to track your cover letter material,
1807 `git format-patch --notes` will include the commit's notes in a similar
1808 manner.
1809
1810 You can then import these into your mail client and send them by
1811 hand. However, if you have a lot to send at once, you may prefer to
1812 use the linkgit:git-send-email[1] script to automate the process.
1813 Consult the mailing list for your project first to determine
1814 their requirements for submitting patches.
1815
1816 [[importing-patches]]
1817 Importing patches to a project
1818 ------------------------------
1819
1820 Git also provides a tool called linkgit:git-am[1] (am stands for
1821 "apply mailbox"), for importing such an emailed series of patches.
1822 Just save all of the patch-containing messages, in order, into a
1823 single mailbox file, say `patches.mbox`, then run
1824
1825 -------------------------------------------------
1826 $ git am -3 patches.mbox
1827 -------------------------------------------------
1828
1829 Git will apply each patch in order; if any conflicts are found, it
1830 will stop, and you can fix the conflicts as described in
1831 "<<resolving-a-merge,Resolving a merge>>". (The `-3` option tells
1832 Git to perform a merge; if you would prefer it just to abort and
1833 leave your tree and index untouched, you may omit that option.)
1834
1835 Once the index is updated with the results of the conflict
1836 resolution, instead of creating a new commit, just run
1837
1838 -------------------------------------------------
1839 $ git am --continue
1840 -------------------------------------------------
1841
1842 and Git will create the commit for you and continue applying the
1843 remaining patches from the mailbox.
1844
1845 The final result will be a series of commits, one for each patch in
1846 the original mailbox, with authorship and commit log message each
1847 taken from the message containing each patch.
1848
1849 [[public-repositories]]
1850 Public Git repositories
1851 -----------------------
1852
1853 Another way to submit changes to a project is to tell the maintainer
1854 of that project to pull the changes from your repository using
1855 linkgit:git-pull[1]. In the section "<<getting-updates-With-git-pull,
1856 Getting updates with `git pull`>>" we described this as a way to get
1857 updates from the "main" repository, but it works just as well in the
1858 other direction.
1859
1860 If you and the maintainer both have accounts on the same machine, then
1861 you can just pull changes from each other's repositories directly;
1862 commands that accept repository URLs as arguments will also accept a
1863 local directory name:
1864
1865 -------------------------------------------------
1866 $ git clone /path/to/repository
1867 $ git pull /path/to/other/repository
1868 -------------------------------------------------
1869
1870 or an ssh URL:
1871
1872 -------------------------------------------------
1873 $ git clone ssh://yourhost/~you/repository
1874 -------------------------------------------------
1875
1876 For projects with few developers, or for synchronizing a few private
1877 repositories, this may be all you need.
1878
1879 However, the more common way to do this is to maintain a separate public
1880 repository (usually on a different host) for others to pull changes
1881 from. This is usually more convenient, and allows you to cleanly
1882 separate private work in progress from publicly visible work.
1883
1884 You will continue to do your day-to-day work in your personal
1885 repository, but periodically "push" changes from your personal
1886 repository into your public repository, allowing other developers to
1887 pull from that repository. So the flow of changes, in a situation
1888 where there is one other developer with a public repository, looks
1889 like this:
1890
1891 you push
1892 your personal repo ------------------> your public repo
1893 ^ |
1894 | |
1895 | you pull | they pull
1896 | |
1897 | |
1898 | they push V
1899 their public repo <------------------- their repo
1900
1901 We explain how to do this in the following sections.
1902
1903 [[setting-up-a-public-repository]]
1904 Setting up a public repository
1905 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1906
1907 Assume your personal repository is in the directory `~/proj`. We
1908 first create a new clone of the repository and tell `git daemon` that it
1909 is meant to be public:
1910
1911 -------------------------------------------------
1912 $ git clone --bare ~/proj proj.git
1913 $ touch proj.git/git-daemon-export-ok
1914 -------------------------------------------------
1915
1916 The resulting directory proj.git contains a "bare" git repository--it is
1917 just the contents of the `.git` directory, without any files checked out
1918 around it.
1919
1920 Next, copy `proj.git` to the server where you plan to host the
1921 public repository. You can use scp, rsync, or whatever is most
1922 convenient.
1923
1924 [[exporting-via-git]]
1925 Exporting a Git repository via the Git protocol
1926 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1927
1928 This is the preferred method.
1929
1930 If someone else administers the server, they should tell you what
1931 directory to put the repository in, and what `git://` URL it will
1932 appear at. You can then skip to the section
1933 "<<pushing-changes-to-a-public-repository,Pushing changes to a public
1934 repository>>", below.
1935
1936 Otherwise, all you need to do is start linkgit:git-daemon[1]; it will
1937 listen on port 9418. By default, it will allow access to any directory
1938 that looks like a Git directory and contains the magic file
1939 git-daemon-export-ok. Passing some directory paths as `git daemon`
1940 arguments will further restrict the exports to those paths.
1941
1942 You can also run `git daemon` as an inetd service; see the
1943 linkgit:git-daemon[1] man page for details. (See especially the
1944 examples section.)
1945
1946 [[exporting-via-http]]
1947 Exporting a git repository via HTTP
1948 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1949
1950 The Git protocol gives better performance and reliability, but on a
1951 host with a web server set up, HTTP exports may be simpler to set up.
1952
1953 All you need to do is place the newly created bare Git repository in
1954 a directory that is exported by the web server, and make some
1955 adjustments to give web clients some extra information they need:
1956
1957 -------------------------------------------------
1958 $ mv proj.git /home/you/public_html/proj.git
1959 $ cd proj.git
1960 $ git --bare update-server-info
1961 $ mv hooks/post-update.sample hooks/post-update
1962 -------------------------------------------------
1963
1964 (For an explanation of the last two lines, see
1965 linkgit:git-update-server-info[1] and linkgit:githooks[5].)
1966
1967 Advertise the URL of `proj.git`. Anybody else should then be able to
1968 clone or pull from that URL, for example with a command line like:
1969
1970 -------------------------------------------------
1971 $ git clone http://yourserver.com/~you/proj.git
1972 -------------------------------------------------
1973
1974 (See also
1975 link:howto/setup-git-server-over-http.html[setup-git-server-over-http]
1976 for a slightly more sophisticated setup using WebDAV which also
1977 allows pushing over HTTP.)
1978
1979 [[pushing-changes-to-a-public-repository]]
1980 Pushing changes to a public repository
1981 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1982
1983 Note that the two techniques outlined above (exporting via
1984 <<exporting-via-http,http>> or <<exporting-via-git,git>>) allow other
1985 maintainers to fetch your latest changes, but they do not allow write
1986 access, which you will need to update the public repository with the
1987 latest changes created in your private repository.
1988
1989 The simplest way to do this is using linkgit:git-push[1] and ssh; to
1990 update the remote branch named `master` with the latest state of your
1991 branch named `master`, run
1992
1993 -------------------------------------------------
1994 $ git push ssh://yourserver.com/~you/proj.git master:master
1995 -------------------------------------------------
1996
1997 or just
1998
1999 -------------------------------------------------
2000 $ git push ssh://yourserver.com/~you/proj.git master
2001 -------------------------------------------------
2002
2003 As with `git fetch`, `git push` will complain if this does not result in a
2004 <<fast-forwards,fast-forward>>; see the following section for details on
2005 handling this case.
2006
2007 Note that the target of a `push` is normally a
2008 <<def_bare_repository,bare>> repository. You can also push to a
2009 repository that has a checked-out working tree, but a push to update the
2010 currently checked-out branch is denied by default to prevent confusion.
2011 See the description of the receive.denyCurrentBranch option
2012 in linkgit:git-config[1] for details.
2013
2014 As with `git fetch`, you may also set up configuration options to
2015 save typing; so, for example:
2016
2017 -------------------------------------------------
2018 $ git remote add public-repo ssh://yourserver.com/~you/proj.git
2019 -------------------------------------------------
2020
2021 adds the following to `.git/config`:
2022
2023 -------------------------------------------------
2024 [remote "public-repo"]
2025 url = yourserver.com:proj.git
2026 fetch = +refs/heads/*:refs/remotes/example/*
2027 -------------------------------------------------
2028
2029 which lets you do the same push with just
2030
2031 -------------------------------------------------
2032 $ git push public-repo master
2033 -------------------------------------------------
2034
2035 See the explanations of the `remote.<name>.url`,
2036 `branch.<name>.remote`, and `remote.<name>.push` options in
2037 linkgit:git-config[1] for details.
2038
2039 [[forcing-push]]
2040 What to do when a push fails
2041 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2042
2043 If a push would not result in a <<fast-forwards,fast-forward>> of the
2044 remote branch, then it will fail with an error like:
2045
2046 -------------------------------------------------
2047 error: remote 'refs/heads/master' is not an ancestor of
2048 local 'refs/heads/master'.
2049 Maybe you are not up-to-date and need to pull first?
2050 error: failed to push to 'ssh://yourserver.com/~you/proj.git'
2051 -------------------------------------------------
2052
2053 This can happen, for example, if you:
2054
2055 - use `git reset --hard` to remove already-published commits, or
2056 - use `git commit --amend` to replace already-published commits
2057 (as in <<fixing-a-mistake-by-rewriting-history>>), or
2058 - use `git rebase` to rebase any already-published commits (as
2059 in <<using-git-rebase>>).
2060
2061 You may force `git push` to perform the update anyway by preceding the
2062 branch name with a plus sign:
2063
2064 -------------------------------------------------
2065 $ git push ssh://yourserver.com/~you/proj.git +master
2066 -------------------------------------------------
2067
2068 Note the addition of the `+` sign. Alternatively, you can use the
2069 `-f` flag to force the remote update, as in:
2070
2071 -------------------------------------------------
2072 $ git push -f ssh://yourserver.com/~you/proj.git master
2073 -------------------------------------------------
2074
2075 Normally whenever a branch head in a public repository is modified, it
2076 is modified to point to a descendant of the commit that it pointed to
2077 before. By forcing a push in this situation, you break that convention.
2078 (See <<problems-With-rewriting-history>>.)
2079
2080 Nevertheless, this is a common practice for people that need a simple
2081 way to publish a work-in-progress patch series, and it is an acceptable
2082 compromise as long as you warn other developers that this is how you
2083 intend to manage the branch.
2084
2085 It's also possible for a push to fail in this way when other people have
2086 the right to push to the same repository. In that case, the correct
2087 solution is to retry the push after first updating your work: either by a
2088 pull, or by a fetch followed by a rebase; see the
2089 <<setting-up-a-shared-repository,next section>> and
2090 linkgit:gitcvs-migration[7] for more.
2091
2092 [[setting-up-a-shared-repository]]
2093 Setting up a shared repository
2094 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2095
2096 Another way to collaborate is by using a model similar to that
2097 commonly used in CVS, where several developers with special rights
2098 all push to and pull from a single shared repository. See
2099 linkgit:gitcvs-migration[7] for instructions on how to
2100 set this up.
2101
2102 However, while there is nothing wrong with Git's support for shared
2103 repositories, this mode of operation is not generally recommended,
2104 simply because the mode of collaboration that Git supports--by
2105 exchanging patches and pulling from public repositories--has so many
2106 advantages over the central shared repository:
2107
2108 - Git's ability to quickly import and merge patches allows a
2109 single maintainer to process incoming changes even at very
2110 high rates. And when that becomes too much, `git pull` provides
2111 an easy way for that maintainer to delegate this job to other
2112 maintainers while still allowing optional review of incoming
2113 changes.
2114 - Since every developer's repository has the same complete copy
2115 of the project history, no repository is special, and it is
2116 trivial for another developer to take over maintenance of a
2117 project, either by mutual agreement, or because a maintainer
2118 becomes unresponsive or difficult to work with.
2119 - The lack of a central group of "committers" means there is
2120 less need for formal decisions about who is "in" and who is
2121 "out".
2122
2123 [[setting-up-gitweb]]
2124 Allowing web browsing of a repository
2125 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2126
2127 The gitweb cgi script provides users an easy way to browse your
2128 project's files and history without having to install Git; see the file
2129 gitweb/INSTALL in the Git source tree for instructions on setting it up.
2130
2131 [[how-to-get-a-git-repository-with-minimal-history]]
2132 How to get a Git repository with minimal history
2133 ------------------------------------------------
2134
2135 A <<def_shallow_clone,shallow clone>>, with its truncated
2136 history, is useful when one is interested only in recent history
2137 of a project and getting full history from the upstream is
2138 expensive.
2139
2140 A <<def_shallow_clone,shallow clone>> is created by specifying
2141 the linkgit:git-clone[1] `--depth` switch. The depth can later be
2142 changed with the linkgit:git-fetch[1] `--depth` switch, or full
2143 history restored with `--unshallow`.
2144
2145 Merging inside a <<def_shallow_clone,shallow clone>> will work as long
2146 as a merge base is in the recent history.
2147 Otherwise, it will be like merging unrelated histories and may
2148 have to result in huge conflicts. This limitation may make such
2149 a repository unsuitable to be used in merge based workflows.
2150
2151 [[sharing-development-examples]]
2152 Examples
2153 --------
2154
2155 [[maintaining-topic-branches]]
2156 Maintaining topic branches for a Linux subsystem maintainer
2157 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2158
2159 This describes how Tony Luck uses Git in his role as maintainer of the
2160 IA64 architecture for the Linux kernel.
2161
2162 He uses two public branches:
2163
2164 - A "test" tree into which patches are initially placed so that they
2165 can get some exposure when integrated with other ongoing development.
2166 This tree is available to Andrew for pulling into -mm whenever he
2167 wants.
2168
2169 - A "release" tree into which tested patches are moved for final sanity
2170 checking, and as a vehicle to send them upstream to Linus (by sending
2171 him a "please pull" request.)
2172
2173 He also uses a set of temporary branches ("topic branches"), each
2174 containing a logical grouping of patches.
2175
2176 To set this up, first create your work tree by cloning Linus's public
2177 tree:
2178
2179 -------------------------------------------------
2180 $ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git work
2181 $ cd work
2182 -------------------------------------------------
2183
2184 Linus's tree will be stored in the remote-tracking branch named origin/master,
2185 and can be updated using linkgit:git-fetch[1]; you can track other
2186 public trees using linkgit:git-remote[1] to set up a "remote" and
2187 linkgit:git-fetch[1] to keep them up-to-date; see
2188 <<repositories-and-branches>>.
2189
2190 Now create the branches in which you are going to work; these start out
2191 at the current tip of origin/master branch, and should be set up (using
2192 the `--track` option to linkgit:git-branch[1]) to merge changes in from
2193 Linus by default.
2194
2195 -------------------------------------------------
2196 $ git branch --track test origin/master
2197 $ git branch --track release origin/master
2198 -------------------------------------------------
2199
2200 These can be easily kept up to date using linkgit:git-pull[1].
2201
2202 -------------------------------------------------
2203 $ git checkout test && git pull
2204 $ git checkout release && git pull
2205 -------------------------------------------------
2206
2207 Important note! If you have any local changes in these branches, then
2208 this merge will create a commit object in the history (with no local
2209 changes Git will simply do a "fast-forward" merge). Many people dislike
2210 the "noise" that this creates in the Linux history, so you should avoid
2211 doing this capriciously in the `release` branch, as these noisy commits
2212 will become part of the permanent history when you ask Linus to pull
2213 from the release branch.
2214
2215 A few configuration variables (see linkgit:git-config[1]) can
2216 make it easy to push both branches to your public tree. (See
2217 <<setting-up-a-public-repository>>.)
2218
2219 -------------------------------------------------
2220 $ cat >> .git/config <<EOF
2221 [remote "mytree"]
2222 url = master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux.git
2223 push = release
2224 push = test
2225 EOF
2226 -------------------------------------------------
2227
2228 Then you can push both the test and release trees using
2229 linkgit:git-push[1]:
2230
2231 -------------------------------------------------
2232 $ git push mytree
2233 -------------------------------------------------
2234
2235 or push just one of the test and release branches using:
2236
2237 -------------------------------------------------
2238 $ git push mytree test
2239 -------------------------------------------------
2240
2241 or
2242
2243 -------------------------------------------------
2244 $ git push mytree release
2245 -------------------------------------------------
2246
2247 Now to apply some patches from the community. Think of a short
2248 snappy name for a branch to hold this patch (or related group of
2249 patches), and create a new branch from a recent stable tag of
2250 Linus's branch. Picking a stable base for your branch will:
2251 1) help you: by avoiding inclusion of unrelated and perhaps lightly
2252 tested changes
2253 2) help future bug hunters that use `git bisect` to find problems
2254
2255 -------------------------------------------------
2256 $ git checkout -b speed-up-spinlocks v2.6.35
2257 -------------------------------------------------
2258
2259 Now you apply the patch(es), run some tests, and commit the change(s). If
2260 the patch is a multi-part series, then you should apply each as a separate
2261 commit to this branch.
2262
2263 -------------------------------------------------
2264 $ ... patch ... test ... commit [ ... patch ... test ... commit ]*
2265 -------------------------------------------------
2266
2267 When you are happy with the state of this change, you can merge it into the
2268 "test" branch in preparation to make it public:
2269
2270 -------------------------------------------------
2271 $ git checkout test && git merge speed-up-spinlocks
2272 -------------------------------------------------
2273
2274 It is unlikely that you would have any conflicts here ... but you might if you
2275 spent a while on this step and had also pulled new versions from upstream.
2276
2277 Sometime later when enough time has passed and testing done, you can pull the
2278 same branch into the `release` tree ready to go upstream. This is where you
2279 see the value of keeping each patch (or patch series) in its own branch. It
2280 means that the patches can be moved into the `release` tree in any order.
2281
2282 -------------------------------------------------
2283 $ git checkout release && git merge speed-up-spinlocks
2284 -------------------------------------------------
2285
2286 After a while, you will have a number of branches, and despite the
2287 well chosen names you picked for each of them, you may forget what
2288 they are for, or what status they are in. To get a reminder of what
2289 changes are in a specific branch, use:
2290
2291 -------------------------------------------------
2292 $ git log linux..branchname | git shortlog
2293 -------------------------------------------------
2294
2295 To see whether it has already been merged into the test or release branches,
2296 use:
2297
2298 -------------------------------------------------
2299 $ git log test..branchname
2300 -------------------------------------------------
2301
2302 or
2303
2304 -------------------------------------------------
2305 $ git log release..branchname
2306 -------------------------------------------------
2307
2308 (If this branch has not yet been merged, you will see some log entries.
2309 If it has been merged, then there will be no output.)
2310
2311 Once a patch completes the great cycle (moving from test to release,
2312 then pulled by Linus, and finally coming back into your local
2313 `origin/master` branch), the branch for this change is no longer needed.
2314 You detect this when the output from:
2315
2316 -------------------------------------------------
2317 $ git log origin..branchname
2318 -------------------------------------------------
2319
2320 is empty. At this point the branch can be deleted:
2321
2322 -------------------------------------------------
2323 $ git branch -d branchname
2324 -------------------------------------------------
2325
2326 Some changes are so trivial that it is not necessary to create a separate
2327 branch and then merge into each of the test and release branches. For
2328 these changes, just apply directly to the `release` branch, and then
2329 merge that into the `test` branch.
2330
2331 After pushing your work to `mytree`, you can use
2332 linkgit:git-request-pull[1] to prepare a "please pull" request message
2333 to send to Linus:
2334
2335 -------------------------------------------------
2336 $ git push mytree
2337 $ git request-pull origin mytree release
2338 -------------------------------------------------
2339
2340 Here are some of the scripts that simplify all this even further.
2341
2342 -------------------------------------------------
2343 ==== update script ====
2344 # Update a branch in my Git tree. If the branch to be updated
2345 # is origin, then pull from kernel.org. Otherwise merge
2346 # origin/master branch into test|release branch
2347
2348 case "$1" in
2349 test|release)
2350 git checkout $1 && git pull . origin
2351 ;;
2352 origin)
2353 before=$(git rev-parse refs/remotes/origin/master)
2354 git fetch origin
2355 after=$(git rev-parse refs/remotes/origin/master)
2356 if [ $before != $after ]
2357 then
2358 git log $before..$after | git shortlog
2359 fi
2360 ;;
2361 *)
2362 echo "usage: $0 origin|test|release" 1>&2
2363 exit 1
2364 ;;
2365 esac
2366 -------------------------------------------------
2367
2368 -------------------------------------------------
2369 ==== merge script ====
2370 # Merge a branch into either the test or release branch
2371
2372 pname=$0
2373
2374 usage()
2375 {
2376 echo "usage: $pname branch test|release" 1>&2
2377 exit 1
2378 }
2379
2380 git show-ref -q --verify -- refs/heads/"$1" || {
2381 echo "Can't see branch <$1>" 1>&2
2382 usage
2383 }
2384
2385 case "$2" in
2386 test|release)
2387 if [ $(git log $2..$1 | wc -c) -eq 0 ]
2388 then
2389 echo $1 already merged into $2 1>&2
2390 exit 1
2391 fi
2392 git checkout $2 && git pull . $1
2393 ;;
2394 *)
2395 usage
2396 ;;
2397 esac
2398 -------------------------------------------------
2399
2400 -------------------------------------------------
2401 ==== status script ====
2402 # report on status of my ia64 Git tree
2403
2404 gb=$(tput setab 2)
2405 rb=$(tput setab 1)
2406 restore=$(tput setab 9)
2407
2408 if [ `git rev-list test..release | wc -c` -gt 0 ]
2409 then
2410 echo $rb Warning: commits in release that are not in test $restore
2411 git log test..release
2412 fi
2413
2414 for branch in `git show-ref --heads | sed 's|^.*/||'`
2415 do
2416 if [ $branch = test -o $branch = release ]
2417 then
2418 continue
2419 fi
2420
2421 echo -n $gb ======= $branch ====== $restore " "
2422 status=
2423 for ref in test release origin/master
2424 do
2425 if [ `git rev-list $ref..$branch | wc -c` -gt 0 ]
2426 then
2427 status=$status${ref:0:1}
2428 fi
2429 done
2430 case $status in
2431 trl)
2432 echo $rb Need to pull into test $restore
2433 ;;
2434 rl)
2435 echo "In test"
2436 ;;
2437 l)
2438 echo "Waiting for linus"
2439 ;;
2440 "")
2441 echo $rb All done $restore
2442 ;;
2443 *)
2444 echo $rb "<$status>" $restore
2445 ;;
2446 esac
2447 git log origin/master..$branch | git shortlog
2448 done
2449 -------------------------------------------------
2450
2451
2452 [[cleaning-up-history]]
2453 Rewriting history and maintaining patch series
2454 ==============================================
2455
2456 Normally commits are only added to a project, never taken away or
2457 replaced. Git is designed with this assumption, and violating it will
2458 cause Git's merge machinery (for example) to do the wrong thing.
2459
2460 However, there is a situation in which it can be useful to violate this
2461 assumption.
2462
2463 [[patch-series]]
2464 Creating the perfect patch series
2465 ---------------------------------
2466
2467 Suppose you are a contributor to a large project, and you want to add a
2468 complicated feature, and to present it to the other developers in a way
2469 that makes it easy for them to read your changes, verify that they are
2470 correct, and understand why you made each change.
2471
2472 If you present all of your changes as a single patch (or commit), they
2473 may find that it is too much to digest all at once.
2474
2475 If you present them with the entire history of your work, complete with
2476 mistakes, corrections, and dead ends, they may be overwhelmed.
2477
2478 So the ideal is usually to produce a series of patches such that:
2479
2480 1. Each patch can be applied in order.
2481
2482 2. Each patch includes a single logical change, together with a
2483 message explaining the change.
2484
2485 3. No patch introduces a regression: after applying any initial
2486 part of the series, the resulting project still compiles and
2487 works, and has no bugs that it didn't have before.
2488
2489 4. The complete series produces the same end result as your own
2490 (probably much messier!) development process did.
2491
2492 We will introduce some tools that can help you do this, explain how to
2493 use them, and then explain some of the problems that can arise because
2494 you are rewriting history.
2495
2496 [[using-git-rebase]]
2497 Keeping a patch series up to date using git rebase
2498 --------------------------------------------------
2499
2500 Suppose that you create a branch `mywork` on a remote-tracking branch
2501 `origin`, and create some commits on top of it:
2502
2503 -------------------------------------------------
2504 $ git checkout -b mywork origin
2505 $ vi file.txt
2506 $ git commit
2507 $ vi otherfile.txt
2508 $ git commit
2509 ...
2510 -------------------------------------------------
2511
2512 You have performed no merges into mywork, so it is just a simple linear
2513 sequence of patches on top of `origin`:
2514
2515 ................................................
2516 o--o--O <-- origin
2517 \
2518 a--b--c <-- mywork
2519 ................................................
2520
2521 Some more interesting work has been done in the upstream project, and
2522 `origin` has advanced:
2523
2524 ................................................
2525 o--o--O--o--o--o <-- origin
2526 \
2527 a--b--c <-- mywork
2528 ................................................
2529
2530 At this point, you could use `pull` to merge your changes back in;
2531 the result would create a new merge commit, like this:
2532
2533 ................................................
2534 o--o--O--o--o--o <-- origin
2535 \ \
2536 a--b--c--m <-- mywork
2537 ................................................
2538
2539 However, if you prefer to keep the history in mywork a simple series of
2540 commits without any merges, you may instead choose to use
2541 linkgit:git-rebase[1]:
2542
2543 -------------------------------------------------
2544 $ git checkout mywork
2545 $ git rebase origin
2546 -------------------------------------------------
2547
2548 This will remove each of your commits from mywork, temporarily saving
2549 them as patches (in a directory named `.git/rebase-apply`), update mywork to
2550 point at the latest version of origin, then apply each of the saved
2551 patches to the new mywork. The result will look like:
2552
2553
2554 ................................................
2555 o--o--O--o--o--o <-- origin
2556 \
2557 a'--b'--c' <-- mywork
2558 ................................................
2559
2560 In the process, it may discover conflicts. In that case it will stop
2561 and allow you to fix the conflicts; after fixing conflicts, use `git add`
2562 to update the index with those contents, and then, instead of
2563 running `git commit`, just run
2564
2565 -------------------------------------------------
2566 $ git rebase --continue
2567 -------------------------------------------------
2568
2569 and Git will continue applying the rest of the patches.
2570
2571 At any point you may use the `--abort` option to abort this process and
2572 return mywork to the state it had before you started the rebase:
2573
2574 -------------------------------------------------
2575 $ git rebase --abort
2576 -------------------------------------------------
2577
2578 If you need to reorder or edit a number of commits in a branch, it may
2579 be easier to use `git rebase -i`, which allows you to reorder and
2580 squash commits, as well as marking them for individual editing during
2581 the rebase. See <<interactive-rebase>> for details, and
2582 <<reordering-patch-series>> for alternatives.
2583
2584 [[rewriting-one-commit]]
2585 Rewriting a single commit
2586 -------------------------
2587
2588 We saw in <<fixing-a-mistake-by-rewriting-history>> that you can replace the
2589 most recent commit using
2590
2591 -------------------------------------------------
2592 $ git commit --amend
2593 -------------------------------------------------
2594
2595 which will replace the old commit by a new commit incorporating your
2596 changes, giving you a chance to edit the old commit message first.
2597 This is useful for fixing typos in your last commit, or for adjusting
2598 the patch contents of a poorly staged commit.
2599
2600 If you need to amend commits from deeper in your history, you can
2601 use <<interactive-rebase,interactive rebase's `edit` instruction>>.
2602
2603 [[reordering-patch-series]]
2604 Reordering or selecting from a patch series
2605 -------------------------------------------
2606
2607 Sometimes you want to edit a commit deeper in your history. One
2608 approach is to use `git format-patch` to create a series of patches
2609 and then reset the state to before the patches:
2610
2611 -------------------------------------------------
2612 $ git format-patch origin
2613 $ git reset --hard origin
2614 -------------------------------------------------
2615
2616 Then modify, reorder, or eliminate patches as needed before applying
2617 them again with linkgit:git-am[1]:
2618
2619 -------------------------------------------------
2620 $ git am *.patch
2621 -------------------------------------------------
2622
2623 [[interactive-rebase]]
2624 Using interactive rebases
2625 -------------------------
2626
2627 You can also edit a patch series with an interactive rebase. This is
2628 the same as <<reordering-patch-series,reordering a patch series using
2629 `format-patch`>>, so use whichever interface you like best.
2630
2631 Rebase your current HEAD on the last commit you want to retain as-is.
2632 For example, if you want to reorder the last 5 commits, use:
2633
2634 -------------------------------------------------
2635 $ git rebase -i HEAD~5
2636 -------------------------------------------------
2637
2638 This will open your editor with a list of steps to be taken to perform
2639 your rebase.
2640
2641 -------------------------------------------------
2642 pick deadbee The oneline of this commit
2643 pick fa1afe1 The oneline of the next commit
2644 ...
2645
2646 # Rebase c0ffeee..deadbee onto c0ffeee
2647 #
2648 # Commands:
2649 # p, pick = use commit
2650 # r, reword = use commit, but edit the commit message
2651 # e, edit = use commit, but stop for amending
2652 # s, squash = use commit, but meld into previous commit
2653 # f, fixup = like "squash", but discard this commit's log message
2654 # x, exec = run command (the rest of the line) using shell
2655 #
2656 # These lines can be re-ordered; they are executed from top to bottom.
2657 #
2658 # If you remove a line here THAT COMMIT WILL BE LOST.
2659 #
2660 # However, if you remove everything, the rebase will be aborted.
2661 #
2662 # Note that empty commits are commented out
2663 -------------------------------------------------
2664
2665 As explained in the comments, you can reorder commits, squash them
2666 together, edit commit messages, etc. by editing the list. Once you
2667 are satisfied, save the list and close your editor, and the rebase
2668 will begin.
2669
2670 The rebase will stop where `pick` has been replaced with `edit` or
2671 when a step in the list fails to mechanically resolve conflicts and
2672 needs your help. When you are done editing and/or resolving conflicts
2673 you can continue with `git rebase --continue`. If you decide that
2674 things are getting too hairy, you can always bail out with `git rebase
2675 --abort`. Even after the rebase is complete, you can still recover
2676 the original branch by using the <<reflogs,reflog>>.
2677
2678 For a more detailed discussion of the procedure and additional tips,
2679 see the "INTERACTIVE MODE" section of linkgit:git-rebase[1].
2680
2681 [[patch-series-tools]]
2682 Other tools
2683 -----------
2684
2685 There are numerous other tools, such as StGit, which exist for the
2686 purpose of maintaining a patch series. These are outside of the scope of
2687 this manual.
2688
2689 [[problems-With-rewriting-history]]
2690 Problems with rewriting history
2691 -------------------------------
2692
2693 The primary problem with rewriting the history of a branch has to do
2694 with merging. Suppose somebody fetches your branch and merges it into
2695 their branch, with a result something like this:
2696
2697 ................................................
2698 o--o--O--o--o--o <-- origin
2699 \ \
2700 t--t--t--m <-- their branch:
2701 ................................................
2702
2703 Then suppose you modify the last three commits:
2704
2705 ................................................
2706 o--o--o <-- new head of origin
2707 /
2708 o--o--O--o--o--o <-- old head of origin
2709 ................................................
2710
2711 If we examined all this history together in one repository, it will
2712 look like:
2713
2714 ................................................
2715 o--o--o <-- new head of origin
2716 /
2717 o--o--O--o--o--o <-- old head of origin
2718 \ \
2719 t--t--t--m <-- their branch:
2720 ................................................
2721
2722 Git has no way of knowing that the new head is an updated version of
2723 the old head; it treats this situation exactly the same as it would if
2724 two developers had independently done the work on the old and new heads
2725 in parallel. At this point, if someone attempts to merge the new head
2726 in to their branch, Git will attempt to merge together the two (old and
2727 new) lines of development, instead of trying to replace the old by the
2728 new. The results are likely to be unexpected.
2729
2730 You may still choose to publish branches whose history is rewritten,
2731 and it may be useful for others to be able to fetch those branches in
2732 order to examine or test them, but they should not attempt to pull such
2733 branches into their own work.
2734
2735 For true distributed development that supports proper merging,
2736 published branches should never be rewritten.
2737
2738 [[bisect-merges]]
2739 Why bisecting merge commits can be harder than bisecting linear history
2740 -----------------------------------------------------------------------
2741
2742 The linkgit:git-bisect[1] command correctly handles history that
2743 includes merge commits. However, when the commit that it finds is a
2744 merge commit, the user may need to work harder than usual to figure out
2745 why that commit introduced a problem.
2746
2747 Imagine this history:
2748
2749 ................................................
2750 ---Z---o---X---...---o---A---C---D
2751 \ /
2752 o---o---Y---...---o---B
2753 ................................................
2754
2755 Suppose that on the upper line of development, the meaning of one
2756 of the functions that exists at Z is changed at commit X. The
2757 commits from Z leading to A change both the function's
2758 implementation and all calling sites that exist at Z, as well
2759 as new calling sites they add, to be consistent. There is no
2760 bug at A.
2761
2762 Suppose that in the meantime on the lower line of development somebody
2763 adds a new calling site for that function at commit Y. The
2764 commits from Z leading to B all assume the old semantics of that
2765 function and the callers and the callee are consistent with each
2766 other. There is no bug at B, either.
2767
2768 Suppose further that the two development lines merge cleanly at C,
2769 so no conflict resolution is required.
2770
2771 Nevertheless, the code at C is broken, because the callers added
2772 on the lower line of development have not been converted to the new
2773 semantics introduced on the upper line of development. So if all
2774 you know is that D is bad, that Z is good, and that
2775 linkgit:git-bisect[1] identifies C as the culprit, how will you
2776 figure out that the problem is due to this change in semantics?
2777
2778 When the result of a `git bisect` is a non-merge commit, you should
2779 normally be able to discover the problem by examining just that commit.
2780 Developers can make this easy by breaking their changes into small
2781 self-contained commits. That won't help in the case above, however,
2782 because the problem isn't obvious from examination of any single
2783 commit; instead, a global view of the development is required. To
2784 make matters worse, the change in semantics in the problematic
2785 function may be just one small part of the changes in the upper
2786 line of development.
2787
2788 On the other hand, if instead of merging at C you had rebased the
2789 history between Z to B on top of A, you would have gotten this
2790 linear history:
2791
2792 ................................................................
2793 ---Z---o---X--...---o---A---o---o---Y*--...---o---B*--D*
2794 ................................................................
2795
2796 Bisecting between Z and D* would hit a single culprit commit Y*,
2797 and understanding why Y* was broken would probably be easier.
2798
2799 Partly for this reason, many experienced Git users, even when
2800 working on an otherwise merge-heavy project, keep the history
2801 linear by rebasing against the latest upstream version before
2802 publishing.
2803
2804 [[advanced-branch-management]]
2805 Advanced branch management
2806 ==========================
2807
2808 [[fetching-individual-branches]]
2809 Fetching individual branches
2810 ----------------------------
2811
2812 Instead of using linkgit:git-remote[1], you can also choose just
2813 to update one branch at a time, and to store it locally under an
2814 arbitrary name:
2815
2816 -------------------------------------------------
2817 $ git fetch origin todo:my-todo-work
2818 -------------------------------------------------
2819
2820 The first argument, `origin`, just tells Git to fetch from the
2821 repository you originally cloned from. The second argument tells Git
2822 to fetch the branch named `todo` from the remote repository, and to
2823 store it locally under the name `refs/heads/my-todo-work`.
2824
2825 You can also fetch branches from other repositories; so
2826
2827 -------------------------------------------------
2828 $ git fetch git://example.com/proj.git master:example-master
2829 -------------------------------------------------
2830
2831 will create a new branch named `example-master` and store in it the
2832 branch named `master` from the repository at the given URL. If you
2833 already have a branch named example-master, it will attempt to
2834 <<fast-forwards,fast-forward>> to the commit given by example.com's
2835 master branch. In more detail:
2836
2837 [[fetch-fast-forwards]]
2838 git fetch and fast-forwards
2839 ---------------------------
2840
2841 In the previous example, when updating an existing branch, `git fetch`
2842 checks to make sure that the most recent commit on the remote
2843 branch is a descendant of the most recent commit on your copy of the
2844 branch before updating your copy of the branch to point at the new
2845 commit. Git calls this process a <<fast-forwards,fast-forward>>.
2846
2847 A fast-forward looks something like this:
2848
2849 ................................................
2850 o--o--o--o <-- old head of the branch
2851 \
2852 o--o--o <-- new head of the branch
2853 ................................................
2854
2855
2856 In some cases it is possible that the new head will *not* actually be
2857 a descendant of the old head. For example, the developer may have
2858 realized she made a serious mistake, and decided to backtrack,
2859 resulting in a situation like:
2860
2861 ................................................
2862 o--o--o--o--a--b <-- old head of the branch
2863 \
2864 o--o--o <-- new head of the branch
2865 ................................................
2866
2867 In this case, `git fetch` will fail, and print out a warning.
2868
2869 In that case, you can still force Git to update to the new head, as
2870 described in the following section. However, note that in the
2871 situation above this may mean losing the commits labeled `a` and `b`,
2872 unless you've already created a reference of your own pointing to
2873 them.
2874
2875 [[forcing-fetch]]
2876 Forcing git fetch to do non-fast-forward updates
2877 ------------------------------------------------
2878
2879 If git fetch fails because the new head of a branch is not a
2880 descendant of the old head, you may force the update with:
2881
2882 -------------------------------------------------
2883 $ git fetch git://example.com/proj.git +master:refs/remotes/example/master
2884 -------------------------------------------------
2885
2886 Note the addition of the `+` sign. Alternatively, you can use the `-f`
2887 flag to force updates of all the fetched branches, as in:
2888
2889 -------------------------------------------------
2890 $ git fetch -f origin
2891 -------------------------------------------------
2892
2893 Be aware that commits that the old version of example/master pointed at
2894 may be lost, as we saw in the previous section.
2895
2896 [[remote-branch-configuration]]
2897 Configuring remote-tracking branches
2898 ------------------------------------
2899
2900 We saw above that `origin` is just a shortcut to refer to the
2901 repository that you originally cloned from. This information is
2902 stored in Git configuration variables, which you can see using
2903 linkgit:git-config[1]:
2904
2905 -------------------------------------------------
2906 $ git config -l
2907 core.repositoryformatversion=0
2908 core.filemode=true
2909 core.logallrefupdates=true
2910 remote.origin.url=git://git.kernel.org/pub/scm/git/git.git
2911 remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*
2912 branch.master.remote=origin
2913 branch.master.merge=refs/heads/master
2914 -------------------------------------------------
2915
2916 If there are other repositories that you also use frequently, you can
2917 create similar configuration options to save typing; for example,
2918
2919 -------------------------------------------------
2920 $ git remote add example git://example.com/proj.git
2921 -------------------------------------------------
2922
2923 adds the following to `.git/config`:
2924
2925 -------------------------------------------------
2926 [remote "example"]
2927 url = git://example.com/proj.git
2928 fetch = +refs/heads/*:refs/remotes/example/*
2929 -------------------------------------------------
2930
2931 Also note that the above configuration can be performed by directly
2932 editing the file `.git/config` instead of using linkgit:git-remote[1].
2933
2934 After configuring the remote, the following three commands will do the
2935 same thing:
2936
2937 -------------------------------------------------
2938 $ git fetch git://example.com/proj.git +refs/heads/*:refs/remotes/example/*
2939 $ git fetch example +refs/heads/*:refs/remotes/example/*
2940 $ git fetch example
2941 -------------------------------------------------
2942
2943 See linkgit:git-config[1] for more details on the configuration
2944 options mentioned above and linkgit:git-fetch[1] for more details on
2945 the refspec syntax.
2946
2947
2948 [[git-concepts]]
2949 Git concepts
2950 ============
2951
2952 Git is built on a small number of simple but powerful ideas. While it
2953 is possible to get things done without understanding them, you will find
2954 Git much more intuitive if you do.
2955
2956 We start with the most important, the <<def_object_database,object
2957 database>> and the <<def_index,index>>.
2958
2959 [[the-object-database]]
2960 The Object Database
2961 -------------------
2962
2963
2964 We already saw in <<understanding-commits>> that all commits are stored
2965 under a 40-digit "object name". In fact, all the information needed to
2966 represent the history of a project is stored in objects with such names.
2967 In each case the name is calculated by taking the SHA-1 hash of the
2968 contents of the object. The SHA-1 hash is a cryptographic hash function.
2969 What that means to us is that it is impossible to find two different
2970 objects with the same name. This has a number of advantages; among
2971 others:
2972
2973 - Git can quickly determine whether two objects are identical or not,
2974 just by comparing names.
2975 - Since object names are computed the same way in every repository, the
2976 same content stored in two repositories will always be stored under
2977 the same name.
2978 - Git can detect errors when it reads an object, by checking that the
2979 object's name is still the SHA-1 hash of its contents.
2980
2981 (See <<object-details>> for the details of the object formatting and
2982 SHA-1 calculation.)
2983
2984 There are four different types of objects: "blob", "tree", "commit", and
2985 "tag".
2986
2987 - A <<def_blob_object,"blob" object>> is used to store file data.
2988 - A <<def_tree_object,"tree" object>> ties one or more
2989 "blob" objects into a directory structure. In addition, a tree object
2990 can refer to other tree objects, thus creating a directory hierarchy.
2991 - A <<def_commit_object,"commit" object>> ties such directory hierarchies
2992 together into a <<def_DAG,directed acyclic graph>> of revisions--each
2993 commit contains the object name of exactly one tree designating the
2994 directory hierarchy at the time of the commit. In addition, a commit
2995 refers to "parent" commit objects that describe the history of how we
2996 arrived at that directory hierarchy.
2997 - A <<def_tag_object,"tag" object>> symbolically identifies and can be
2998 used to sign other objects. It contains the object name and type of
2999 another object, a symbolic name (of course!) and, optionally, a
3000 signature.
3001
3002 The object types in some more detail:
3003
3004 [[commit-object]]
3005 Commit Object
3006 ~~~~~~~~~~~~~
3007
3008 The "commit" object links a physical state of a tree with a description
3009 of how we got there and why. Use the `--pretty=raw` option to
3010 linkgit:git-show[1] or linkgit:git-log[1] to examine your favorite
3011 commit:
3012
3013 ------------------------------------------------
3014 $ git show -s --pretty=raw 2be7fcb476
3015 commit 2be7fcb4764f2dbcee52635b91fedb1b3dcf7ab4
3016 tree fb3a8bdd0ceddd019615af4d57a53f43d8cee2bf
3017 parent 257a84d9d02e90447b149af58b271c19405edb6a
3018 author Dave Watson <dwatson@mimvista.com> 1187576872 -0400
3019 committer Junio C Hamano <gitster@pobox.com> 1187591163 -0700
3020
3021 Fix misspelling of 'suppress' in docs
3022
3023 Signed-off-by: Junio C Hamano <gitster@pobox.com>
3024 ------------------------------------------------
3025
3026 As you can see, a commit is defined by:
3027
3028 - a tree: The SHA-1 name of a tree object (as defined below), representing
3029 the contents of a directory at a certain point in time.
3030 - parent(s): The SHA-1 name(s) of some number of commits which represent the
3031 immediately previous step(s) in the history of the project. The
3032 example above has one parent; merge commits may have more than
3033 one. A commit with no parents is called a "root" commit, and
3034 represents the initial revision of a project. Each project must have
3035 at least one root. A project can also have multiple roots, though
3036 that isn't common (or necessarily a good idea).
3037 - an author: The name of the person responsible for this change, together
3038 with its date.
3039 - a committer: The name of the person who actually created the commit,
3040 with the date it was done. This may be different from the author, for
3041 example, if the author was someone who wrote a patch and emailed it
3042 to the person who used it to create the commit.
3043 - a comment describing this commit.
3044
3045 Note that a commit does not itself contain any information about what
3046 actually changed; all changes are calculated by comparing the contents
3047 of the tree referred to by this commit with the trees associated with
3048 its parents. In particular, Git does not attempt to record file renames
3049 explicitly, though it can identify cases where the existence of the same
3050 file data at changing paths suggests a rename. (See, for example, the
3051 `-M` option to linkgit:git-diff[1]).
3052
3053 A commit is usually created by linkgit:git-commit[1], which creates a
3054 commit whose parent is normally the current HEAD, and whose tree is
3055 taken from the content currently stored in the index.
3056
3057 [[tree-object]]
3058 Tree Object
3059 ~~~~~~~~~~~
3060
3061 The ever-versatile linkgit:git-show[1] command can also be used to
3062 examine tree objects, but linkgit:git-ls-tree[1] will give you more
3063 details:
3064
3065 ------------------------------------------------
3066 $ git ls-tree fb3a8bdd0ce
3067 100644 blob 63c918c667fa005ff12ad89437f2fdc80926e21c .gitignore
3068 100644 blob 5529b198e8d14decbe4ad99db3f7fb632de0439d .mailmap
3069 100644 blob 6ff87c4664981e4397625791c8ea3bbb5f2279a3 COPYING
3070 040000 tree 2fb783e477100ce076f6bf57e4a6f026013dc745 Documentation
3071 100755 blob 3c0032cec592a765692234f1cba47dfdcc3a9200 GIT-VERSION-GEN
3072 100644 blob 289b046a443c0647624607d471289b2c7dcd470b INSTALL
3073 100644 blob 4eb463797adc693dc168b926b6932ff53f17d0b1 Makefile
3074 100644 blob 548142c327a6790ff8821d67c2ee1eff7a656b52 README
3075 ...
3076 ------------------------------------------------
3077
3078 As you can see, a tree object contains a list of entries, each with a
3079 mode, object type, SHA-1 name, and name, sorted by name. It represents
3080 the contents of a single directory tree.
3081
3082 The object type may be a blob, representing the contents of a file, or
3083 another tree, representing the contents of a subdirectory. Since trees
3084 and blobs, like all other objects, are named by the SHA-1 hash of their
3085 contents, two trees have the same SHA-1 name if and only if their
3086 contents (including, recursively, the contents of all subdirectories)
3087 are identical. This allows Git to quickly determine the differences
3088 between two related tree objects, since it can ignore any entries with
3089 identical object names.
3090
3091 (Note: in the presence of submodules, trees may also have commits as
3092 entries. See <<submodules>> for documentation.)
3093
3094 Note that the files all have mode 644 or 755: Git actually only pays
3095 attention to the executable bit.
3096
3097 [[blob-object]]
3098 Blob Object
3099 ~~~~~~~~~~~
3100
3101 You can use linkgit:git-show[1] to examine the contents of a blob; take,
3102 for example, the blob in the entry for `COPYING` from the tree above:
3103
3104 ------------------------------------------------
3105 $ git show 6ff87c4664
3106
3107 Note that the only valid version of the GPL as far as this project
3108 is concerned is _this_ particular version of the license (ie v2, not
3109 v2.2 or v3.x or whatever), unless explicitly otherwise stated.
3110 ...
3111 ------------------------------------------------
3112
3113 A "blob" object is nothing but a binary blob of data. It doesn't refer
3114 to anything else or have attributes of any kind.
3115
3116 Since the blob is entirely defined by its data, if two files in a
3117 directory tree (or in multiple different versions of the repository)
3118 have the same contents, they will share the same blob object. The object
3119 is totally independent of its location in the directory tree, and
3120 renaming a file does not change the object that file is associated with.
3121
3122 Note that any tree or blob object can be examined using
3123 linkgit:git-show[1] with the <revision>:<path> syntax. This can
3124 sometimes be useful for browsing the contents of a tree that is not
3125 currently checked out.
3126
3127 [[trust]]
3128 Trust
3129 ~~~~~
3130
3131 If you receive the SHA-1 name of a blob from one source, and its contents
3132 from another (possibly untrusted) source, you can still trust that those
3133 contents are correct as long as the SHA-1 name agrees. This is because
3134 the SHA-1 is designed so that it is infeasible to find different contents
3135 that produce the same hash.
3136
3137 Similarly, you need only trust the SHA-1 name of a top-level tree object
3138 to trust the contents of the entire directory that it refers to, and if
3139 you receive the SHA-1 name of a commit from a trusted source, then you
3140 can easily verify the entire history of commits reachable through
3141 parents of that commit, and all of those contents of the trees referred
3142 to by those commits.
3143
3144 So to introduce some real trust in the system, the only thing you need
3145 to do is to digitally sign just 'one' special note, which includes the
3146 name of a top-level commit. Your digital signature shows others
3147 that you trust that commit, and the immutability of the history of
3148 commits tells others that they can trust the whole history.
3149
3150 In other words, you can easily validate a whole archive by just
3151 sending out a single email that tells the people the name (SHA-1 hash)
3152 of the top commit, and digitally sign that email using something
3153 like GPG/PGP.
3154
3155 To assist in this, Git also provides the tag object...
3156
3157 [[tag-object]]
3158 Tag Object
3159 ~~~~~~~~~~
3160
3161 A tag object contains an object, object type, tag name, the name of the
3162 person ("tagger") who created the tag, and a message, which may contain
3163 a signature, as can be seen using linkgit:git-cat-file[1]:
3164
3165 ------------------------------------------------
3166 $ git cat-file tag v1.5.0
3167 object 437b1b20df4b356c9342dac8d38849f24ef44f27
3168 type commit
3169 tag v1.5.0
3170 tagger Junio C Hamano <junkio@cox.net> 1171411200 +0000
3171
3172 GIT 1.5.0
3173 -----BEGIN PGP SIGNATURE-----
3174 Version: GnuPG v1.4.6 (GNU/Linux)
3175
3176 iD8DBQBF0lGqwMbZpPMRm5oRAuRiAJ9ohBLd7s2kqjkKlq1qqC57SbnmzQCdG4ui
3177 nLE/L9aUXdWeTFPron96DLA=
3178 =2E+0
3179 -----END PGP SIGNATURE-----
3180 ------------------------------------------------
3181
3182 See the linkgit:git-tag[1] command to learn how to create and verify tag
3183 objects. (Note that linkgit:git-tag[1] can also be used to create
3184 "lightweight tags", which are not tag objects at all, but just simple
3185 references whose names begin with `refs/tags/`).
3186
3187 [[pack-files]]
3188 How Git stores objects efficiently: pack files
3189 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3190
3191 Newly created objects are initially created in a file named after the
3192 object's SHA-1 hash (stored in `.git/objects`).
3193
3194 Unfortunately this system becomes inefficient once a project has a
3195 lot of objects. Try this on an old project:
3196
3197 ------------------------------------------------
3198 $ git count-objects
3199 6930 objects, 47620 kilobytes
3200 ------------------------------------------------
3201
3202 The first number is the number of objects which are kept in
3203 individual files. The second is the amount of space taken up by
3204 those "loose" objects.
3205
3206 You can save space and make Git faster by moving these loose objects in
3207 to a "pack file", which stores a group of objects in an efficient
3208 compressed format; the details of how pack files are formatted can be
3209 found in link:technical/pack-format.html[pack format].
3210
3211 To put the loose objects into a pack, just run git repack:
3212
3213 ------------------------------------------------
3214 $ git repack
3215 Counting objects: 6020, done.
3216 Delta compression using up to 4 threads.
3217 Compressing objects: 100% (6020/6020), done.
3218 Writing objects: 100% (6020/6020), done.
3219 Total 6020 (delta 4070), reused 0 (delta 0)
3220 ------------------------------------------------
3221
3222 This creates a single "pack file" in .git/objects/pack/
3223 containing all currently unpacked objects. You can then run
3224
3225 ------------------------------------------------
3226 $ git prune
3227 ------------------------------------------------
3228
3229 to remove any of the "loose" objects that are now contained in the
3230 pack. This will also remove any unreferenced objects (which may be
3231 created when, for example, you use `git reset` to remove a commit).
3232 You can verify that the loose objects are gone by looking at the
3233 `.git/objects` directory or by running
3234
3235 ------------------------------------------------
3236 $ git count-objects
3237 0 objects, 0 kilobytes
3238 ------------------------------------------------
3239
3240 Although the object files are gone, any commands that refer to those
3241 objects will work exactly as they did before.
3242
3243 The linkgit:git-gc[1] command performs packing, pruning, and more for
3244 you, so is normally the only high-level command you need.
3245
3246 [[dangling-objects]]
3247 Dangling objects
3248 ~~~~~~~~~~~~~~~~
3249
3250 The linkgit:git-fsck[1] command will sometimes complain about dangling
3251 objects. They are not a problem.
3252
3253 The most common cause of dangling objects is that you've rebased a
3254 branch, or you have pulled from somebody else who rebased a branch--see
3255 <<cleaning-up-history>>. In that case, the old head of the original
3256 branch still exists, as does everything it pointed to. The branch
3257 pointer itself just doesn't, since you replaced it with another one.
3258
3259 There are also other situations that cause dangling objects. For
3260 example, a "dangling blob" may arise because you did a `git add` of a
3261 file, but then, before you actually committed it and made it part of the
3262 bigger picture, you changed something else in that file and committed
3263 that *updated* thing--the old state that you added originally ends up
3264 not being pointed to by any commit or tree, so it's now a dangling blob
3265 object.
3266
3267 Similarly, when the "recursive" merge strategy runs, and finds that
3268 there are criss-cross merges and thus more than one merge base (which is
3269 fairly unusual, but it does happen), it will generate one temporary
3270 midway tree (or possibly even more, if you had lots of criss-crossing
3271 merges and more than two merge bases) as a temporary internal merge
3272 base, and again, those are real objects, but the end result will not end
3273 up pointing to them, so they end up "dangling" in your repository.
3274
3275 Generally, dangling objects aren't anything to worry about. They can
3276 even be very useful: if you screw something up, the dangling objects can
3277 be how you recover your old tree (say, you did a rebase, and realized
3278 that you really didn't want to--you can look at what dangling objects
3279 you have, and decide to reset your head to some old dangling state).
3280
3281 For commits, you can just use:
3282
3283 ------------------------------------------------
3284 $ gitk <dangling-commit-sha-goes-here> --not --all
3285 ------------------------------------------------
3286
3287 This asks for all the history reachable from the given commit but not
3288 from any branch, tag, or other reference. If you decide it's something
3289 you want, you can always create a new reference to it, e.g.,
3290
3291 ------------------------------------------------
3292 $ git branch recovered-branch <dangling-commit-sha-goes-here>
3293 ------------------------------------------------
3294
3295 For blobs and trees, you can't do the same, but you can still examine
3296 them. You can just do
3297
3298 ------------------------------------------------
3299 $ git show <dangling-blob/tree-sha-goes-here>
3300 ------------------------------------------------
3301
3302 to show what the contents of the blob were (or, for a tree, basically
3303 what the `ls` for that directory was), and that may give you some idea
3304 of what the operation was that left that dangling object.
3305
3306 Usually, dangling blobs and trees aren't very interesting. They're
3307 almost always the result of either being a half-way mergebase (the blob
3308 will often even have the conflict markers from a merge in it, if you
3309 have had conflicting merges that you fixed up by hand), or simply
3310 because you interrupted a `git fetch` with ^C or something like that,
3311 leaving _some_ of the new objects in the object database, but just
3312 dangling and useless.
3313
3314 Anyway, once you are sure that you're not interested in any dangling
3315 state, you can just prune all unreachable objects:
3316
3317 ------------------------------------------------
3318 $ git prune
3319 ------------------------------------------------
3320
3321 and they'll be gone. (You should only run `git prune` on a quiescent
3322 repository--it's kind of like doing a filesystem fsck recovery: you
3323 don't want to do that while the filesystem is mounted.
3324 `git prune` is designed not to cause any harm in such cases of concurrent
3325 accesses to a repository but you might receive confusing or scary messages.)
3326
3327 [[recovering-from-repository-corruption]]
3328 Recovering from repository corruption
3329 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3330
3331 By design, Git treats data trusted to it with caution. However, even in
3332 the absence of bugs in Git itself, it is still possible that hardware or
3333 operating system errors could corrupt data.
3334
3335 The first defense against such problems is backups. You can back up a
3336 Git directory using clone, or just using cp, tar, or any other backup
3337 mechanism.
3338
3339 As a last resort, you can search for the corrupted objects and attempt
3340 to replace them by hand. Back up your repository before attempting this
3341 in case you corrupt things even more in the process.
3342
3343 We'll assume that the problem is a single missing or corrupted blob,
3344 which is sometimes a solvable problem. (Recovering missing trees and
3345 especially commits is *much* harder).
3346
3347 Before starting, verify that there is corruption, and figure out where
3348 it is with linkgit:git-fsck[1]; this may be time-consuming.
3349
3350 Assume the output looks like this:
3351
3352 ------------------------------------------------
3353 $ git fsck --full --no-dangling
3354 broken link from tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff8
3355 to blob 4b9458b3786228369c63936db65827de3cc06200
3356 missing blob 4b9458b3786228369c63936db65827de3cc06200
3357 ------------------------------------------------
3358
3359 Now you know that blob 4b9458b3 is missing, and that the tree 2d9263c6
3360 points to it. If you could find just one copy of that missing blob
3361 object, possibly in some other repository, you could move it into
3362 `.git/objects/4b/9458b3...` and be done. Suppose you can't. You can
3363 still examine the tree that pointed to it with linkgit:git-ls-tree[1],
3364 which might output something like:
3365
3366 ------------------------------------------------
3367 $ git ls-tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff8
3368 100644 blob 8d14531846b95bfa3564b58ccfb7913a034323b8 .gitignore
3369 100644 blob ebf9bf84da0aab5ed944264a5db2a65fe3a3e883 .mailmap
3370 100644 blob ca442d313d86dc67e0a2e5d584b465bd382cbf5c COPYING
3371 ...
3372 100644 blob 4b9458b3786228369c63936db65827de3cc06200 myfile
3373 ...
3374 ------------------------------------------------
3375
3376 So now you know that the missing blob was the data for a file named
3377 `myfile`. And chances are you can also identify the directory--let's
3378 say it's in `somedirectory`. If you're lucky the missing copy might be
3379 the same as the copy you have checked out in your working tree at
3380 `somedirectory/myfile`; you can test whether that's right with
3381 linkgit:git-hash-object[1]:
3382
3383 ------------------------------------------------
3384 $ git hash-object -w somedirectory/myfile
3385 ------------------------------------------------
3386
3387 which will create and store a blob object with the contents of
3388 somedirectory/myfile, and output the SHA-1 of that object. if you're
3389 extremely lucky it might be 4b9458b3786228369c63936db65827de3cc06200, in
3390 which case you've guessed right, and the corruption is fixed!
3391
3392 Otherwise, you need more information. How do you tell which version of
3393 the file has been lost?
3394
3395 The easiest way to do this is with:
3396
3397 ------------------------------------------------
3398 $ git log --raw --all --full-history -- somedirectory/myfile
3399 ------------------------------------------------
3400
3401 Because you're asking for raw output, you'll now get something like
3402
3403 ------------------------------------------------
3404 commit abc
3405 Author:
3406 Date:
3407 ...
3408 :100644 100644 4b9458b... newsha... M somedirectory/myfile
3409
3410
3411 commit xyz
3412 Author:
3413 Date:
3414
3415 ...
3416 :100644 100644 oldsha... 4b9458b... M somedirectory/myfile
3417 ------------------------------------------------
3418
3419 This tells you that the immediately following version of the file was
3420 "newsha", and that the immediately preceding version was "oldsha".
3421 You also know the commit messages that went with the change from oldsha
3422 to 4b9458b and with the change from 4b9458b to newsha.
3423
3424 If you've been committing small enough changes, you may now have a good
3425 shot at reconstructing the contents of the in-between state 4b9458b.
3426
3427 If you can do that, you can now recreate the missing object with
3428
3429 ------------------------------------------------
3430 $ git hash-object -w <recreated-file>
3431 ------------------------------------------------
3432
3433 and your repository is good again!
3434
3435 (Btw, you could have ignored the `fsck`, and started with doing a
3436
3437 ------------------------------------------------
3438 $ git log --raw --all
3439 ------------------------------------------------
3440
3441 and just looked for the sha of the missing object (4b9458b..) in that
3442 whole thing. It's up to you--Git does *have* a lot of information, it is
3443 just missing one particular blob version.
3444
3445 [[the-index]]
3446 The index
3447 -----------
3448
3449 The index is a binary file (generally kept in `.git/index`) containing a
3450 sorted list of path names, each with permissions and the SHA-1 of a blob
3451 object; linkgit:git-ls-files[1] can show you the contents of the index:
3452
3453 -------------------------------------------------
3454 $ git ls-files --stage
3455 100644 63c918c667fa005ff12ad89437f2fdc80926e21c 0 .gitignore
3456 100644 5529b198e8d14decbe4ad99db3f7fb632de0439d 0 .mailmap
3457 100644 6ff87c4664981e4397625791c8ea3bbb5f2279a3 0 COPYING
3458 100644 a37b2152bd26be2c2289e1f57a292534a51a93c7 0 Documentation/.gitignore
3459 100644 fbefe9a45b00a54b58d94d06eca48b03d40a50e0 0 Documentation/Makefile
3460 ...
3461 100644 2511aef8d89ab52be5ec6a5e46236b4b6bcd07ea 0 xdiff/xtypes.h
3462 100644 2ade97b2574a9f77e7ae4002a4e07a6a38e46d07 0 xdiff/xutils.c
3463 100644 d5de8292e05e7c36c4b68857c1cf9855e3d2f70a 0 xdiff/xutils.h
3464 -------------------------------------------------
3465
3466 Note that in older documentation you may see the index called the
3467 "current directory cache" or just the "cache". It has three important
3468 properties:
3469
3470 1. The index contains all the information necessary to generate a single
3471 (uniquely determined) tree object.
3472 +
3473 For example, running linkgit:git-commit[1] generates this tree object
3474 from the index, stores it in the object database, and uses it as the
3475 tree object associated with the new commit.
3476
3477 2. The index enables fast comparisons between the tree object it defines
3478 and the working tree.
3479 +
3480 It does this by storing some additional data for each entry (such as
3481 the last modified time). This data is not displayed above, and is not
3482 stored in the created tree object, but it can be used to determine
3483 quickly which files in the working directory differ from what was
3484 stored in the index, and thus save Git from having to read all of the
3485 data from such files to look for changes.
3486
3487 3. It can efficiently represent information about merge conflicts
3488 between different tree objects, allowing each pathname to be
3489 associated with sufficient information about the trees involved that
3490 you can create a three-way merge between them.
3491 +
3492 We saw in <<conflict-resolution>> that during a merge the index can
3493 store multiple versions of a single file (called "stages"). The third
3494 column in the linkgit:git-ls-files[1] output above is the stage
3495 number, and will take on values other than 0 for files with merge
3496 conflicts.
3497
3498 The index is thus a sort of temporary staging area, which is filled with
3499 a tree which you are in the process of working on.
3500
3501 If you blow the index away entirely, you generally haven't lost any
3502 information as long as you have the name of the tree that it described.
3503
3504 [[submodules]]
3505 Submodules
3506 ==========
3507
3508 Large projects are often composed of smaller, self-contained modules. For
3509 example, an embedded Linux distribution's source tree would include every
3510 piece of software in the distribution with some local modifications; a movie
3511 player might need to build against a specific, known-working version of a
3512 decompression library; several independent programs might all share the same
3513 build scripts.
3514
3515 With centralized revision control systems this is often accomplished by
3516 including every module in one single repository. Developers can check out
3517 all modules or only the modules they need to work with. They can even modify
3518 files across several modules in a single commit while moving things around
3519 or updating APIs and translations.
3520
3521 Git does not allow partial checkouts, so duplicating this approach in Git
3522 would force developers to keep a local copy of modules they are not
3523 interested in touching. Commits in an enormous checkout would be slower
3524 than you'd expect as Git would have to scan every directory for changes.
3525 If modules have a lot of local history, clones would take forever.
3526
3527 On the plus side, distributed revision control systems can much better
3528 integrate with external sources. In a centralized model, a single arbitrary
3529 snapshot of the external project is exported from its own revision control
3530 and then imported into the local revision control on a vendor branch. All
3531 the history is hidden. With distributed revision control you can clone the
3532 entire external history and much more easily follow development and re-merge
3533 local changes.
3534
3535 Git's submodule support allows a repository to contain, as a subdirectory, a
3536 checkout of an external project. Submodules maintain their own identity;
3537 the submodule support just stores the submodule repository location and
3538 commit ID, so other developers who clone the containing project
3539 ("superproject") can easily clone all the submodules at the same revision.
3540 Partial checkouts of the superproject are possible: you can tell Git to
3541 clone none, some or all of the submodules.
3542
3543 The linkgit:git-submodule[1] command is available since Git 1.5.3. Users
3544 with Git 1.5.2 can look up the submodule commits in the repository and
3545 manually check them out; earlier versions won't recognize the submodules at
3546 all.
3547
3548 To see how submodule support works, create four example
3549 repositories that can be used later as a submodule:
3550
3551 -------------------------------------------------
3552 $ mkdir ~/git
3553 $ cd ~/git
3554 $ for i in a b c d
3555 do
3556 mkdir $i
3557 cd $i
3558 git init
3559 echo "module $i" > $i.txt
3560 git add $i.txt
3561 git commit -m "Initial commit, submodule $i"
3562 cd ..
3563 done
3564 -------------------------------------------------
3565
3566 Now create the superproject and add all the submodules:
3567
3568 -------------------------------------------------
3569 $ mkdir super
3570 $ cd super
3571 $ git init
3572 $ for i in a b c d
3573 do
3574 git submodule add ~/git/$i $i
3575 done
3576 -------------------------------------------------
3577
3578 NOTE: Do not use local URLs here if you plan to publish your superproject!
3579
3580 See what files `git submodule` created:
3581
3582 -------------------------------------------------
3583 $ ls -a
3584 . .. .git .gitmodules a b c d
3585 -------------------------------------------------
3586
3587 The `git submodule add <repo> <path>` command does a couple of things:
3588
3589 - It clones the submodule from `<repo>` to the given `<path>` under the
3590 current directory and by default checks out the master branch.
3591 - It adds the submodule's clone path to the linkgit:gitmodules[5] file and
3592 adds this file to the index, ready to be committed.
3593 - It adds the submodule's current commit ID to the index, ready to be
3594 committed.
3595
3596 Commit the superproject:
3597
3598 -------------------------------------------------
3599 $ git commit -m "Add submodules a, b, c and d."
3600 -------------------------------------------------
3601
3602 Now clone the superproject:
3603
3604 -------------------------------------------------
3605 $ cd ..
3606 $ git clone super cloned
3607 $ cd cloned
3608 -------------------------------------------------
3609
3610 The submodule directories are there, but they're empty:
3611
3612 -------------------------------------------------
3613 $ ls -a a
3614 . ..
3615 $ git submodule status
3616 -d266b9873ad50488163457f025db7cdd9683d88b a
3617 -e81d457da15309b4fef4249aba9b50187999670d b
3618 -c1536a972b9affea0f16e0680ba87332dc059146 c
3619 -d96249ff5d57de5de093e6baff9e0aafa5276a74 d
3620 -------------------------------------------------
3621
3622 NOTE: The commit object names shown above would be different for you, but they
3623 should match the HEAD commit object names of your repositories. You can check
3624 it by running `git ls-remote ../a`.
3625
3626 Pulling down the submodules is a two-step process. First run `git submodule
3627 init` to add the submodule repository URLs to `.git/config`:
3628
3629 -------------------------------------------------
3630 $ git submodule init
3631 -------------------------------------------------
3632
3633 Now use `git submodule update` to clone the repositories and check out the
3634 commits specified in the superproject:
3635
3636 -------------------------------------------------
3637 $ git submodule update
3638 $ cd a
3639 $ ls -a
3640 . .. .git a.txt
3641 -------------------------------------------------
3642
3643 One major difference between `git submodule update` and `git submodule add` is
3644 that `git submodule update` checks out a specific commit, rather than the tip
3645 of a branch. It's like checking out a tag: the head is detached, so you're not
3646 working on a branch.
3647
3648 -------------------------------------------------
3649 $ git branch
3650 * (detached from d266b98)
3651 master
3652 -------------------------------------------------
3653
3654 If you want to make a change within a submodule and you have a detached head,
3655 then you should create or checkout a branch, make your changes, publish the
3656 change within the submodule, and then update the superproject to reference the
3657 new commit:
3658
3659 -------------------------------------------------
3660 $ git checkout master
3661 -------------------------------------------------
3662
3663 or
3664
3665 -------------------------------------------------
3666 $ git checkout -b fix-up
3667 -------------------------------------------------
3668
3669 then
3670
3671 -------------------------------------------------
3672 $ echo "adding a line again" >> a.txt
3673 $ git commit -a -m "Updated the submodule from within the superproject."
3674 $ git push
3675 $ cd ..
3676 $ git diff
3677 diff --git a/a b/a
3678 index d266b98..261dfac 160000
3679 --- a/a
3680 +++ b/a
3681 @@ -1 +1 @@
3682 -Subproject commit d266b9873ad50488163457f025db7cdd9683d88b
3683 +Subproject commit 261dfac35cb99d380eb966e102c1197139f7fa24
3684 $ git add a
3685 $ git commit -m "Updated submodule a."
3686 $ git push
3687 -------------------------------------------------
3688
3689 You have to run `git submodule update` after `git pull` if you want to update
3690 submodules, too.
3691
3692 Pitfalls with submodules
3693 ------------------------
3694
3695 Always publish the submodule change before publishing the change to the
3696 superproject that references it. If you forget to publish the submodule change,
3697 others won't be able to clone the repository:
3698
3699 -------------------------------------------------
3700 $ cd ~/git/super/a
3701 $ echo i added another line to this file >> a.txt
3702 $ git commit -a -m "doing it wrong this time"
3703 $ cd ..
3704 $ git add a
3705 $ git commit -m "Updated submodule a again."
3706 $ git push
3707 $ cd ~/git/cloned
3708 $ git pull
3709 $ git submodule update
3710 error: pathspec '261dfac35cb99d380eb966e102c1197139f7fa24' did not match any file(s) known to git.
3711 Did you forget to 'git add'?
3712 Unable to checkout '261dfac35cb99d380eb966e102c1197139f7fa24' in submodule path 'a'
3713 -------------------------------------------------
3714
3715 In older Git versions it could be easily forgotten to commit new or modified
3716 files in a submodule, which silently leads to similar problems as not pushing
3717 the submodule changes. Starting with Git 1.7.0 both `git status` and `git diff`
3718 in the superproject show submodules as modified when they contain new or
3719 modified files to protect against accidentally committing such a state. `git
3720 diff` will also add a `-dirty` to the work tree side when generating patch
3721 output or used with the `--submodule` option:
3722
3723 -------------------------------------------------
3724 $ git diff
3725 diff --git a/sub b/sub
3726 --- a/sub
3727 +++ b/sub
3728 @@ -1 +1 @@
3729 -Subproject commit 3f356705649b5d566d97ff843cf193359229a453
3730 +Subproject commit 3f356705649b5d566d97ff843cf193359229a453-dirty
3731 $ git diff --submodule
3732 Submodule sub 3f35670..3f35670-dirty:
3733 -------------------------------------------------
3734
3735 You also should not rewind branches in a submodule beyond commits that were
3736 ever recorded in any superproject.
3737
3738 It's not safe to run `git submodule update` if you've made and committed
3739 changes within a submodule without checking out a branch first. They will be
3740 silently overwritten:
3741
3742 -------------------------------------------------
3743 $ cat a.txt
3744 module a
3745 $ echo line added from private2 >> a.txt
3746 $ git commit -a -m "line added inside private2"
3747 $ cd ..
3748 $ git submodule update
3749 Submodule path 'a': checked out 'd266b9873ad50488163457f025db7cdd9683d88b'
3750 $ cd a
3751 $ cat a.txt
3752 module a
3753 -------------------------------------------------
3754
3755 NOTE: The changes are still visible in the submodule's reflog.
3756
3757 If you have uncommitted changes in your submodule working tree, `git
3758 submodule update` will not overwrite them. Instead, you get the usual
3759 warning about not being able switch from a dirty branch.
3760
3761 [[low-level-operations]]
3762 Low-level Git operations
3763 ========================
3764
3765 Many of the higher-level commands were originally implemented as shell
3766 scripts using a smaller core of low-level Git commands. These can still
3767 be useful when doing unusual things with Git, or just as a way to
3768 understand its inner workings.
3769
3770 [[object-manipulation]]
3771 Object access and manipulation
3772 ------------------------------
3773
3774 The linkgit:git-cat-file[1] command can show the contents of any object,
3775 though the higher-level linkgit:git-show[1] is usually more useful.
3776
3777 The linkgit:git-commit-tree[1] command allows constructing commits with
3778 arbitrary parents and trees.
3779
3780 A tree can be created with linkgit:git-write-tree[1] and its data can be
3781 accessed by linkgit:git-ls-tree[1]. Two trees can be compared with
3782 linkgit:git-diff-tree[1].
3783
3784 A tag is created with linkgit:git-mktag[1], and the signature can be
3785 verified by linkgit:git-verify-tag[1], though it is normally simpler to
3786 use linkgit:git-tag[1] for both.
3787
3788 [[the-workflow]]
3789 The Workflow
3790 ------------
3791
3792 High-level operations such as linkgit:git-commit[1],
3793 linkgit:git-checkout[1] and linkgit:git-reset[1] work by moving data
3794 between the working tree, the index, and the object database. Git
3795 provides low-level operations which perform each of these steps
3796 individually.
3797
3798 Generally, all Git operations work on the index file. Some operations
3799 work *purely* on the index file (showing the current state of the
3800 index), but most operations move data between the index file and either
3801 the database or the working directory. Thus there are four main
3802 combinations:
3803
3804 [[working-directory-to-index]]
3805 working directory -> index
3806 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3807
3808 The linkgit:git-update-index[1] command updates the index with
3809 information from the working directory. You generally update the
3810 index information by just specifying the filename you want to update,
3811 like so:
3812
3813 -------------------------------------------------
3814 $ git update-index filename
3815 -------------------------------------------------
3816
3817 but to avoid common mistakes with filename globbing etc., the command
3818 will not normally add totally new entries or remove old entries,
3819 i.e. it will normally just update existing cache entries.
3820
3821 To tell Git that yes, you really do realize that certain files no
3822 longer exist, or that new files should be added, you
3823 should use the `--remove` and `--add` flags respectively.
3824
3825 NOTE! A `--remove` flag does 'not' mean that subsequent filenames will
3826 necessarily be removed: if the files still exist in your directory
3827 structure, the index will be updated with their new status, not
3828 removed. The only thing `--remove` means is that update-index will be
3829 considering a removed file to be a valid thing, and if the file really
3830 does not exist any more, it will update the index accordingly.
3831
3832 As a special case, you can also do `git update-index --refresh`, which
3833 will refresh the "stat" information of each index to match the current
3834 stat information. It will 'not' update the object status itself, and
3835 it will only update the fields that are used to quickly test whether
3836 an object still matches its old backing store object.
3837
3838 The previously introduced linkgit:git-add[1] is just a wrapper for
3839 linkgit:git-update-index[1].
3840
3841 [[index-to-object-database]]
3842 index -> object database
3843 ~~~~~~~~~~~~~~~~~~~~~~~~
3844
3845 You write your current index file to a "tree" object with the program
3846
3847 -------------------------------------------------
3848 $ git write-tree
3849 -------------------------------------------------
3850
3851 that doesn't come with any options--it will just write out the
3852 current index into the set of tree objects that describe that state,
3853 and it will return the name of the resulting top-level tree. You can
3854 use that tree to re-generate the index at any time by going in the
3855 other direction:
3856
3857 [[object-database-to-index]]
3858 object database -> index
3859 ~~~~~~~~~~~~~~~~~~~~~~~~
3860
3861 You read a "tree" file from the object database, and use that to
3862 populate (and overwrite--don't do this if your index contains any
3863 unsaved state that you might want to restore later!) your current
3864 index. Normal operation is just
3865
3866 -------------------------------------------------
3867 $ git read-tree <SHA-1 of tree>
3868 -------------------------------------------------
3869
3870 and your index file will now be equivalent to the tree that you saved
3871 earlier. However, that is only your 'index' file: your working
3872 directory contents have not been modified.
3873
3874 [[index-to-working-directory]]
3875 index -> working directory
3876 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3877
3878 You update your working directory from the index by "checking out"
3879 files. This is not a very common operation, since normally you'd just
3880 keep your files updated, and rather than write to your working
3881 directory, you'd tell the index files about the changes in your
3882 working directory (i.e. `git update-index`).
3883
3884 However, if you decide to jump to a new version, or check out somebody
3885 else's version, or just restore a previous tree, you'd populate your
3886 index file with read-tree, and then you need to check out the result
3887 with
3888
3889 -------------------------------------------------
3890 $ git checkout-index filename
3891 -------------------------------------------------
3892
3893 or, if you want to check out all of the index, use `-a`.
3894
3895 NOTE! `git checkout-index` normally refuses to overwrite old files, so
3896 if you have an old version of the tree already checked out, you will
3897 need to use the `-f` flag ('before' the `-a` flag or the filename) to
3898 'force' the checkout.
3899
3900
3901 Finally, there are a few odds and ends which are not purely moving
3902 from one representation to the other:
3903
3904 [[tying-it-all-together]]
3905 Tying it all together
3906 ~~~~~~~~~~~~~~~~~~~~~
3907
3908 To commit a tree you have instantiated with `git write-tree`, you'd
3909 create a "commit" object that refers to that tree and the history
3910 behind it--most notably the "parent" commits that preceded it in
3911 history.
3912
3913 Normally a "commit" has one parent: the previous state of the tree
3914 before a certain change was made. However, sometimes it can have two
3915 or more parent commits, in which case we call it a "merge", due to the
3916 fact that such a commit brings together ("merges") two or more
3917 previous states represented by other commits.
3918
3919 In other words, while a "tree" represents a particular directory state
3920 of a working directory, a "commit" represents that state in time,
3921 and explains how we got there.
3922
3923 You create a commit object by giving it the tree that describes the
3924 state at the time of the commit, and a list of parents:
3925
3926 -------------------------------------------------
3927 $ git commit-tree <tree> -p <parent> [(-p <parent2>)...]
3928 -------------------------------------------------
3929
3930 and then giving the reason for the commit on stdin (either through
3931 redirection from a pipe or file, or by just typing it at the tty).
3932
3933 `git commit-tree` will return the name of the object that represents
3934 that commit, and you should save it away for later use. Normally,
3935 you'd commit a new `HEAD` state, and while Git doesn't care where you
3936 save the note about that state, in practice we tend to just write the
3937 result to the file pointed at by `.git/HEAD`, so that we can always see
3938 what the last committed state was.
3939
3940 Here is a picture that illustrates how various pieces fit together:
3941
3942 ------------
3943
3944 commit-tree
3945 commit obj
3946 +----+
3947 | |
3948 | |
3949 V V
3950 +-----------+
3951 | Object DB |
3952 | Backing |
3953 | Store |
3954 +-----------+
3955 ^
3956 write-tree | |
3957 tree obj | |
3958 | | read-tree
3959 | | tree obj
3960 V
3961 +-----------+
3962 | Index |
3963 | "cache" |
3964 +-----------+
3965 update-index ^
3966 blob obj | |
3967 | |
3968 checkout-index -u | | checkout-index
3969 stat | | blob obj
3970 V
3971 +-----------+
3972 | Working |
3973 | Directory |
3974 +-----------+
3975
3976 ------------
3977
3978
3979 [[examining-the-data]]
3980 Examining the data
3981 ------------------
3982
3983 You can examine the data represented in the object database and the
3984 index with various helper tools. For every object, you can use
3985 linkgit:git-cat-file[1] to examine details about the
3986 object:
3987
3988 -------------------------------------------------
3989 $ git cat-file -t <objectname>
3990 -------------------------------------------------
3991
3992 shows the type of the object, and once you have the type (which is
3993 usually implicit in where you find the object), you can use
3994
3995 -------------------------------------------------
3996 $ git cat-file blob|tree|commit|tag <objectname>
3997 -------------------------------------------------
3998
3999 to show its contents. NOTE! Trees have binary content, and as a result
4000 there is a special helper for showing that content, called
4001 `git ls-tree`, which turns the binary content into a more easily
4002 readable form.
4003
4004 It's especially instructive to look at "commit" objects, since those
4005 tend to be small and fairly self-explanatory. In particular, if you
4006 follow the convention of having the top commit name in `.git/HEAD`,
4007 you can do
4008
4009 -------------------------------------------------
4010 $ git cat-file commit HEAD
4011 -------------------------------------------------
4012
4013 to see what the top commit was.
4014
4015 [[merging-multiple-trees]]
4016 Merging multiple trees
4017 ----------------------
4018
4019 Git can help you perform a three-way merge, which can in turn be
4020 used for a many-way merge by repeating the merge procedure several
4021 times. The usual situation is that you only do one three-way merge
4022 (reconciling two lines of history) and commit the result, but if
4023 you like to, you can merge several branches in one go.
4024
4025 To perform a three-way merge, you start with the two commits you
4026 want to merge, find their closest common parent (a third commit),
4027 and compare the trees corresponding to these three commits.
4028
4029 To get the "base" for the merge, look up the common parent of two
4030 commits:
4031
4032 -------------------------------------------------
4033 $ git merge-base <commit1> <commit2>
4034 -------------------------------------------------
4035
4036 This prints the name of a commit they are both based on. You should
4037 now look up the tree objects of those commits, which you can easily
4038 do with
4039
4040 -------------------------------------------------
4041 $ git cat-file commit <commitname> | head -1
4042 -------------------------------------------------
4043
4044 since the tree object information is always the first line in a commit
4045 object.
4046
4047 Once you know the three trees you are going to merge (the one "original"
4048 tree, aka the common tree, and the two "result" trees, aka the branches
4049 you want to merge), you do a "merge" read into the index. This will
4050 complain if it has to throw away your old index contents, so you should
4051 make sure that you've committed those--in fact you would normally
4052 always do a merge against your last commit (which should thus match what
4053 you have in your current index anyway).
4054
4055 To do the merge, do
4056
4057 -------------------------------------------------
4058 $ git read-tree -m -u <origtree> <yourtree> <targettree>
4059 -------------------------------------------------
4060
4061 which will do all trivial merge operations for you directly in the
4062 index file, and you can just write the result out with
4063 `git write-tree`.
4064
4065
4066 [[merging-multiple-trees-2]]
4067 Merging multiple trees, continued
4068 ---------------------------------
4069
4070 Sadly, many merges aren't trivial. If there are files that have
4071 been added, moved or removed, or if both branches have modified the
4072 same file, you will be left with an index tree that contains "merge
4073 entries" in it. Such an index tree can 'NOT' be written out to a tree
4074 object, and you will have to resolve any such merge clashes using
4075 other tools before you can write out the result.
4076
4077 You can examine such index state with `git ls-files --unmerged`
4078 command. An example:
4079
4080 ------------------------------------------------
4081 $ git read-tree -m $orig HEAD $target
4082 $ git ls-files --unmerged
4083 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello.c
4084 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello.c
4085 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello.c
4086 ------------------------------------------------
4087
4088 Each line of the `git ls-files --unmerged` output begins with
4089 the blob mode bits, blob SHA-1, 'stage number', and the
4090 filename. The 'stage number' is Git's way to say which tree it
4091 came from: stage 1 corresponds to the `$orig` tree, stage 2 to
4092 the `HEAD` tree, and stage 3 to the `$target` tree.
4093
4094 Earlier we said that trivial merges are done inside
4095 `git read-tree -m`. For example, if the file did not change
4096 from `$orig` to `HEAD` or `$target`, or if the file changed
4097 from `$orig` to `HEAD` and `$orig` to `$target` the same way,
4098 obviously the final outcome is what is in `HEAD`. What the
4099 above example shows is that file `hello.c` was changed from
4100 `$orig` to `HEAD` and `$orig` to `$target` in a different way.
4101 You could resolve this by running your favorite 3-way merge
4102 program, e.g. `diff3`, `merge`, or Git's own merge-file, on
4103 the blob objects from these three stages yourself, like this:
4104
4105 ------------------------------------------------
4106 $ git cat-file blob 263414f... >hello.c~1
4107 $ git cat-file blob 06fa6a2... >hello.c~2
4108 $ git cat-file blob cc44c73... >hello.c~3
4109 $ git merge-file hello.c~2 hello.c~1 hello.c~3
4110 ------------------------------------------------
4111
4112 This would leave the merge result in `hello.c~2` file, along
4113 with conflict markers if there are conflicts. After verifying
4114 the merge result makes sense, you can tell Git what the final
4115 merge result for this file is by:
4116
4117 -------------------------------------------------
4118 $ mv -f hello.c~2 hello.c
4119 $ git update-index hello.c
4120 -------------------------------------------------
4121
4122 When a path is in the "unmerged" state, running `git update-index` for
4123 that path tells Git to mark the path resolved.
4124
4125 The above is the description of a Git merge at the lowest level,
4126 to help you understand what conceptually happens under the hood.
4127 In practice, nobody, not even Git itself, runs `git cat-file` three times
4128 for this. There is a `git merge-index` program that extracts the
4129 stages to temporary files and calls a "merge" script on it:
4130
4131 -------------------------------------------------
4132 $ git merge-index git-merge-one-file hello.c
4133 -------------------------------------------------
4134
4135 and that is what higher level `git merge -s resolve` is implemented with.
4136
4137 [[hacking-git]]
4138 Hacking Git
4139 ===========
4140
4141 This chapter covers internal details of the Git implementation which
4142 probably only Git developers need to understand.
4143
4144 [[object-details]]
4145 Object storage format
4146 ---------------------
4147
4148 All objects have a statically determined "type" which identifies the
4149 format of the object (i.e. how it is used, and how it can refer to other
4150 objects). There are currently four different object types: "blob",
4151 "tree", "commit", and "tag".
4152
4153 Regardless of object type, all objects share the following
4154 characteristics: they are all deflated with zlib, and have a header
4155 that not only specifies their type, but also provides size information
4156 about the data in the object. It's worth noting that the SHA-1 hash
4157 that is used to name the object is the hash of the original data
4158 plus this header, so `sha1sum` 'file' does not match the object name
4159 for 'file'.
4160
4161 As a result, the general consistency of an object can always be tested
4162 independently of the contents or the type of the object: all objects can
4163 be validated by verifying that (a) their hashes match the content of the
4164 file and (b) the object successfully inflates to a stream of bytes that
4165 forms a sequence of
4166 `<ascii type without space> + <space> + <ascii decimal size> +
4167 <byte\0> + <binary object data>`.
4168
4169 The structured objects can further have their structure and
4170 connectivity to other objects verified. This is generally done with
4171 the `git fsck` program, which generates a full dependency graph
4172 of all objects, and verifies their internal consistency (in addition
4173 to just verifying their superficial consistency through the hash).
4174
4175 [[birdview-on-the-source-code]]
4176 A birds-eye view of Git's source code
4177 -------------------------------------
4178
4179 It is not always easy for new developers to find their way through Git's
4180 source code. This section gives you a little guidance to show where to
4181 start.
4182
4183 A good place to start is with the contents of the initial commit, with:
4184
4185 ----------------------------------------------------
4186 $ git checkout e83c5163
4187 ----------------------------------------------------
4188
4189 The initial revision lays the foundation for almost everything Git has
4190 today, but is small enough to read in one sitting.
4191
4192 Note that terminology has changed since that revision. For example, the
4193 README in that revision uses the word "changeset" to describe what we
4194 now call a <<def_commit_object,commit>>.
4195
4196 Also, we do not call it "cache" any more, but rather "index"; however, the
4197 file is still called `cache.h`. Remark: Not much reason to change it now,
4198 especially since there is no good single name for it anyway, because it is
4199 basically _the_ header file which is included by _all_ of Git's C sources.
4200
4201 If you grasp the ideas in that initial commit, you should check out a
4202 more recent version and skim `cache.h`, `object.h` and `commit.h`.
4203
4204 In the early days, Git (in the tradition of UNIX) was a bunch of programs
4205 which were extremely simple, and which you used in scripts, piping the
4206 output of one into another. This turned out to be good for initial
4207 development, since it was easier to test new things. However, recently
4208 many of these parts have become builtins, and some of the core has been
4209 "libified", i.e. put into libgit.a for performance, portability reasons,
4210 and to avoid code duplication.
4211
4212 By now, you know what the index is (and find the corresponding data
4213 structures in `cache.h`), and that there are just a couple of object types
4214 (blobs, trees, commits and tags) which inherit their common structure from
4215 `struct object`, which is their first member (and thus, you can cast e.g.
4216 `(struct object *)commit` to achieve the _same_ as `&commit->object`, i.e.
4217 get at the object name and flags).
4218
4219 Now is a good point to take a break to let this information sink in.
4220
4221 Next step: get familiar with the object naming. Read <<naming-commits>>.
4222 There are quite a few ways to name an object (and not only revisions!).
4223 All of these are handled in `sha1_name.c`. Just have a quick look at
4224 the function `get_sha1()`. A lot of the special handling is done by
4225 functions like `get_sha1_basic()` or the likes.
4226
4227 This is just to get you into the groove for the most libified part of Git:
4228 the revision walker.
4229
4230 Basically, the initial version of `git log` was a shell script:
4231
4232 ----------------------------------------------------------------
4233 $ git-rev-list --pretty $(git-rev-parse --default HEAD "$@") | \
4234 LESS=-S ${PAGER:-less}
4235 ----------------------------------------------------------------
4236
4237 What does this mean?
4238
4239 `git rev-list` is the original version of the revision walker, which
4240 _always_ printed a list of revisions to stdout. It is still functional,
4241 and needs to, since most new Git commands start out as scripts using
4242 `git rev-list`.
4243
4244 `git rev-parse` is not as important any more; it was only used to filter out
4245 options that were relevant for the different plumbing commands that were
4246 called by the script.
4247
4248 Most of what `git rev-list` did is contained in `revision.c` and
4249 `revision.h`. It wraps the options in a struct named `rev_info`, which
4250 controls how and what revisions are walked, and more.
4251
4252 The original job of `git rev-parse` is now taken by the function
4253 `setup_revisions()`, which parses the revisions and the common command line
4254 options for the revision walker. This information is stored in the struct
4255 `rev_info` for later consumption. You can do your own command line option
4256 parsing after calling `setup_revisions()`. After that, you have to call
4257 `prepare_revision_walk()` for initialization, and then you can get the
4258 commits one by one with the function `get_revision()`.
4259
4260 If you are interested in more details of the revision walking process,
4261 just have a look at the first implementation of `cmd_log()`; call
4262 `git show v1.3.0~155^2~4` and scroll down to that function (note that you
4263 no longer need to call `setup_pager()` directly).
4264
4265 Nowadays, `git log` is a builtin, which means that it is _contained_ in the
4266 command `git`. The source side of a builtin is
4267
4268 - a function called `cmd_<bla>`, typically defined in `builtin/<bla.c>`
4269 (note that older versions of Git used to have it in `builtin-<bla>.c`
4270 instead), and declared in `builtin.h`.
4271
4272 - an entry in the `commands[]` array in `git.c`, and
4273
4274 - an entry in `BUILTIN_OBJECTS` in the `Makefile`.
4275
4276 Sometimes, more than one builtin is contained in one source file. For
4277 example, `cmd_whatchanged()` and `cmd_log()` both reside in `builtin/log.c`,
4278 since they share quite a bit of code. In that case, the commands which are
4279 _not_ named like the `.c` file in which they live have to be listed in
4280 `BUILT_INS` in the `Makefile`.
4281
4282 `git log` looks more complicated in C than it does in the original script,
4283 but that allows for a much greater flexibility and performance.
4284
4285 Here again it is a good point to take a pause.
4286
4287 Lesson three is: study the code. Really, it is the best way to learn about
4288 the organization of Git (after you know the basic concepts).
4289
4290 So, think about something which you are interested in, say, "how can I
4291 access a blob just knowing the object name of it?". The first step is to
4292 find a Git command with which you can do it. In this example, it is either
4293 `git show` or `git cat-file`.
4294
4295 For the sake of clarity, let's stay with `git cat-file`, because it
4296
4297 - is plumbing, and
4298
4299 - was around even in the initial commit (it literally went only through
4300 some 20 revisions as `cat-file.c`, was renamed to `builtin/cat-file.c`
4301 when made a builtin, and then saw less than 10 versions).
4302
4303 So, look into `builtin/cat-file.c`, search for `cmd_cat_file()` and look what
4304 it does.
4305
4306 ------------------------------------------------------------------
4307 git_config(git_default_config);
4308 if (argc != 3)
4309 usage("git cat-file [-t|-s|-e|-p|<type>] <sha1>");
4310 if (get_sha1(argv[2], sha1))
4311 die("Not a valid object name %s", argv[2]);
4312 ------------------------------------------------------------------
4313
4314 Let's skip over the obvious details; the only really interesting part
4315 here is the call to `get_sha1()`. It tries to interpret `argv[2]` as an
4316 object name, and if it refers to an object which is present in the current
4317 repository, it writes the resulting SHA-1 into the variable `sha1`.
4318
4319 Two things are interesting here:
4320
4321 - `get_sha1()` returns 0 on _success_. This might surprise some new
4322 Git hackers, but there is a long tradition in UNIX to return different
4323 negative numbers in case of different errors--and 0 on success.
4324
4325 - the variable `sha1` in the function signature of `get_sha1()` is `unsigned
4326 char *`, but is actually expected to be a pointer to `unsigned
4327 char[20]`. This variable will contain the 160-bit SHA-1 of the given
4328 commit. Note that whenever a SHA-1 is passed as `unsigned char *`, it
4329 is the binary representation, as opposed to the ASCII representation in
4330 hex characters, which is passed as `char *`.
4331
4332 You will see both of these things throughout the code.
4333
4334 Now, for the meat:
4335
4336 -----------------------------------------------------------------------------
4337 case 0:
4338 buf = read_object_with_reference(sha1, argv[1], &size, NULL);
4339 -----------------------------------------------------------------------------
4340
4341 This is how you read a blob (actually, not only a blob, but any type of
4342 object). To know how the function `read_object_with_reference()` actually
4343 works, find the source code for it (something like `git grep
4344 read_object_with | grep ":[a-z]"` in the Git repository), and read
4345 the source.
4346
4347 To find out how the result can be used, just read on in `cmd_cat_file()`:
4348
4349 -----------------------------------
4350 write_or_die(1, buf, size);
4351 -----------------------------------
4352
4353 Sometimes, you do not know where to look for a feature. In many such cases,
4354 it helps to search through the output of `git log`, and then `git show` the
4355 corresponding commit.
4356
4357 Example: If you know that there was some test case for `git bundle`, but
4358 do not remember where it was (yes, you _could_ `git grep bundle t/`, but that
4359 does not illustrate the point!):
4360
4361 ------------------------
4362 $ git log --no-merges t/
4363 ------------------------
4364
4365 In the pager (`less`), just search for "bundle", go a few lines back,
4366 and see that it is in commit 18449ab0... Now just copy this object name,
4367 and paste it into the command line
4368
4369 -------------------
4370 $ git show 18449ab0
4371 -------------------
4372
4373 Voila.
4374
4375 Another example: Find out what to do in order to make some script a
4376 builtin:
4377
4378 -------------------------------------------------
4379 $ git log --no-merges --diff-filter=A builtin/*.c
4380 -------------------------------------------------
4381
4382 You see, Git is actually the best tool to find out about the source of Git
4383 itself!
4384
4385 [[glossary]]
4386 Git Glossary
4387 ============
4388
4389 include::glossary-content.txt[]
4390
4391 [[git-quick-start]]
4392 Appendix A: Git Quick Reference
4393 ===============================
4394
4395 This is a quick summary of the major commands; the previous chapters
4396 explain how these work in more detail.
4397
4398 [[quick-creating-a-new-repository]]
4399 Creating a new repository
4400 -------------------------
4401
4402 From a tarball:
4403
4404 -----------------------------------------------
4405 $ tar xzf project.tar.gz
4406 $ cd project
4407 $ git init
4408 Initialized empty Git repository in .git/
4409 $ git add .
4410 $ git commit
4411 -----------------------------------------------
4412
4413 From a remote repository:
4414
4415 -----------------------------------------------
4416 $ git clone git://example.com/pub/project.git
4417 $ cd project
4418 -----------------------------------------------
4419
4420 [[managing-branches]]
4421 Managing branches
4422 -----------------
4423
4424 -----------------------------------------------
4425 $ git branch # list all local branches in this repo
4426 $ git checkout test # switch working directory to branch "test"
4427 $ git branch new # create branch "new" starting at current HEAD
4428 $ git branch -d new # delete branch "new"
4429 -----------------------------------------------
4430
4431 Instead of basing a new branch on current HEAD (the default), use:
4432
4433 -----------------------------------------------
4434 $ git branch new test # branch named "test"
4435 $ git branch new v2.6.15 # tag named v2.6.15
4436 $ git branch new HEAD^ # commit before the most recent
4437 $ git branch new HEAD^^ # commit before that
4438 $ git branch new test~10 # ten commits before tip of branch "test"
4439 -----------------------------------------------
4440
4441 Create and switch to a new branch at the same time:
4442
4443 -----------------------------------------------
4444 $ git checkout -b new v2.6.15
4445 -----------------------------------------------
4446
4447 Update and examine branches from the repository you cloned from:
4448
4449 -----------------------------------------------
4450 $ git fetch # update
4451 $ git branch -r # list
4452 origin/master
4453 origin/next
4454 ...
4455 $ git checkout -b masterwork origin/master
4456 -----------------------------------------------
4457
4458 Fetch a branch from a different repository, and give it a new
4459 name in your repository:
4460
4461 -----------------------------------------------
4462 $ git fetch git://example.com/project.git theirbranch:mybranch
4463 $ git fetch git://example.com/project.git v2.6.15:mybranch
4464 -----------------------------------------------
4465
4466 Keep a list of repositories you work with regularly:
4467
4468 -----------------------------------------------
4469 $ git remote add example git://example.com/project.git
4470 $ git remote # list remote repositories
4471 example
4472 origin
4473 $ git remote show example # get details
4474 * remote example
4475 URL: git://example.com/project.git
4476 Tracked remote branches
4477 master
4478 next
4479 ...
4480 $ git fetch example # update branches from example
4481 $ git branch -r # list all remote branches
4482 -----------------------------------------------
4483
4484
4485 [[exploring-history]]
4486 Exploring history
4487 -----------------
4488
4489 -----------------------------------------------
4490 $ gitk # visualize and browse history
4491 $ git log # list all commits
4492 $ git log src/ # ...modifying src/
4493 $ git log v2.6.15..v2.6.16 # ...in v2.6.16, not in v2.6.15
4494 $ git log master..test # ...in branch test, not in branch master
4495 $ git log test..master # ...in branch master, but not in test
4496 $ git log test...master # ...in one branch, not in both
4497 $ git log -S'foo()' # ...where difference contain "foo()"
4498 $ git log --since="2 weeks ago"
4499 $ git log -p # show patches as well
4500 $ git show # most recent commit
4501 $ git diff v2.6.15..v2.6.16 # diff between two tagged versions
4502 $ git diff v2.6.15..HEAD # diff with current head
4503 $ git grep "foo()" # search working directory for "foo()"
4504 $ git grep v2.6.15 "foo()" # search old tree for "foo()"
4505 $ git show v2.6.15:a.txt # look at old version of a.txt
4506 -----------------------------------------------
4507
4508 Search for regressions:
4509
4510 -----------------------------------------------
4511 $ git bisect start
4512 $ git bisect bad # current version is bad
4513 $ git bisect good v2.6.13-rc2 # last known good revision
4514 Bisecting: 675 revisions left to test after this
4515 # test here, then:
4516 $ git bisect good # if this revision is good, or
4517 $ git bisect bad # if this revision is bad.
4518 # repeat until done.
4519 -----------------------------------------------
4520
4521 [[making-changes]]
4522 Making changes
4523 --------------
4524
4525 Make sure Git knows who to blame:
4526
4527 ------------------------------------------------
4528 $ cat >>~/.gitconfig <<\EOF
4529 [user]
4530 name = Your Name Comes Here
4531 email = you@yourdomain.example.com
4532 EOF
4533 ------------------------------------------------
4534
4535 Select file contents to include in the next commit, then make the
4536 commit:
4537
4538 -----------------------------------------------
4539 $ git add a.txt # updated file
4540 $ git add b.txt # new file
4541 $ git rm c.txt # old file
4542 $ git commit
4543 -----------------------------------------------
4544
4545 Or, prepare and create the commit in one step:
4546
4547 -----------------------------------------------
4548 $ git commit d.txt # use latest content only of d.txt
4549 $ git commit -a # use latest content of all tracked files
4550 -----------------------------------------------
4551
4552 [[merging]]
4553 Merging
4554 -------
4555
4556 -----------------------------------------------
4557 $ git merge test # merge branch "test" into the current branch
4558 $ git pull git://example.com/project.git master
4559 # fetch and merge in remote branch
4560 $ git pull . test # equivalent to git merge test
4561 -----------------------------------------------
4562
4563 [[sharing-your-changes]]
4564 Sharing your changes
4565 --------------------
4566
4567 Importing or exporting patches:
4568
4569 -----------------------------------------------
4570 $ git format-patch origin..HEAD # format a patch for each commit
4571 # in HEAD but not in origin
4572 $ git am mbox # import patches from the mailbox "mbox"
4573 -----------------------------------------------
4574
4575 Fetch a branch in a different Git repository, then merge into the
4576 current branch:
4577
4578 -----------------------------------------------
4579 $ git pull git://example.com/project.git theirbranch
4580 -----------------------------------------------
4581
4582 Store the fetched branch into a local branch before merging into the
4583 current branch:
4584
4585 -----------------------------------------------
4586 $ git pull git://example.com/project.git theirbranch:mybranch
4587 -----------------------------------------------
4588
4589 After creating commits on a local branch, update the remote
4590 branch with your commits:
4591
4592 -----------------------------------------------
4593 $ git push ssh://example.com/project.git mybranch:theirbranch
4594 -----------------------------------------------
4595
4596 When remote and local branch are both named "test":
4597
4598 -----------------------------------------------
4599 $ git push ssh://example.com/project.git test
4600 -----------------------------------------------
4601
4602 Shortcut version for a frequently used remote repository:
4603
4604 -----------------------------------------------
4605 $ git remote add example ssh://example.com/project.git
4606 $ git push example test
4607 -----------------------------------------------
4608
4609 [[repository-maintenance]]
4610 Repository maintenance
4611 ----------------------
4612
4613 Check for corruption:
4614
4615 -----------------------------------------------
4616 $ git fsck
4617 -----------------------------------------------
4618
4619 Recompress, remove unused cruft:
4620
4621 -----------------------------------------------
4622 $ git gc
4623 -----------------------------------------------
4624
4625
4626 [[todo]]
4627 Appendix B: Notes and todo list for this manual
4628 ===============================================
4629
4630 This is a work in progress.
4631
4632 The basic requirements:
4633
4634 - It must be readable in order, from beginning to end, by someone
4635 intelligent with a basic grasp of the UNIX command line, but without
4636 any special knowledge of Git. If necessary, any other prerequisites
4637 should be specifically mentioned as they arise.
4638 - Whenever possible, section headings should clearly describe the task
4639 they explain how to do, in language that requires no more knowledge
4640 than necessary: for example, "importing patches into a project" rather
4641 than "the `git am` command"
4642
4643 Think about how to create a clear chapter dependency graph that will
4644 allow people to get to important topics without necessarily reading
4645 everything in between.
4646
4647 Scan `Documentation/` for other stuff left out; in particular:
4648
4649 - howto's
4650 - some of `technical/`?
4651 - hooks
4652 - list of commands in linkgit:git[1]
4653
4654 Scan email archives for other stuff left out
4655
4656 Scan man pages to see if any assume more background than this manual
4657 provides.
4658
4659 Add more good examples. Entire sections of just cookbook examples
4660 might be a good idea; maybe make an "advanced examples" section a
4661 standard end-of-chapter section?
4662
4663 Include cross-references to the glossary, where appropriate.
4664
4665 Add a section on working with other version control systems, including
4666 CVS, Subversion, and just imports of series of release tarballs.
4667
4668 More details on gitweb?
4669
4670 Write a chapter on using plumbing and writing scripts.
4671
4672 Alternates, clone -reference, etc.
4673
4674 More on recovery from repository corruption. See:
4675 http://marc.info/?l=git&m=117263864820799&w=2
4676 http://marc.info/?l=git&m=117147855503798&w=2