cf02e14ed0669bb5874dd9bff700f5a36bf1fa6b
   1Git User's Manual (for version 1.5.3 or newer)
   2______________________________________________
   3
   4
   5Git is a fast distributed revision control system.
   6
   7This manual is designed to be readable by someone with basic UNIX
   8command-line skills, but no previous knowledge of git.
   9
  10<<repositories-and-branches>> and <<exploring-git-history>> explain how
  11to fetch and study a project using git--read these chapters to learn how
  12to build and test a particular version of a software project, search for
  13regressions, and so on.
  14
  15People needing to do actual development will also want to read
  16<<Developing-with-git>> and <<sharing-development>>.
  17
  18Further chapters cover more specialized topics.
  19
  20Comprehensive reference documentation is available through the man
  21pages.  For a command such as "git clone", just use
  22
  23------------------------------------------------
  24$ man git-clone
  25------------------------------------------------
  26
  27See also <<git-quick-start>> for a brief overview of git commands,
  28without any explanation.
  29
  30Finally, see <<todo>> for ways that you can help make this manual more
  31complete.
  32
  33
  34[[repositories-and-branches]]
  35Repositories and Branches
  36=========================
  37
  38[[how-to-get-a-git-repository]]
  39How to get a git repository
  40---------------------------
  41
  42It will be useful to have a git repository to experiment with as you
  43read this manual.
  44
  45The best way to get one is by using the gitlink:git-clone[1] command to
  46download a copy of an existing repository.  If you don't already have a
  47project in mind, here are some interesting examples:
  48
  49------------------------------------------------
  50        # git itself (approx. 10MB download):
  51$ git clone git://git.kernel.org/pub/scm/git/git.git
  52        # the linux kernel (approx. 150MB download):
  53$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
  54------------------------------------------------
  55
  56The initial clone may be time-consuming for a large project, but you
  57will only need to clone once.
  58
  59The clone command creates a new directory named after the project
  60("git" or "linux-2.6" in the examples above).  After you cd into this
  61directory, you will see that it contains a copy of the project files,
  62together with a special top-level directory named ".git", which
  63contains all the information about the history of the project.
  64
  65[[how-to-check-out]]
  66How to check out a different version of a project
  67-------------------------------------------------
  68
  69Git is best thought of as a tool for storing the history of a collection
  70of files.  It stores the history as a compressed collection of
  71interrelated snapshots of the project's contents.  In git each such
  72version is called a <<def_commit,commit>>.
  73
  74A single git repository may contain multiple branches.  It keeps track
  75of them by keeping a list of <<def_head,heads>> which reference the
  76latest commit on each branch; the gitlink:git-branch[1] command shows
  77you the list of branch heads:
  78
  79------------------------------------------------
  80$ git branch
  81* master
  82------------------------------------------------
  83
  84A freshly cloned repository contains a single branch head, by default
  85named "master", with the working directory initialized to the state of
  86the project referred to by that branch head.
  87
  88Most projects also use <<def_tag,tags>>.  Tags, like heads, are
  89references into the project's history, and can be listed using the
  90gitlink:git-tag[1] command:
  91
  92------------------------------------------------
  93$ git tag -l
  94v2.6.11
  95v2.6.11-tree
  96v2.6.12
  97v2.6.12-rc2
  98v2.6.12-rc3
  99v2.6.12-rc4
 100v2.6.12-rc5
 101v2.6.12-rc6
 102v2.6.13
 103...
 104------------------------------------------------
 105
 106Tags are expected to always point at the same version of a project,
 107while heads are expected to advance as development progresses.
 108
 109Create a new branch head pointing to one of these versions and check it
 110out using gitlink:git-checkout[1]:
 111
 112------------------------------------------------
 113$ git checkout -b new v2.6.13
 114------------------------------------------------
 115
 116The working directory then reflects the contents that the project had
 117when it was tagged v2.6.13, and gitlink:git-branch[1] shows two
 118branches, with an asterisk marking the currently checked-out branch:
 119
 120------------------------------------------------
 121$ git branch
 122  master
 123* new
 124------------------------------------------------
 125
 126If you decide that you'd rather see version 2.6.17, you can modify
 127the current branch to point at v2.6.17 instead, with
 128
 129------------------------------------------------
 130$ git reset --hard v2.6.17
 131------------------------------------------------
 132
 133Note that if the current branch head was your only reference to a
 134particular point in history, then resetting that branch may leave you
 135with no way to find the history it used to point to; so use this command
 136carefully.
 137
 138[[understanding-commits]]
 139Understanding History: Commits
 140------------------------------
 141
 142Every change in the history of a project is represented by a commit.
 143The gitlink:git-show[1] command shows the most recent commit on the
 144current branch:
 145
 146------------------------------------------------
 147$ git show
 148commit 17cf781661e6d38f737f15f53ab552f1e95960d7
 149Author: Linus Torvalds <torvalds@ppc970.osdl.org.(none)>
 150Date:   Tue Apr 19 14:11:06 2005 -0700
 151
 152    Remove duplicate getenv(DB_ENVIRONMENT) call
 153
 154    Noted by Tony Luck.
 155
 156diff --git a/init-db.c b/init-db.c
 157index 65898fa..b002dc6 100644
 158--- a/init-db.c
 159+++ b/init-db.c
 160@@ -7,7 +7,7 @@
 161 
 162 int main(int argc, char **argv)
 163 {
 164-       char *sha1_dir = getenv(DB_ENVIRONMENT), *path;
 165+       char *sha1_dir, *path;
 166        int len, i;
 167 
 168        if (mkdir(".git", 0755) < 0) {
 169------------------------------------------------
 170
 171As you can see, a commit shows who made the latest change, what they
 172did, and why.
 173
 174Every commit has a 40-hexdigit id, sometimes called the "object name" or the
 175"SHA1 id", shown on the first line of the "git show" output.  You can usually
 176refer to a commit by a shorter name, such as a tag or a branch name, but this
 177longer name can also be useful.  Most importantly, it is a globally unique
 178name for this commit: so if you tell somebody else the object name (for
 179example in email), then you are guaranteed that name will refer to the same
 180commit in their repository that it does in yours (assuming their repository
 181has that commit at all).  Since the object name is computed as a hash over the
 182contents of the commit, you are guaranteed that the commit can never change
 183without its name also changing.
 184
 185In fact, in <<git-concepts>> we shall see that everything stored in git
 186history, including file data and directory contents, is stored in an object
 187with a name that is a hash of its contents.
 188
 189[[understanding-reachability]]
 190Understanding history: commits, parents, and reachability
 191~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 192
 193Every commit (except the very first commit in a project) also has a
 194parent commit which shows what happened before this commit.
 195Following the chain of parents will eventually take you back to the
 196beginning of the project.
 197
 198However, the commits do not form a simple list; git allows lines of
 199development to diverge and then reconverge, and the point where two
 200lines of development reconverge is called a "merge".  The commit
 201representing a merge can therefore have more than one parent, with
 202each parent representing the most recent commit on one of the lines
 203of development leading to that point.
 204
 205The best way to see how this works is using the gitlink:gitk[1]
 206command; running gitk now on a git repository and looking for merge
 207commits will help understand how the git organizes history.
 208
 209In the following, we say that commit X is "reachable" from commit Y
 210if commit X is an ancestor of commit Y.  Equivalently, you could say
 211that Y is a descendant of X, or that there is a chain of parents
 212leading from commit Y to commit X.
 213
 214[[history-diagrams]]
 215Understanding history: History diagrams
 216~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 217
 218We will sometimes represent git history using diagrams like the one
 219below.  Commits are shown as "o", and the links between them with
 220lines drawn with - / and \.  Time goes left to right:
 221
 222
 223................................................
 224         o--o--o <-- Branch A
 225        /
 226 o--o--o <-- master
 227        \
 228         o--o--o <-- Branch B
 229................................................
 230
 231If we need to talk about a particular commit, the character "o" may
 232be replaced with another letter or number.
 233
 234[[what-is-a-branch]]
 235Understanding history: What is a branch?
 236~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 237
 238When we need to be precise, we will use the word "branch" to mean a line
 239of development, and "branch head" (or just "head") to mean a reference
 240to the most recent commit on a branch.  In the example above, the branch
 241head named "A" is a pointer to one particular commit, but we refer to
 242the line of three commits leading up to that point as all being part of
 243"branch A".
 244
 245However, when no confusion will result, we often just use the term
 246"branch" both for branches and for branch heads.
 247
 248[[manipulating-branches]]
 249Manipulating branches
 250---------------------
 251
 252Creating, deleting, and modifying branches is quick and easy; here's
 253a summary of the commands:
 254
 255git branch::
 256        list all branches
 257git branch <branch>::
 258        create a new branch named <branch>, referencing the same
 259        point in history as the current branch
 260git branch <branch> <start-point>::
 261        create a new branch named <branch>, referencing
 262        <start-point>, which may be specified any way you like,
 263        including using a branch name or a tag name
 264git branch -d <branch>::
 265        delete the branch <branch>; if the branch you are deleting
 266        points to a commit which is not reachable from the current
 267        branch, this command will fail with a warning.
 268git branch -D <branch>::
 269        even if the branch points to a commit not reachable
 270        from the current branch, you may know that that commit
 271        is still reachable from some other branch or tag.  In that
 272        case it is safe to use this command to force git to delete
 273        the branch.
 274git checkout <branch>::
 275        make the current branch <branch>, updating the working
 276        directory to reflect the version referenced by <branch>
 277git checkout -b <new> <start-point>::
 278        create a new branch <new> referencing <start-point>, and
 279        check it out.
 280
 281The special symbol "HEAD" can always be used to refer to the current
 282branch.  In fact, git uses a file named "HEAD" in the .git directory to
 283remember which branch is current:
 284
 285------------------------------------------------
 286$ cat .git/HEAD
 287ref: refs/heads/master
 288------------------------------------------------
 289
 290[[detached-head]]
 291Examining an old version without creating a new branch
 292------------------------------------------------------
 293
 294The git-checkout command normally expects a branch head, but will also
 295accept an arbitrary commit; for example, you can check out the commit
 296referenced by a tag:
 297
 298------------------------------------------------
 299$ git checkout v2.6.17
 300Note: moving to "v2.6.17" which isn't a local branch
 301If you want to create a new branch from this checkout, you may do so
 302(now or later) by using -b with the checkout command again. Example:
 303  git checkout -b <new_branch_name>
 304HEAD is now at 427abfa... Linux v2.6.17
 305------------------------------------------------
 306
 307The HEAD then refers to the SHA1 of the commit instead of to a branch,
 308and git branch shows that you are no longer on a branch:
 309
 310------------------------------------------------
 311$ cat .git/HEAD
 312427abfa28afedffadfca9dd8b067eb6d36bac53f
 313$ git branch
 314* (no branch)
 315  master
 316------------------------------------------------
 317
 318In this case we say that the HEAD is "detached".
 319
 320This is an easy way to check out a particular version without having to
 321make up a name for the new branch.   You can still create a new branch
 322(or tag) for this version later if you decide to.
 323
 324[[examining-remote-branches]]
 325Examining branches from a remote repository
 326-------------------------------------------
 327
 328The "master" branch that was created at the time you cloned is a copy
 329of the HEAD in the repository that you cloned from.  That repository
 330may also have had other branches, though, and your local repository
 331keeps branches which track each of those remote branches, which you
 332can view using the "-r" option to gitlink:git-branch[1]:
 333
 334------------------------------------------------
 335$ git branch -r
 336  origin/HEAD
 337  origin/html
 338  origin/maint
 339  origin/man
 340  origin/master
 341  origin/next
 342  origin/pu
 343  origin/todo
 344------------------------------------------------
 345
 346You cannot check out these remote-tracking branches, but you can
 347examine them on a branch of your own, just as you would a tag:
 348
 349------------------------------------------------
 350$ git checkout -b my-todo-copy origin/todo
 351------------------------------------------------
 352
 353Note that the name "origin" is just the name that git uses by default
 354to refer to the repository that you cloned from.
 355
 356[[how-git-stores-references]]
 357Naming branches, tags, and other references
 358-------------------------------------------
 359
 360Branches, remote-tracking branches, and tags are all references to
 361commits.  All references are named with a slash-separated path name
 362starting with "refs"; the names we've been using so far are actually
 363shorthand:
 364
 365        - The branch "test" is short for "refs/heads/test".
 366        - The tag "v2.6.18" is short for "refs/tags/v2.6.18".
 367        - "origin/master" is short for "refs/remotes/origin/master".
 368
 369The full name is occasionally useful if, for example, there ever
 370exists a tag and a branch with the same name.
 371
 372(Newly created refs are actually stored in the .git/refs directory,
 373under the path given by their name.  However, for efficiency reasons
 374they may also be packed together in a single file; see
 375gitlink:git-pack-refs[1]).
 376
 377As another useful shortcut, the "HEAD" of a repository can be referred
 378to just using the name of that repository.  So, for example, "origin"
 379is usually a shortcut for the HEAD branch in the repository "origin".
 380
 381For the complete list of paths which git checks for references, and
 382the order it uses to decide which to choose when there are multiple
 383references with the same shorthand name, see the "SPECIFYING
 384REVISIONS" section of gitlink:git-rev-parse[1].
 385
 386[[Updating-a-repository-with-git-fetch]]
 387Updating a repository with git fetch
 388------------------------------------
 389
 390Eventually the developer cloned from will do additional work in her
 391repository, creating new commits and advancing the branches to point
 392at the new commits.
 393
 394The command "git fetch", with no arguments, will update all of the
 395remote-tracking branches to the latest version found in her
 396repository.  It will not touch any of your own branches--not even the
 397"master" branch that was created for you on clone.
 398
 399[[fetching-branches]]
 400Fetching branches from other repositories
 401-----------------------------------------
 402
 403You can also track branches from repositories other than the one you
 404cloned from, using gitlink:git-remote[1]:
 405
 406-------------------------------------------------
 407$ git remote add linux-nfs git://linux-nfs.org/pub/nfs-2.6.git
 408$ git fetch linux-nfs
 409* refs/remotes/linux-nfs/master: storing branch 'master' ...
 410  commit: bf81b46
 411-------------------------------------------------
 412
 413New remote-tracking branches will be stored under the shorthand name
 414that you gave "git remote add", in this case linux-nfs:
 415
 416-------------------------------------------------
 417$ git branch -r
 418linux-nfs/master
 419origin/master
 420-------------------------------------------------
 421
 422If you run "git fetch <remote>" later, the tracking branches for the
 423named <remote> will be updated.
 424
 425If you examine the file .git/config, you will see that git has added
 426a new stanza:
 427
 428-------------------------------------------------
 429$ cat .git/config
 430...
 431[remote "linux-nfs"]
 432        url = git://linux-nfs.org/pub/nfs-2.6.git
 433        fetch = +refs/heads/*:refs/remotes/linux-nfs/*
 434...
 435-------------------------------------------------
 436
 437This is what causes git to track the remote's branches; you may modify
 438or delete these configuration options by editing .git/config with a
 439text editor.  (See the "CONFIGURATION FILE" section of
 440gitlink:git-config[1] for details.)
 441
 442[[exploring-git-history]]
 443Exploring git history
 444=====================
 445
 446Git is best thought of as a tool for storing the history of a
 447collection of files.  It does this by storing compressed snapshots of
 448the contents of a file hierarchy, together with "commits" which show
 449the relationships between these snapshots.
 450
 451Git provides extremely flexible and fast tools for exploring the
 452history of a project.
 453
 454We start with one specialized tool that is useful for finding the
 455commit that introduced a bug into a project.
 456
 457[[using-bisect]]
 458How to use bisect to find a regression
 459--------------------------------------
 460
 461Suppose version 2.6.18 of your project worked, but the version at
 462"master" crashes.  Sometimes the best way to find the cause of such a
 463regression is to perform a brute-force search through the project's
 464history to find the particular commit that caused the problem.  The
 465gitlink:git-bisect[1] command can help you do this:
 466
 467-------------------------------------------------
 468$ git bisect start
 469$ git bisect good v2.6.18
 470$ git bisect bad master
 471Bisecting: 3537 revisions left to test after this
 472[65934a9a028b88e83e2b0f8b36618fe503349f8e] BLOCK: Make USB storage depend on SCSI rather than selecting it [try #6]
 473-------------------------------------------------
 474
 475If you run "git branch" at this point, you'll see that git has
 476temporarily moved you to a new branch named "bisect".  This branch
 477points to a commit (with commit id 65934...) that is reachable from
 478v2.6.19 but not from v2.6.18.  Compile and test it, and see whether
 479it crashes.  Assume it does crash.  Then:
 480
 481-------------------------------------------------
 482$ git bisect bad
 483Bisecting: 1769 revisions left to test after this
 484[7eff82c8b1511017ae605f0c99ac275a7e21b867] i2c-core: Drop useless bitmaskings
 485-------------------------------------------------
 486
 487checks out an older version.  Continue like this, telling git at each
 488stage whether the version it gives you is good or bad, and notice
 489that the number of revisions left to test is cut approximately in
 490half each time.
 491
 492After about 13 tests (in this case), it will output the commit id of
 493the guilty commit.  You can then examine the commit with
 494gitlink:git-show[1], find out who wrote it, and mail them your bug
 495report with the commit id.  Finally, run
 496
 497-------------------------------------------------
 498$ git bisect reset
 499-------------------------------------------------
 500
 501to return you to the branch you were on before and delete the
 502temporary "bisect" branch.
 503
 504Note that the version which git-bisect checks out for you at each
 505point is just a suggestion, and you're free to try a different
 506version if you think it would be a good idea.  For example,
 507occasionally you may land on a commit that broke something unrelated;
 508run
 509
 510-------------------------------------------------
 511$ git bisect visualize
 512-------------------------------------------------
 513
 514which will run gitk and label the commit it chose with a marker that
 515says "bisect".  Chose a safe-looking commit nearby, note its commit
 516id, and check it out with:
 517
 518-------------------------------------------------
 519$ git reset --hard fb47ddb2db...
 520-------------------------------------------------
 521
 522then test, run "bisect good" or "bisect bad" as appropriate, and
 523continue.
 524
 525[[naming-commits]]
 526Naming commits
 527--------------
 528
 529We have seen several ways of naming commits already:
 530
 531        - 40-hexdigit object name
 532        - branch name: refers to the commit at the head of the given
 533          branch
 534        - tag name: refers to the commit pointed to by the given tag
 535          (we've seen branches and tags are special cases of
 536          <<how-git-stores-references,references>>).
 537        - HEAD: refers to the head of the current branch
 538
 539There are many more; see the "SPECIFYING REVISIONS" section of the
 540gitlink:git-rev-parse[1] man page for the complete list of ways to
 541name revisions.  Some examples:
 542
 543-------------------------------------------------
 544$ git show fb47ddb2 # the first few characters of the object name
 545                    # are usually enough to specify it uniquely
 546$ git show HEAD^    # the parent of the HEAD commit
 547$ git show HEAD^^   # the grandparent
 548$ git show HEAD~4   # the great-great-grandparent
 549-------------------------------------------------
 550
 551Recall that merge commits may have more than one parent; by default,
 552^ and ~ follow the first parent listed in the commit, but you can
 553also choose:
 554
 555-------------------------------------------------
 556$ git show HEAD^1   # show the first parent of HEAD
 557$ git show HEAD^2   # show the second parent of HEAD
 558-------------------------------------------------
 559
 560In addition to HEAD, there are several other special names for
 561commits:
 562
 563Merges (to be discussed later), as well as operations such as
 564git-reset, which change the currently checked-out commit, generally
 565set ORIG_HEAD to the value HEAD had before the current operation.
 566
 567The git-fetch operation always stores the head of the last fetched
 568branch in FETCH_HEAD.  For example, if you run git fetch without
 569specifying a local branch as the target of the operation
 570
 571-------------------------------------------------
 572$ git fetch git://example.com/proj.git theirbranch
 573-------------------------------------------------
 574
 575the fetched commits will still be available from FETCH_HEAD.
 576
 577When we discuss merges we'll also see the special name MERGE_HEAD,
 578which refers to the other branch that we're merging in to the current
 579branch.
 580
 581The gitlink:git-rev-parse[1] command is a low-level command that is
 582occasionally useful for translating some name for a commit to the object
 583name for that commit:
 584
 585-------------------------------------------------
 586$ git rev-parse origin
 587e05db0fd4f31dde7005f075a84f96b360d05984b
 588-------------------------------------------------
 589
 590[[creating-tags]]
 591Creating tags
 592-------------
 593
 594We can also create a tag to refer to a particular commit; after
 595running
 596
 597-------------------------------------------------
 598$ git tag stable-1 1b2e1d63ff
 599-------------------------------------------------
 600
 601You can use stable-1 to refer to the commit 1b2e1d63ff.
 602
 603This creates a "lightweight" tag.  If you would also like to include a
 604comment with the tag, and possibly sign it cryptographically, then you
 605should create a tag object instead; see the gitlink:git-tag[1] man page
 606for details.
 607
 608[[browsing-revisions]]
 609Browsing revisions
 610------------------
 611
 612The gitlink:git-log[1] command can show lists of commits.  On its
 613own, it shows all commits reachable from the parent commit; but you
 614can also make more specific requests:
 615
 616-------------------------------------------------
 617$ git log v2.5..        # commits since (not reachable from) v2.5
 618$ git log test..master  # commits reachable from master but not test
 619$ git log master..test  # ...reachable from test but not master
 620$ git log master...test # ...reachable from either test or master,
 621                        #    but not both
 622$ git log --since="2 weeks ago" # commits from the last 2 weeks
 623$ git log Makefile      # commits which modify Makefile
 624$ git log fs/           # ... which modify any file under fs/
 625$ git log -S'foo()'     # commits which add or remove any file data
 626                        # matching the string 'foo()'
 627-------------------------------------------------
 628
 629And of course you can combine all of these; the following finds
 630commits since v2.5 which touch the Makefile or any file under fs:
 631
 632-------------------------------------------------
 633$ git log v2.5.. Makefile fs/
 634-------------------------------------------------
 635
 636You can also ask git log to show patches:
 637
 638-------------------------------------------------
 639$ git log -p
 640-------------------------------------------------
 641
 642See the "--pretty" option in the gitlink:git-log[1] man page for more
 643display options.
 644
 645Note that git log starts with the most recent commit and works
 646backwards through the parents; however, since git history can contain
 647multiple independent lines of development, the particular order that
 648commits are listed in may be somewhat arbitrary.
 649
 650[[generating-diffs]]
 651Generating diffs
 652----------------
 653
 654You can generate diffs between any two versions using
 655gitlink:git-diff[1]:
 656
 657-------------------------------------------------
 658$ git diff master..test
 659-------------------------------------------------
 660
 661Sometimes what you want instead is a set of patches:
 662
 663-------------------------------------------------
 664$ git format-patch master..test
 665-------------------------------------------------
 666
 667will generate a file with a patch for each commit reachable from test
 668but not from master.  Note that if master also has commits which are
 669not reachable from test, then the combined result of these patches
 670will not be the same as the diff produced by the git-diff example.
 671
 672[[viewing-old-file-versions]]
 673Viewing old file versions
 674-------------------------
 675
 676You can always view an old version of a file by just checking out the
 677correct revision first.  But sometimes it is more convenient to be
 678able to view an old version of a single file without checking
 679anything out; this command does that:
 680
 681-------------------------------------------------
 682$ git show v2.5:fs/locks.c
 683-------------------------------------------------
 684
 685Before the colon may be anything that names a commit, and after it
 686may be any path to a file tracked by git.
 687
 688[[history-examples]]
 689Examples
 690--------
 691
 692[[counting-commits-on-a-branch]]
 693Counting the number of commits on a branch
 694~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 695
 696Suppose you want to know how many commits you've made on "mybranch"
 697since it diverged from "origin":
 698
 699-------------------------------------------------
 700$ git log --pretty=oneline origin..mybranch | wc -l
 701-------------------------------------------------
 702
 703Alternatively, you may often see this sort of thing done with the
 704lower-level command gitlink:git-rev-list[1], which just lists the SHA1's
 705of all the given commits:
 706
 707-------------------------------------------------
 708$ git rev-list origin..mybranch | wc -l
 709-------------------------------------------------
 710
 711[[checking-for-equal-branches]]
 712Check whether two branches point at the same history
 713~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 714
 715Suppose you want to check whether two branches point at the same point
 716in history.
 717
 718-------------------------------------------------
 719$ git diff origin..master
 720-------------------------------------------------
 721
 722will tell you whether the contents of the project are the same at the
 723two branches; in theory, however, it's possible that the same project
 724contents could have been arrived at by two different historical
 725routes.  You could compare the object names:
 726
 727-------------------------------------------------
 728$ git rev-list origin
 729e05db0fd4f31dde7005f075a84f96b360d05984b
 730$ git rev-list master
 731e05db0fd4f31dde7005f075a84f96b360d05984b
 732-------------------------------------------------
 733
 734Or you could recall that the ... operator selects all commits
 735contained reachable from either one reference or the other but not
 736both: so
 737
 738-------------------------------------------------
 739$ git log origin...master
 740-------------------------------------------------
 741
 742will return no commits when the two branches are equal.
 743
 744[[finding-tagged-descendants]]
 745Find first tagged version including a given fix
 746~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 747
 748Suppose you know that the commit e05db0fd fixed a certain problem.
 749You'd like to find the earliest tagged release that contains that
 750fix.
 751
 752Of course, there may be more than one answer--if the history branched
 753after commit e05db0fd, then there could be multiple "earliest" tagged
 754releases.
 755
 756You could just visually inspect the commits since e05db0fd:
 757
 758-------------------------------------------------
 759$ gitk e05db0fd..
 760-------------------------------------------------
 761
 762Or you can use gitlink:git-name-rev[1], which will give the commit a
 763name based on any tag it finds pointing to one of the commit's
 764descendants:
 765
 766-------------------------------------------------
 767$ git name-rev --tags e05db0fd
 768e05db0fd tags/v1.5.0-rc1^0~23
 769-------------------------------------------------
 770
 771The gitlink:git-describe[1] command does the opposite, naming the
 772revision using a tag on which the given commit is based:
 773
 774-------------------------------------------------
 775$ git describe e05db0fd
 776v1.5.0-rc0-260-ge05db0f
 777-------------------------------------------------
 778
 779but that may sometimes help you guess which tags might come after the
 780given commit.
 781
 782If you just want to verify whether a given tagged version contains a
 783given commit, you could use gitlink:git-merge-base[1]:
 784
 785-------------------------------------------------
 786$ git merge-base e05db0fd v1.5.0-rc1
 787e05db0fd4f31dde7005f075a84f96b360d05984b
 788-------------------------------------------------
 789
 790The merge-base command finds a common ancestor of the given commits,
 791and always returns one or the other in the case where one is a
 792descendant of the other; so the above output shows that e05db0fd
 793actually is an ancestor of v1.5.0-rc1.
 794
 795Alternatively, note that
 796
 797-------------------------------------------------
 798$ git log v1.5.0-rc1..e05db0fd
 799-------------------------------------------------
 800
 801will produce empty output if and only if v1.5.0-rc1 includes e05db0fd,
 802because it outputs only commits that are not reachable from v1.5.0-rc1.
 803
 804As yet another alternative, the gitlink:git-show-branch[1] command lists
 805the commits reachable from its arguments with a display on the left-hand
 806side that indicates which arguments that commit is reachable from.  So,
 807you can run something like
 808
 809-------------------------------------------------
 810$ git show-branch e05db0fd v1.5.0-rc0 v1.5.0-rc1 v1.5.0-rc2
 811! [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
 812available
 813 ! [v1.5.0-rc0] GIT v1.5.0 preview
 814  ! [v1.5.0-rc1] GIT v1.5.0-rc1
 815   ! [v1.5.0-rc2] GIT v1.5.0-rc2
 816...
 817-------------------------------------------------
 818
 819then search for a line that looks like
 820
 821-------------------------------------------------
 822+ ++ [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
 823available
 824-------------------------------------------------
 825
 826Which shows that e05db0fd is reachable from itself, from v1.5.0-rc1, and
 827from v1.5.0-rc2, but not from v1.5.0-rc0.
 828
 829[[showing-commits-unique-to-a-branch]]
 830Showing commits unique to a given branch
 831~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 832
 833Suppose you would like to see all the commits reachable from the branch
 834head named "master" but not from any other head in your repository.
 835
 836We can list all the heads in this repository with
 837gitlink:git-show-ref[1]:
 838
 839-------------------------------------------------
 840$ git show-ref --heads
 841bf62196b5e363d73353a9dcf094c59595f3153b7 refs/heads/core-tutorial
 842db768d5504c1bb46f63ee9d6e1772bd047e05bf9 refs/heads/maint
 843a07157ac624b2524a059a3414e99f6f44bebc1e7 refs/heads/master
 84424dbc180ea14dc1aebe09f14c8ecf32010690627 refs/heads/tutorial-2
 8451e87486ae06626c2f31eaa63d26fc0fd646c8af2 refs/heads/tutorial-fixes
 846-------------------------------------------------
 847
 848We can get just the branch-head names, and remove "master", with
 849the help of the standard utilities cut and grep:
 850
 851-------------------------------------------------
 852$ git show-ref --heads | cut -d' ' -f2 | grep -v '^refs/heads/master'
 853refs/heads/core-tutorial
 854refs/heads/maint
 855refs/heads/tutorial-2
 856refs/heads/tutorial-fixes
 857-------------------------------------------------
 858
 859And then we can ask to see all the commits reachable from master
 860but not from these other heads:
 861
 862-------------------------------------------------
 863$ gitk master --not $( git show-ref --heads | cut -d' ' -f2 |
 864                                grep -v '^refs/heads/master' )
 865-------------------------------------------------
 866
 867Obviously, endless variations are possible; for example, to see all
 868commits reachable from some head but not from any tag in the repository:
 869
 870-------------------------------------------------
 871$ gitk $( git show-ref --heads ) --not  $( git show-ref --tags )
 872-------------------------------------------------
 873
 874(See gitlink:git-rev-parse[1] for explanations of commit-selecting
 875syntax such as `--not`.)
 876
 877[[making-a-release]]
 878Creating a changelog and tarball for a software release
 879~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 880
 881The gitlink:git-archive[1] command can create a tar or zip archive from
 882any version of a project; for example:
 883
 884-------------------------------------------------
 885$ git archive --format=tar --prefix=project/ HEAD | gzip >latest.tar.gz
 886-------------------------------------------------
 887
 888will use HEAD to produce a tar archive in which each filename is
 889preceded by "project/".
 890
 891If you're releasing a new version of a software project, you may want
 892to simultaneously make a changelog to include in the release
 893announcement.
 894
 895Linus Torvalds, for example, makes new kernel releases by tagging them,
 896then running:
 897
 898-------------------------------------------------
 899$ release-script 2.6.12 2.6.13-rc6 2.6.13-rc7
 900-------------------------------------------------
 901
 902where release-script is a shell script that looks like:
 903
 904-------------------------------------------------
 905#!/bin/sh
 906stable="$1"
 907last="$2"
 908new="$3"
 909echo "# git tag v$new"
 910echo "git archive --prefix=linux-$new/ v$new | gzip -9 > ../linux-$new.tar.gz"
 911echo "git diff v$stable v$new | gzip -9 > ../patch-$new.gz"
 912echo "git log --no-merges v$new ^v$last > ../ChangeLog-$new"
 913echo "git shortlog --no-merges v$new ^v$last > ../ShortLog"
 914echo "git diff --stat --summary -M v$last v$new > ../diffstat-$new"
 915-------------------------------------------------
 916
 917and then he just cut-and-pastes the output commands after verifying that
 918they look OK.
 919
 920[[Finding-comments-with-given-content]]
 921Finding commits referencing a file with given content
 922~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 923
 924Somebody hands you a copy of a file, and asks which commits modified a
 925file such that it contained the given content either before or after the
 926commit.  You can find out with this:
 927
 928-------------------------------------------------
 929$  git log --raw --abbrev=40 --pretty=oneline -- filename |
 930        grep -B 1 `git hash-object filename`
 931-------------------------------------------------
 932
 933Figuring out why this works is left as an exercise to the (advanced)
 934student.  The gitlink:git-log[1], gitlink:git-diff-tree[1], and
 935gitlink:git-hash-object[1] man pages may prove helpful.
 936
 937[[Developing-with-git]]
 938Developing with git
 939===================
 940
 941[[telling-git-your-name]]
 942Telling git your name
 943---------------------
 944
 945Before creating any commits, you should introduce yourself to git.  The
 946easiest way to do so is to make sure the following lines appear in a
 947file named .gitconfig in your home directory:
 948
 949------------------------------------------------
 950[user]
 951        name = Your Name Comes Here
 952        email = you@yourdomain.example.com
 953------------------------------------------------
 954
 955(See the "CONFIGURATION FILE" section of gitlink:git-config[1] for
 956details on the configuration file.)
 957
 958
 959[[creating-a-new-repository]]
 960Creating a new repository
 961-------------------------
 962
 963Creating a new repository from scratch is very easy:
 964
 965-------------------------------------------------
 966$ mkdir project
 967$ cd project
 968$ git init
 969-------------------------------------------------
 970
 971If you have some initial content (say, a tarball):
 972
 973-------------------------------------------------
 974$ tar -xzvf project.tar.gz
 975$ cd project
 976$ git init
 977$ git add . # include everything below ./ in the first commit:
 978$ git commit
 979-------------------------------------------------
 980
 981[[how-to-make-a-commit]]
 982How to make a commit
 983--------------------
 984
 985Creating a new commit takes three steps:
 986
 987        1. Making some changes to the working directory using your
 988           favorite editor.
 989        2. Telling git about your changes.
 990        3. Creating the commit using the content you told git about
 991           in step 2.
 992
 993In practice, you can interleave and repeat steps 1 and 2 as many
 994times as you want: in order to keep track of what you want committed
 995at step 3, git maintains a snapshot of the tree's contents in a
 996special staging area called "the index."
 997
 998At the beginning, the content of the index will be identical to
 999that of the HEAD.  The command "git diff --cached", which shows
1000the difference between the HEAD and the index, should therefore
1001produce no output at that point.
1002
1003Modifying the index is easy:
1004
1005To update the index with the new contents of a modified file, use
1006
1007-------------------------------------------------
1008$ git add path/to/file
1009-------------------------------------------------
1010
1011To add the contents of a new file to the index, use
1012
1013-------------------------------------------------
1014$ git add path/to/file
1015-------------------------------------------------
1016
1017To remove a file from the index and from the working tree,
1018
1019-------------------------------------------------
1020$ git rm path/to/file
1021-------------------------------------------------
1022
1023After each step you can verify that
1024
1025-------------------------------------------------
1026$ git diff --cached
1027-------------------------------------------------
1028
1029always shows the difference between the HEAD and the index file--this
1030is what you'd commit if you created the commit now--and that
1031
1032-------------------------------------------------
1033$ git diff
1034-------------------------------------------------
1035
1036shows the difference between the working tree and the index file.
1037
1038Note that "git add" always adds just the current contents of a file
1039to the index; further changes to the same file will be ignored unless
1040you run git-add on the file again.
1041
1042When you're ready, just run
1043
1044-------------------------------------------------
1045$ git commit
1046-------------------------------------------------
1047
1048and git will prompt you for a commit message and then create the new
1049commit.  Check to make sure it looks like what you expected with
1050
1051-------------------------------------------------
1052$ git show
1053-------------------------------------------------
1054
1055As a special shortcut,
1056
1057-------------------------------------------------
1058$ git commit -a
1059-------------------------------------------------
1060
1061will update the index with any files that you've modified or removed
1062and create a commit, all in one step.
1063
1064A number of commands are useful for keeping track of what you're
1065about to commit:
1066
1067-------------------------------------------------
1068$ git diff --cached # difference between HEAD and the index; what
1069                    # would be committed if you ran "commit" now.
1070$ git diff          # difference between the index file and your
1071                    # working directory; changes that would not
1072                    # be included if you ran "commit" now.
1073$ git diff HEAD     # difference between HEAD and working tree; what
1074                    # would be committed if you ran "commit -a" now.
1075$ git status        # a brief per-file summary of the above.
1076-------------------------------------------------
1077
1078You can also use gitlink:git-gui[1] to create commits, view changes in
1079the index and the working tree files, and individually select diff hunks
1080for inclusion in the index (by right-clicking on the diff hunk and
1081choosing "Stage Hunk For Commit").
1082
1083[[creating-good-commit-messages]]
1084Creating good commit messages
1085-----------------------------
1086
1087Though not required, it's a good idea to begin the commit message
1088with a single short (less than 50 character) line summarizing the
1089change, followed by a blank line and then a more thorough
1090description.  Tools that turn commits into email, for example, use
1091the first line on the Subject line and the rest of the commit in the
1092body.
1093
1094[[ignoring-files]]
1095Ignoring files
1096--------------
1097
1098A project will often generate files that you do 'not' want to track with git.
1099This typically includes files generated by a build process or temporary
1100backup files made by your editor. Of course, 'not' tracking files with git
1101is just a matter of 'not' calling "`git add`" on them. But it quickly becomes
1102annoying to have these untracked files lying around; e.g. they make
1103"`git add .`" and "`git commit -a`" practically useless, and they keep
1104showing up in the output of "`git status`".
1105
1106You can tell git to ignore certain files by creating a file called .gitignore
1107in the top level of your working directory, with contents such as:
1108
1109-------------------------------------------------
1110# Lines starting with '#' are considered comments.
1111# Ignore any file named foo.txt.
1112foo.txt
1113# Ignore (generated) html files,
1114*.html
1115# except foo.html which is maintained by hand.
1116!foo.html
1117# Ignore objects and archives.
1118*.[oa]
1119-------------------------------------------------
1120
1121See gitlink:gitignore[5] for a detailed explanation of the syntax.  You can
1122also place .gitignore files in other directories in your working tree, and they
1123will apply to those directories and their subdirectories.  The `.gitignore`
1124files can be added to your repository like any other files (just run `git add
1125.gitignore` and `git commit`, as usual), which is convenient when the exclude
1126patterns (such as patterns matching build output files) would also make sense
1127for other users who clone your repository.
1128
1129If you wish the exclude patterns to affect only certain repositories
1130(instead of every repository for a given project), you may instead put
1131them in a file in your repository named .git/info/exclude, or in any file
1132specified by the `core.excludesfile` configuration variable.  Some git
1133commands can also take exclude patterns directly on the command line.
1134See gitlink:gitignore[5] for the details.
1135
1136[[how-to-merge]]
1137How to merge
1138------------
1139
1140You can rejoin two diverging branches of development using
1141gitlink:git-merge[1]:
1142
1143-------------------------------------------------
1144$ git merge branchname
1145-------------------------------------------------
1146
1147merges the development in the branch "branchname" into the current
1148branch.  If there are conflicts--for example, if the same file is
1149modified in two different ways in the remote branch and the local
1150branch--then you are warned; the output may look something like this:
1151
1152-------------------------------------------------
1153$ git merge next
1154 100% (4/4) done
1155Auto-merged file.txt
1156CONFLICT (content): Merge conflict in file.txt
1157Automatic merge failed; fix conflicts and then commit the result.
1158-------------------------------------------------
1159
1160Conflict markers are left in the problematic files, and after
1161you resolve the conflicts manually, you can update the index
1162with the contents and run git commit, as you normally would when
1163creating a new file.
1164
1165If you examine the resulting commit using gitk, you will see that it
1166has two parents, one pointing to the top of the current branch, and
1167one to the top of the other branch.
1168
1169[[resolving-a-merge]]
1170Resolving a merge
1171-----------------
1172
1173When a merge isn't resolved automatically, git leaves the index and
1174the working tree in a special state that gives you all the
1175information you need to help resolve the merge.
1176
1177Files with conflicts are marked specially in the index, so until you
1178resolve the problem and update the index, gitlink:git-commit[1] will
1179fail:
1180
1181-------------------------------------------------
1182$ git commit
1183file.txt: needs merge
1184-------------------------------------------------
1185
1186Also, gitlink:git-status[1] will list those files as "unmerged", and the
1187files with conflicts will have conflict markers added, like this:
1188
1189-------------------------------------------------
1190<<<<<<< HEAD:file.txt
1191Hello world
1192=======
1193Goodbye
1194>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1195-------------------------------------------------
1196
1197All you need to do is edit the files to resolve the conflicts, and then
1198
1199-------------------------------------------------
1200$ git add file.txt
1201$ git commit
1202-------------------------------------------------
1203
1204Note that the commit message will already be filled in for you with
1205some information about the merge.  Normally you can just use this
1206default message unchanged, but you may add additional commentary of
1207your own if desired.
1208
1209The above is all you need to know to resolve a simple merge.  But git
1210also provides more information to help resolve conflicts:
1211
1212[[conflict-resolution]]
1213Getting conflict-resolution help during a merge
1214~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1215
1216All of the changes that git was able to merge automatically are
1217already added to the index file, so gitlink:git-diff[1] shows only
1218the conflicts.  It uses an unusual syntax:
1219
1220-------------------------------------------------
1221$ git diff
1222diff --cc file.txt
1223index 802992c,2b60207..0000000
1224--- a/file.txt
1225+++ b/file.txt
1226@@@ -1,1 -1,1 +1,5 @@@
1227++<<<<<<< HEAD:file.txt
1228 +Hello world
1229++=======
1230+ Goodbye
1231++>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1232-------------------------------------------------
1233
1234Recall that the commit which will be committed after we resolve this
1235conflict will have two parents instead of the usual one: one parent
1236will be HEAD, the tip of the current branch; the other will be the
1237tip of the other branch, which is stored temporarily in MERGE_HEAD.
1238
1239During the merge, the index holds three versions of each file.  Each of
1240these three "file stages" represents a different version of the file:
1241
1242-------------------------------------------------
1243$ git show :1:file.txt  # the file in a common ancestor of both branches
1244$ git show :2:file.txt  # the version from HEAD, but including any
1245                        # nonconflicting changes from MERGE_HEAD
1246$ git show :3:file.txt  # the version from MERGE_HEAD, but including any
1247                        # nonconflicting changes from HEAD.
1248-------------------------------------------------
1249
1250Since the stage 2 and stage 3 versions have already been updated with
1251nonconflicting changes, the only remaining differences between them are
1252the important ones; thus gitlink:git-diff[1] can use the information in
1253the index to show only those conflicts.
1254
1255The diff above shows the differences between the working-tree version of
1256file.txt and the stage 2 and stage 3 versions.  So instead of preceding
1257each line by a single "+" or "-", it now uses two columns: the first
1258column is used for differences between the first parent and the working
1259directory copy, and the second for differences between the second parent
1260and the working directory copy.  (See the "COMBINED DIFF FORMAT" section
1261of gitlink:git-diff-files[1] for a details of the format.)
1262
1263After resolving the conflict in the obvious way (but before updating the
1264index), the diff will look like:
1265
1266-------------------------------------------------
1267$ git diff
1268diff --cc file.txt
1269index 802992c,2b60207..0000000
1270--- a/file.txt
1271+++ b/file.txt
1272@@@ -1,1 -1,1 +1,1 @@@
1273- Hello world
1274 -Goodbye
1275++Goodbye world
1276-------------------------------------------------
1277
1278This shows that our resolved version deleted "Hello world" from the
1279first parent, deleted "Goodbye" from the second parent, and added
1280"Goodbye world", which was previously absent from both.
1281
1282Some special diff options allow diffing the working directory against
1283any of these stages:
1284
1285-------------------------------------------------
1286$ git diff -1 file.txt          # diff against stage 1
1287$ git diff --base file.txt      # same as the above
1288$ git diff -2 file.txt          # diff against stage 2
1289$ git diff --ours file.txt      # same as the above
1290$ git diff -3 file.txt          # diff against stage 3
1291$ git diff --theirs file.txt    # same as the above.
1292-------------------------------------------------
1293
1294The gitlink:git-log[1] and gitk[1] commands also provide special help
1295for merges:
1296
1297-------------------------------------------------
1298$ git log --merge
1299$ gitk --merge
1300-------------------------------------------------
1301
1302These will display all commits which exist only on HEAD or on
1303MERGE_HEAD, and which touch an unmerged file.
1304
1305You may also use gitlink:git-mergetool[1], which lets you merge the
1306unmerged files using external tools such as emacs or kdiff3.
1307
1308Each time you resolve the conflicts in a file and update the index:
1309
1310-------------------------------------------------
1311$ git add file.txt
1312-------------------------------------------------
1313
1314the different stages of that file will be "collapsed", after which
1315git-diff will (by default) no longer show diffs for that file.
1316
1317[[undoing-a-merge]]
1318Undoing a merge
1319---------------
1320
1321If you get stuck and decide to just give up and throw the whole mess
1322away, you can always return to the pre-merge state with
1323
1324-------------------------------------------------
1325$ git reset --hard HEAD
1326-------------------------------------------------
1327
1328Or, if you've already committed the merge that you want to throw away,
1329
1330-------------------------------------------------
1331$ git reset --hard ORIG_HEAD
1332-------------------------------------------------
1333
1334However, this last command can be dangerous in some cases--never
1335throw away a commit you have already committed if that commit may
1336itself have been merged into another branch, as doing so may confuse
1337further merges.
1338
1339[[fast-forwards]]
1340Fast-forward merges
1341-------------------
1342
1343There is one special case not mentioned above, which is treated
1344differently.  Normally, a merge results in a merge commit, with two
1345parents, one pointing at each of the two lines of development that
1346were merged.
1347
1348However, if the current branch is a descendant of the other--so every
1349commit present in the one is already contained in the other--then git
1350just performs a "fast forward"; the head of the current branch is moved
1351forward to point at the head of the merged-in branch, without any new
1352commits being created.
1353
1354[[fixing-mistakes]]
1355Fixing mistakes
1356---------------
1357
1358If you've messed up the working tree, but haven't yet committed your
1359mistake, you can return the entire working tree to the last committed
1360state with
1361
1362-------------------------------------------------
1363$ git reset --hard HEAD
1364-------------------------------------------------
1365
1366If you make a commit that you later wish you hadn't, there are two
1367fundamentally different ways to fix the problem:
1368
1369        1. You can create a new commit that undoes whatever was done
1370        by the previous commit.  This is the correct thing if your
1371        mistake has already been made public.
1372
1373        2. You can go back and modify the old commit.  You should
1374        never do this if you have already made the history public;
1375        git does not normally expect the "history" of a project to
1376        change, and cannot correctly perform repeated merges from
1377        a branch that has had its history changed.
1378
1379[[reverting-a-commit]]
1380Fixing a mistake with a new commit
1381~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1382
1383Creating a new commit that reverts an earlier change is very easy;
1384just pass the gitlink:git-revert[1] command a reference to the bad
1385commit; for example, to revert the most recent commit:
1386
1387-------------------------------------------------
1388$ git revert HEAD
1389-------------------------------------------------
1390
1391This will create a new commit which undoes the change in HEAD.  You
1392will be given a chance to edit the commit message for the new commit.
1393
1394You can also revert an earlier change, for example, the next-to-last:
1395
1396-------------------------------------------------
1397$ git revert HEAD^
1398-------------------------------------------------
1399
1400In this case git will attempt to undo the old change while leaving
1401intact any changes made since then.  If more recent changes overlap
1402with the changes to be reverted, then you will be asked to fix
1403conflicts manually, just as in the case of <<resolving-a-merge,
1404resolving a merge>>.
1405
1406[[fixing-a-mistake-by-editing-history]]
1407Fixing a mistake by editing history
1408~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1409
1410If the problematic commit is the most recent commit, and you have not
1411yet made that commit public, then you may just
1412<<undoing-a-merge,destroy it using git-reset>>.
1413
1414Alternatively, you
1415can edit the working directory and update the index to fix your
1416mistake, just as if you were going to <<how-to-make-a-commit,create a
1417new commit>>, then run
1418
1419-------------------------------------------------
1420$ git commit --amend
1421-------------------------------------------------
1422
1423which will replace the old commit by a new commit incorporating your
1424changes, giving you a chance to edit the old commit message first.
1425
1426Again, you should never do this to a commit that may already have
1427been merged into another branch; use gitlink:git-revert[1] instead in
1428that case.
1429
1430It is also possible to edit commits further back in the history, but
1431this is an advanced topic to be left for
1432<<cleaning-up-history,another chapter>>.
1433
1434[[checkout-of-path]]
1435Checking out an old version of a file
1436~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1437
1438In the process of undoing a previous bad change, you may find it
1439useful to check out an older version of a particular file using
1440gitlink:git-checkout[1].  We've used git checkout before to switch
1441branches, but it has quite different behavior if it is given a path
1442name: the command
1443
1444-------------------------------------------------
1445$ git checkout HEAD^ path/to/file
1446-------------------------------------------------
1447
1448replaces path/to/file by the contents it had in the commit HEAD^, and
1449also updates the index to match.  It does not change branches.
1450
1451If you just want to look at an old version of the file, without
1452modifying the working directory, you can do that with
1453gitlink:git-show[1]:
1454
1455-------------------------------------------------
1456$ git show HEAD^:path/to/file
1457-------------------------------------------------
1458
1459which will display the given version of the file.
1460
1461[[interrupted-work]]
1462Temporarily setting aside work in progress
1463~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1464
1465While you are in the middle of working on something complicated, you
1466find an unrelated but obvious and trivial bug.  You would like to fix it
1467before continuing.  You can use gitlink:git-stash[1] to save the current
1468state of your work, and after fixing the bug (or, optionally after doing
1469so on a different branch and then coming back), unstash the
1470work-in-progress changes.
1471
1472------------------------------------------------
1473$ git stash "work in progress for foo feature"
1474------------------------------------------------
1475
1476This command will save your changes away to the `stash`, and
1477reset your working tree and the index to match the tip of your
1478current branch.  Then you can make your fix as usual.
1479
1480------------------------------------------------
1481... edit and test ...
1482$ git commit -a -m "blorpl: typofix"
1483------------------------------------------------
1484
1485After that, you can go back to what you were working on with
1486`git stash apply`:
1487
1488------------------------------------------------
1489$ git stash apply
1490------------------------------------------------
1491
1492
1493[[ensuring-good-performance]]
1494Ensuring good performance
1495-------------------------
1496
1497On large repositories, git depends on compression to keep the history
1498information from taking up to much space on disk or in memory.
1499
1500This compression is not performed automatically.  Therefore you
1501should occasionally run gitlink:git-gc[1]:
1502
1503-------------------------------------------------
1504$ git gc
1505-------------------------------------------------
1506
1507to recompress the archive.  This can be very time-consuming, so
1508you may prefer to run git-gc when you are not doing other work.
1509
1510
1511[[ensuring-reliability]]
1512Ensuring reliability
1513--------------------
1514
1515[[checking-for-corruption]]
1516Checking the repository for corruption
1517~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1518
1519The gitlink:git-fsck[1] command runs a number of self-consistency checks
1520on the repository, and reports on any problems.  This may take some
1521time.  The most common warning by far is about "dangling" objects:
1522
1523-------------------------------------------------
1524$ git fsck
1525dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1526dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1527dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1528dangling blob 218761f9d90712d37a9c5e36f406f92202db07eb
1529dangling commit bf093535a34a4d35731aa2bd90fe6b176302f14f
1530dangling commit 8e4bec7f2ddaa268bef999853c25755452100f8e
1531dangling tree d50bb86186bf27b681d25af89d3b5b68382e4085
1532dangling tree b24c2473f1fd3d91352a624795be026d64c8841f
1533...
1534-------------------------------------------------
1535
1536Dangling objects are not a problem.  At worst they may take up a little
1537extra disk space.  They can sometimes provide a last-resort method for
1538recovering lost work--see <<dangling-objects>> for details.  However, if
1539you wish, you can remove them with gitlink:git-prune[1] or the --prune
1540option to gitlink:git-gc[1]:
1541
1542-------------------------------------------------
1543$ git gc --prune
1544-------------------------------------------------
1545
1546This may be time-consuming.  Unlike most other git operations (including
1547git-gc when run without any options), it is not safe to prune while
1548other git operations are in progress in the same repository.
1549
1550[[recovering-lost-changes]]
1551Recovering lost changes
1552~~~~~~~~~~~~~~~~~~~~~~~
1553
1554[[reflogs]]
1555Reflogs
1556^^^^^^^
1557
1558Say you modify a branch with gitlink:git-reset[1] --hard, and then
1559realize that the branch was the only reference you had to that point in
1560history.
1561
1562Fortunately, git also keeps a log, called a "reflog", of all the
1563previous values of each branch.  So in this case you can still find the
1564old history using, for example,
1565
1566-------------------------------------------------
1567$ git log master@{1}
1568-------------------------------------------------
1569
1570This lists the commits reachable from the previous version of the head.
1571This syntax can be used to with any git command that accepts a commit,
1572not just with git log.  Some other examples:
1573
1574-------------------------------------------------
1575$ git show master@{2}           # See where the branch pointed 2,
1576$ git show master@{3}           # 3, ... changes ago.
1577$ gitk master@{yesterday}       # See where it pointed yesterday,
1578$ gitk master@{"1 week ago"}    # ... or last week
1579$ git log --walk-reflogs master # show reflog entries for master
1580-------------------------------------------------
1581
1582A separate reflog is kept for the HEAD, so
1583
1584-------------------------------------------------
1585$ git show HEAD@{"1 week ago"}
1586-------------------------------------------------
1587
1588will show what HEAD pointed to one week ago, not what the current branch
1589pointed to one week ago.  This allows you to see the history of what
1590you've checked out.
1591
1592The reflogs are kept by default for 30 days, after which they may be
1593pruned.  See gitlink:git-reflog[1] and gitlink:git-gc[1] to learn
1594how to control this pruning, and see the "SPECIFYING REVISIONS"
1595section of gitlink:git-rev-parse[1] for details.
1596
1597Note that the reflog history is very different from normal git history.
1598While normal history is shared by every repository that works on the
1599same project, the reflog history is not shared: it tells you only about
1600how the branches in your local repository have changed over time.
1601
1602[[dangling-object-recovery]]
1603Examining dangling objects
1604^^^^^^^^^^^^^^^^^^^^^^^^^^
1605
1606In some situations the reflog may not be able to save you.  For example,
1607suppose you delete a branch, then realize you need the history it
1608contained.  The reflog is also deleted; however, if you have not yet
1609pruned the repository, then you may still be able to find the lost
1610commits in the dangling objects that git-fsck reports.  See
1611<<dangling-objects>> for the details.
1612
1613-------------------------------------------------
1614$ git fsck
1615dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1616dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1617dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1618...
1619-------------------------------------------------
1620
1621You can examine
1622one of those dangling commits with, for example,
1623
1624------------------------------------------------
1625$ gitk 7281251ddd --not --all
1626------------------------------------------------
1627
1628which does what it sounds like: it says that you want to see the commit
1629history that is described by the dangling commit(s), but not the
1630history that is described by all your existing branches and tags.  Thus
1631you get exactly the history reachable from that commit that is lost.
1632(And notice that it might not be just one commit: we only report the
1633"tip of the line" as being dangling, but there might be a whole deep
1634and complex commit history that was dropped.)
1635
1636If you decide you want the history back, you can always create a new
1637reference pointing to it, for example, a new branch:
1638
1639------------------------------------------------
1640$ git branch recovered-branch 7281251ddd
1641------------------------------------------------
1642
1643Other types of dangling objects (blobs and trees) are also possible, and
1644dangling objects can arise in other situations.
1645
1646
1647[[sharing-development]]
1648Sharing development with others
1649===============================
1650
1651[[getting-updates-with-git-pull]]
1652Getting updates with git pull
1653-----------------------------
1654
1655After you clone a repository and make a few changes of your own, you
1656may wish to check the original repository for updates and merge them
1657into your own work.
1658
1659We have already seen <<Updating-a-repository-with-git-fetch,how to
1660keep remote tracking branches up to date>> with gitlink:git-fetch[1],
1661and how to merge two branches.  So you can merge in changes from the
1662original repository's master branch with:
1663
1664-------------------------------------------------
1665$ git fetch
1666$ git merge origin/master
1667-------------------------------------------------
1668
1669However, the gitlink:git-pull[1] command provides a way to do this in
1670one step:
1671
1672-------------------------------------------------
1673$ git pull origin master
1674-------------------------------------------------
1675
1676In fact, if you have "master" checked out, then by default "git pull"
1677merges from the HEAD branch of the origin repository.  So often you can
1678accomplish the above with just a simple
1679
1680-------------------------------------------------
1681$ git pull
1682-------------------------------------------------
1683
1684More generally, a branch that is created from a remote branch will pull
1685by default from that branch.  See the descriptions of the
1686branch.<name>.remote and branch.<name>.merge options in
1687gitlink:git-config[1], and the discussion of the --track option in
1688gitlink:git-checkout[1], to learn how to control these defaults.
1689
1690In addition to saving you keystrokes, "git pull" also helps you by
1691producing a default commit message documenting the branch and
1692repository that you pulled from.
1693
1694(But note that no such commit will be created in the case of a
1695<<fast-forwards,fast forward>>; instead, your branch will just be
1696updated to point to the latest commit from the upstream branch.)
1697
1698The git-pull command can also be given "." as the "remote" repository,
1699in which case it just merges in a branch from the current repository; so
1700the commands
1701
1702-------------------------------------------------
1703$ git pull . branch
1704$ git merge branch
1705-------------------------------------------------
1706
1707are roughly equivalent.  The former is actually very commonly used.
1708
1709[[submitting-patches]]
1710Submitting patches to a project
1711-------------------------------
1712
1713If you just have a few changes, the simplest way to submit them may
1714just be to send them as patches in email:
1715
1716First, use gitlink:git-format-patch[1]; for example:
1717
1718-------------------------------------------------
1719$ git format-patch origin
1720-------------------------------------------------
1721
1722will produce a numbered series of files in the current directory, one
1723for each patch in the current branch but not in origin/HEAD.
1724
1725You can then import these into your mail client and send them by
1726hand.  However, if you have a lot to send at once, you may prefer to
1727use the gitlink:git-send-email[1] script to automate the process.
1728Consult the mailing list for your project first to determine how they
1729prefer such patches be handled.
1730
1731[[importing-patches]]
1732Importing patches to a project
1733------------------------------
1734
1735Git also provides a tool called gitlink:git-am[1] (am stands for
1736"apply mailbox"), for importing such an emailed series of patches.
1737Just save all of the patch-containing messages, in order, into a
1738single mailbox file, say "patches.mbox", then run
1739
1740-------------------------------------------------
1741$ git am -3 patches.mbox
1742-------------------------------------------------
1743
1744Git will apply each patch in order; if any conflicts are found, it
1745will stop, and you can fix the conflicts as described in
1746"<<resolving-a-merge,Resolving a merge>>".  (The "-3" option tells
1747git to perform a merge; if you would prefer it just to abort and
1748leave your tree and index untouched, you may omit that option.)
1749
1750Once the index is updated with the results of the conflict
1751resolution, instead of creating a new commit, just run
1752
1753-------------------------------------------------
1754$ git am --resolved
1755-------------------------------------------------
1756
1757and git will create the commit for you and continue applying the
1758remaining patches from the mailbox.
1759
1760The final result will be a series of commits, one for each patch in
1761the original mailbox, with authorship and commit log message each
1762taken from the message containing each patch.
1763
1764[[public-repositories]]
1765Public git repositories
1766-----------------------
1767
1768Another way to submit changes to a project is to tell the maintainer
1769of that project to pull the changes from your repository using
1770gitlink:git-pull[1].  In the section "<<getting-updates-with-git-pull,
1771Getting updates with git pull>>" we described this as a way to get
1772updates from the "main" repository, but it works just as well in the
1773other direction.
1774
1775If you and the maintainer both have accounts on the same machine, then
1776you can just pull changes from each other's repositories directly;
1777commands that accept repository URLs as arguments will also accept a
1778local directory name:
1779
1780-------------------------------------------------
1781$ git clone /path/to/repository
1782$ git pull /path/to/other/repository
1783-------------------------------------------------
1784
1785or an ssh url:
1786
1787-------------------------------------------------
1788$ git clone ssh://yourhost/~you/repository
1789-------------------------------------------------
1790
1791For projects with few developers, or for synchronizing a few private
1792repositories, this may be all you need.
1793
1794However, the more common way to do this is to maintain a separate public
1795repository (usually on a different host) for others to pull changes
1796from.  This is usually more convenient, and allows you to cleanly
1797separate private work in progress from publicly visible work.
1798
1799You will continue to do your day-to-day work in your personal
1800repository, but periodically "push" changes from your personal
1801repository into your public repository, allowing other developers to
1802pull from that repository.  So the flow of changes, in a situation
1803where there is one other developer with a public repository, looks
1804like this:
1805
1806                        you push
1807  your personal repo ------------------> your public repo
1808        ^                                     |
1809        |                                     |
1810        | you pull                            | they pull
1811        |                                     |
1812        |                                     |
1813        |               they push             V
1814  their public repo <------------------- their repo
1815
1816We explain how to do this in the following sections.
1817
1818[[setting-up-a-public-repository]]
1819Setting up a public repository
1820~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1821
1822Assume your personal repository is in the directory ~/proj.  We
1823first create a new clone of the repository and tell git-daemon that it
1824is meant to be public:
1825
1826-------------------------------------------------
1827$ git clone --bare ~/proj proj.git
1828$ touch proj.git/git-daemon-export-ok
1829-------------------------------------------------
1830
1831The resulting directory proj.git contains a "bare" git repository--it is
1832just the contents of the ".git" directory, without any files checked out
1833around it.
1834
1835Next, copy proj.git to the server where you plan to host the
1836public repository.  You can use scp, rsync, or whatever is most
1837convenient.
1838
1839[[exporting-via-git]]
1840Exporting a git repository via the git protocol
1841~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1842
1843This is the preferred method.
1844
1845If someone else administers the server, they should tell you what
1846directory to put the repository in, and what git:// url it will appear
1847at.  You can then skip to the section
1848"<<pushing-changes-to-a-public-repository,Pushing changes to a public
1849repository>>", below.
1850
1851Otherwise, all you need to do is start gitlink:git-daemon[1]; it will
1852listen on port 9418.  By default, it will allow access to any directory
1853that looks like a git directory and contains the magic file
1854git-daemon-export-ok.  Passing some directory paths as git-daemon
1855arguments will further restrict the exports to those paths.
1856
1857You can also run git-daemon as an inetd service; see the
1858gitlink:git-daemon[1] man page for details.  (See especially the
1859examples section.)
1860
1861[[exporting-via-http]]
1862Exporting a git repository via http
1863~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1864
1865The git protocol gives better performance and reliability, but on a
1866host with a web server set up, http exports may be simpler to set up.
1867
1868All you need to do is place the newly created bare git repository in
1869a directory that is exported by the web server, and make some
1870adjustments to give web clients some extra information they need:
1871
1872-------------------------------------------------
1873$ mv proj.git /home/you/public_html/proj.git
1874$ cd proj.git
1875$ git --bare update-server-info
1876$ chmod a+x hooks/post-update
1877-------------------------------------------------
1878
1879(For an explanation of the last two lines, see
1880gitlink:git-update-server-info[1], and the documentation
1881link:hooks.html[Hooks used by git].)
1882
1883Advertise the url of proj.git.  Anybody else should then be able to
1884clone or pull from that url, for example with a command line like:
1885
1886-------------------------------------------------
1887$ git clone http://yourserver.com/~you/proj.git
1888-------------------------------------------------
1889
1890(See also
1891link:howto/setup-git-server-over-http.txt[setup-git-server-over-http]
1892for a slightly more sophisticated setup using WebDAV which also
1893allows pushing over http.)
1894
1895[[pushing-changes-to-a-public-repository]]
1896Pushing changes to a public repository
1897~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1898
1899Note that the two techniques outlined above (exporting via
1900<<exporting-via-http,http>> or <<exporting-via-git,git>>) allow other
1901maintainers to fetch your latest changes, but they do not allow write
1902access, which you will need to update the public repository with the
1903latest changes created in your private repository.
1904
1905The simplest way to do this is using gitlink:git-push[1] and ssh; to
1906update the remote branch named "master" with the latest state of your
1907branch named "master", run
1908
1909-------------------------------------------------
1910$ git push ssh://yourserver.com/~you/proj.git master:master
1911-------------------------------------------------
1912
1913or just
1914
1915-------------------------------------------------
1916$ git push ssh://yourserver.com/~you/proj.git master
1917-------------------------------------------------
1918
1919As with git-fetch, git-push will complain if this does not result in
1920a <<fast-forwards,fast forward>>.  Normally this is a sign of
1921something wrong.  However, if you are sure you know what you're
1922doing, you may force git-push to perform the update anyway by
1923proceeding the branch name by a plus sign:
1924
1925-------------------------------------------------
1926$ git push ssh://yourserver.com/~you/proj.git +master
1927-------------------------------------------------
1928
1929Note that the target of a "push" is normally a
1930<<def_bare_repository,bare>> repository.  You can also push to a
1931repository that has a checked-out working tree, but the working tree
1932will not be updated by the push.  This may lead to unexpected results if
1933the branch you push to is the currently checked-out branch!
1934
1935As with git-fetch, you may also set up configuration options to
1936save typing; so, for example, after
1937
1938-------------------------------------------------
1939$ cat >>.git/config <<EOF
1940[remote "public-repo"]
1941        url = ssh://yourserver.com/~you/proj.git
1942EOF
1943-------------------------------------------------
1944
1945you should be able to perform the above push with just
1946
1947-------------------------------------------------
1948$ git push public-repo master
1949-------------------------------------------------
1950
1951See the explanations of the remote.<name>.url, branch.<name>.remote,
1952and remote.<name>.push options in gitlink:git-config[1] for
1953details.
1954
1955[[setting-up-a-shared-repository]]
1956Setting up a shared repository
1957~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1958
1959Another way to collaborate is by using a model similar to that
1960commonly used in CVS, where several developers with special rights
1961all push to and pull from a single shared repository.  See
1962link:cvs-migration.html[git for CVS users] for instructions on how to
1963set this up.
1964
1965However, while there is nothing wrong with git's support for shared
1966repositories, this mode of operation is not generally recommended,
1967simply because the mode of collaboration that git supports--by
1968exchanging patches and pulling from public repositories--has so many
1969advantages over the central shared repository:
1970
1971        - Git's ability to quickly import and merge patches allows a
1972          single maintainer to process incoming changes even at very
1973          high rates.  And when that becomes too much, git-pull provides
1974          an easy way for that maintainer to delegate this job to other
1975          maintainers while still allowing optional review of incoming
1976          changes.
1977        - Since every developer's repository has the same complete copy
1978          of the project history, no repository is special, and it is
1979          trivial for another developer to take over maintenance of a
1980          project, either by mutual agreement, or because a maintainer
1981          becomes unresponsive or difficult to work with.
1982        - The lack of a central group of "committers" means there is
1983          less need for formal decisions about who is "in" and who is
1984          "out".
1985
1986[[setting-up-gitweb]]
1987Allowing web browsing of a repository
1988~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1989
1990The gitweb cgi script provides users an easy way to browse your
1991project's files and history without having to install git; see the file
1992gitweb/INSTALL in the git source tree for instructions on setting it up.
1993
1994[[sharing-development-examples]]
1995Examples
1996--------
1997
1998[[maintaining-topic-branches]]
1999Maintaining topic branches for a Linux subsystem maintainer
2000~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2001
2002This describes how Tony Luck uses git in his role as maintainer of the
2003IA64 architecture for the Linux kernel.
2004
2005He uses two public branches:
2006
2007 - A "test" tree into which patches are initially placed so that they
2008   can get some exposure when integrated with other ongoing development.
2009   This tree is available to Andrew for pulling into -mm whenever he
2010   wants.
2011
2012 - A "release" tree into which tested patches are moved for final sanity
2013   checking, and as a vehicle to send them upstream to Linus (by sending
2014   him a "please pull" request.)
2015
2016He also uses a set of temporary branches ("topic branches"), each
2017containing a logical grouping of patches.
2018
2019To set this up, first create your work tree by cloning Linus's public
2020tree:
2021
2022-------------------------------------------------
2023$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git work
2024$ cd work
2025-------------------------------------------------
2026
2027Linus's tree will be stored in the remote branch named origin/master,
2028and can be updated using gitlink:git-fetch[1]; you can track other
2029public trees using gitlink:git-remote[1] to set up a "remote" and
2030gitlink:git-fetch[1] to keep them up-to-date; see
2031<<repositories-and-branches>>.
2032
2033Now create the branches in which you are going to work; these start out
2034at the current tip of origin/master branch, and should be set up (using
2035the --track option to gitlink:git-branch[1]) to merge changes in from
2036Linus by default.
2037
2038-------------------------------------------------
2039$ git branch --track test origin/master
2040$ git branch --track release origin/master
2041-------------------------------------------------
2042
2043These can be easily kept up to date using gitlink:git-pull[1]
2044
2045-------------------------------------------------
2046$ git checkout test && git pull
2047$ git checkout release && git pull
2048-------------------------------------------------
2049
2050Important note!  If you have any local changes in these branches, then
2051this merge will create a commit object in the history (with no local
2052changes git will simply do a "Fast forward" merge).  Many people dislike
2053the "noise" that this creates in the Linux history, so you should avoid
2054doing this capriciously in the "release" branch, as these noisy commits
2055will become part of the permanent history when you ask Linus to pull
2056from the release branch.
2057
2058A few configuration variables (see gitlink:git-config[1]) can
2059make it easy to push both branches to your public tree.  (See
2060<<setting-up-a-public-repository>>.)
2061
2062-------------------------------------------------
2063$ cat >> .git/config <<EOF
2064[remote "mytree"]
2065        url =  master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6.git
2066        push = release
2067        push = test
2068EOF
2069-------------------------------------------------
2070
2071Then you can push both the test and release trees using
2072gitlink:git-push[1]:
2073
2074-------------------------------------------------
2075$ git push mytree
2076-------------------------------------------------
2077
2078or push just one of the test and release branches using:
2079
2080-------------------------------------------------
2081$ git push mytree test
2082-------------------------------------------------
2083
2084or
2085
2086-------------------------------------------------
2087$ git push mytree release
2088-------------------------------------------------
2089
2090Now to apply some patches from the community.  Think of a short
2091snappy name for a branch to hold this patch (or related group of
2092patches), and create a new branch from the current tip of Linus's
2093branch:
2094
2095-------------------------------------------------
2096$ git checkout -b speed-up-spinlocks origin
2097-------------------------------------------------
2098
2099Now you apply the patch(es), run some tests, and commit the change(s).  If
2100the patch is a multi-part series, then you should apply each as a separate
2101commit to this branch.
2102
2103-------------------------------------------------
2104$ ... patch ... test  ... commit [ ... patch ... test ... commit ]*
2105-------------------------------------------------
2106
2107When you are happy with the state of this change, you can pull it into the
2108"test" branch in preparation to make it public:
2109
2110-------------------------------------------------
2111$ git checkout test && git pull . speed-up-spinlocks
2112-------------------------------------------------
2113
2114It is unlikely that you would have any conflicts here ... but you might if you
2115spent a while on this step and had also pulled new versions from upstream.
2116
2117Some time later when enough time has passed and testing done, you can pull the
2118same branch into the "release" tree ready to go upstream.  This is where you
2119see the value of keeping each patch (or patch series) in its own branch.  It
2120means that the patches can be moved into the "release" tree in any order.
2121
2122-------------------------------------------------
2123$ git checkout release && git pull . speed-up-spinlocks
2124-------------------------------------------------
2125
2126After a while, you will have a number of branches, and despite the
2127well chosen names you picked for each of them, you may forget what
2128they are for, or what status they are in.  To get a reminder of what
2129changes are in a specific branch, use:
2130
2131-------------------------------------------------
2132$ git log linux..branchname | git-shortlog
2133-------------------------------------------------
2134
2135To see whether it has already been merged into the test or release branches
2136use:
2137
2138-------------------------------------------------
2139$ git log test..branchname
2140-------------------------------------------------
2141
2142or
2143
2144-------------------------------------------------
2145$ git log release..branchname
2146-------------------------------------------------
2147
2148(If this branch has not yet been merged you will see some log entries.
2149If it has been merged, then there will be no output.)
2150
2151Once a patch completes the great cycle (moving from test to release,
2152then pulled by Linus, and finally coming back into your local
2153"origin/master" branch) the branch for this change is no longer needed.
2154You detect this when the output from:
2155
2156-------------------------------------------------
2157$ git log origin..branchname
2158-------------------------------------------------
2159
2160is empty.  At this point the branch can be deleted:
2161
2162-------------------------------------------------
2163$ git branch -d branchname
2164-------------------------------------------------
2165
2166Some changes are so trivial that it is not necessary to create a separate
2167branch and then merge into each of the test and release branches.  For
2168these changes, just apply directly to the "release" branch, and then
2169merge that into the "test" branch.
2170
2171To create diffstat and shortlog summaries of changes to include in a "please
2172pull" request to Linus you can use:
2173
2174-------------------------------------------------
2175$ git diff --stat origin..release
2176-------------------------------------------------
2177
2178and
2179
2180-------------------------------------------------
2181$ git log -p origin..release | git shortlog
2182-------------------------------------------------
2183
2184Here are some of the scripts that simplify all this even further.
2185
2186-------------------------------------------------
2187==== update script ====
2188# Update a branch in my GIT tree.  If the branch to be updated
2189# is origin, then pull from kernel.org.  Otherwise merge
2190# origin/master branch into test|release branch
2191
2192case "$1" in
2193test|release)
2194        git checkout $1 && git pull . origin
2195        ;;
2196origin)
2197        before=$(git rev-parse refs/remotes/origin/master)
2198        git fetch origin
2199        after=$(git rev-parse refs/remotes/origin/master)
2200        if [ $before != $after ]
2201        then
2202                git log $before..$after | git shortlog
2203        fi
2204        ;;
2205*)
2206        echo "Usage: $0 origin|test|release" 1>&2
2207        exit 1
2208        ;;
2209esac
2210-------------------------------------------------
2211
2212-------------------------------------------------
2213==== merge script ====
2214# Merge a branch into either the test or release branch
2215
2216pname=$0
2217
2218usage()
2219{
2220        echo "Usage: $pname branch test|release" 1>&2
2221        exit 1
2222}
2223
2224git show-ref -q --verify -- refs/heads/"$1" || {
2225        echo "Can't see branch <$1>" 1>&2
2226        usage
2227}
2228
2229case "$2" in
2230test|release)
2231        if [ $(git log $2..$1 | wc -c) -eq 0 ]
2232        then
2233                echo $1 already merged into $2 1>&2
2234                exit 1
2235        fi
2236        git checkout $2 && git pull . $1
2237        ;;
2238*)
2239        usage
2240        ;;
2241esac
2242-------------------------------------------------
2243
2244-------------------------------------------------
2245==== status script ====
2246# report on status of my ia64 GIT tree
2247
2248gb=$(tput setab 2)
2249rb=$(tput setab 1)
2250restore=$(tput setab 9)
2251
2252if [ `git rev-list test..release | wc -c` -gt 0 ]
2253then
2254        echo $rb Warning: commits in release that are not in test $restore
2255        git log test..release
2256fi
2257
2258for branch in `git show-ref --heads | sed 's|^.*/||'`
2259do
2260        if [ $branch = test -o $branch = release ]
2261        then
2262                continue
2263        fi
2264
2265        echo -n $gb ======= $branch ====== $restore " "
2266        status=
2267        for ref in test release origin/master
2268        do
2269                if [ `git rev-list $ref..$branch | wc -c` -gt 0 ]
2270                then
2271                        status=$status${ref:0:1}
2272                fi
2273        done
2274        case $status in
2275        trl)
2276                echo $rb Need to pull into test $restore
2277                ;;
2278        rl)
2279                echo "In test"
2280                ;;
2281        l)
2282                echo "Waiting for linus"
2283                ;;
2284        "")
2285                echo $rb All done $restore
2286                ;;
2287        *)
2288                echo $rb "<$status>" $restore
2289                ;;
2290        esac
2291        git log origin/master..$branch | git shortlog
2292done
2293-------------------------------------------------
2294
2295
2296[[cleaning-up-history]]
2297Rewriting history and maintaining patch series
2298==============================================
2299
2300Normally commits are only added to a project, never taken away or
2301replaced.  Git is designed with this assumption, and violating it will
2302cause git's merge machinery (for example) to do the wrong thing.
2303
2304However, there is a situation in which it can be useful to violate this
2305assumption.
2306
2307[[patch-series]]
2308Creating the perfect patch series
2309---------------------------------
2310
2311Suppose you are a contributor to a large project, and you want to add a
2312complicated feature, and to present it to the other developers in a way
2313that makes it easy for them to read your changes, verify that they are
2314correct, and understand why you made each change.
2315
2316If you present all of your changes as a single patch (or commit), they
2317may find that it is too much to digest all at once.
2318
2319If you present them with the entire history of your work, complete with
2320mistakes, corrections, and dead ends, they may be overwhelmed.
2321
2322So the ideal is usually to produce a series of patches such that:
2323
2324        1. Each patch can be applied in order.
2325
2326        2. Each patch includes a single logical change, together with a
2327           message explaining the change.
2328
2329        3. No patch introduces a regression: after applying any initial
2330           part of the series, the resulting project still compiles and
2331           works, and has no bugs that it didn't have before.
2332
2333        4. The complete series produces the same end result as your own
2334           (probably much messier!) development process did.
2335
2336We will introduce some tools that can help you do this, explain how to
2337use them, and then explain some of the problems that can arise because
2338you are rewriting history.
2339
2340[[using-git-rebase]]
2341Keeping a patch series up to date using git-rebase
2342--------------------------------------------------
2343
2344Suppose that you create a branch "mywork" on a remote-tracking branch
2345"origin", and create some commits on top of it:
2346
2347-------------------------------------------------
2348$ git checkout -b mywork origin
2349$ vi file.txt
2350$ git commit
2351$ vi otherfile.txt
2352$ git commit
2353...
2354-------------------------------------------------
2355
2356You have performed no merges into mywork, so it is just a simple linear
2357sequence of patches on top of "origin":
2358
2359................................................
2360 o--o--o <-- origin
2361        \
2362         o--o--o <-- mywork
2363................................................
2364
2365Some more interesting work has been done in the upstream project, and
2366"origin" has advanced:
2367
2368................................................
2369 o--o--O--o--o--o <-- origin
2370        \
2371         a--b--c <-- mywork
2372................................................
2373
2374At this point, you could use "pull" to merge your changes back in;
2375the result would create a new merge commit, like this:
2376
2377................................................
2378 o--o--O--o--o--o <-- origin
2379        \        \
2380         a--b--c--m <-- mywork
2381................................................
2382
2383However, if you prefer to keep the history in mywork a simple series of
2384commits without any merges, you may instead choose to use
2385gitlink:git-rebase[1]:
2386
2387-------------------------------------------------
2388$ git checkout mywork
2389$ git rebase origin
2390-------------------------------------------------
2391
2392This will remove each of your commits from mywork, temporarily saving
2393them as patches (in a directory named ".dotest"), update mywork to
2394point at the latest version of origin, then apply each of the saved
2395patches to the new mywork.  The result will look like:
2396
2397
2398................................................
2399 o--o--O--o--o--o <-- origin
2400                 \
2401                  a'--b'--c' <-- mywork
2402................................................
2403
2404In the process, it may discover conflicts.  In that case it will stop
2405and allow you to fix the conflicts; after fixing conflicts, use "git
2406add" to update the index with those contents, and then, instead of
2407running git-commit, just run
2408
2409-------------------------------------------------
2410$ git rebase --continue
2411-------------------------------------------------
2412
2413and git will continue applying the rest of the patches.
2414
2415At any point you may use the --abort option to abort this process and
2416return mywork to the state it had before you started the rebase:
2417
2418-------------------------------------------------
2419$ git rebase --abort
2420-------------------------------------------------
2421
2422[[modifying-one-commit]]
2423Modifying a single commit
2424-------------------------
2425
2426We saw in <<fixing-a-mistake-by-editing-history>> that you can replace the
2427most recent commit using
2428
2429-------------------------------------------------
2430$ git commit --amend
2431-------------------------------------------------
2432
2433which will replace the old commit by a new commit incorporating your
2434changes, giving you a chance to edit the old commit message first.
2435
2436You can also use a combination of this and gitlink:git-rebase[1] to edit
2437commits further back in your history.  First, tag the problematic commit with
2438
2439-------------------------------------------------
2440$ git tag bad mywork~5
2441-------------------------------------------------
2442
2443(Either gitk or git-log may be useful for finding the commit.)
2444
2445Then check out that commit, edit it, and rebase the rest of the series
2446on top of it (note that we could check out the commit on a temporary
2447branch, but instead we're using a <<detached-head,detached head>>):
2448
2449-------------------------------------------------
2450$ git checkout bad
2451$ # make changes here and update the index
2452$ git commit --amend
2453$ git rebase --onto HEAD bad mywork
2454-------------------------------------------------
2455
2456When you're done, you'll be left with mywork checked out, with the top
2457patches on mywork reapplied on top of your modified commit.  You can
2458then clean up with
2459
2460-------------------------------------------------
2461$ git tag -d bad
2462-------------------------------------------------
2463
2464Note that the immutable nature of git history means that you haven't really
2465"modified" existing commits; instead, you have replaced the old commits with
2466new commits having new object names.
2467
2468[[reordering-patch-series]]
2469Reordering or selecting from a patch series
2470-------------------------------------------
2471
2472Given one existing commit, the gitlink:git-cherry-pick[1] command
2473allows you to apply the change introduced by that commit and create a
2474new commit that records it.  So, for example, if "mywork" points to a
2475series of patches on top of "origin", you might do something like:
2476
2477-------------------------------------------------
2478$ git checkout -b mywork-new origin
2479$ gitk origin..mywork &
2480-------------------------------------------------
2481
2482And browse through the list of patches in the mywork branch using gitk,
2483applying them (possibly in a different order) to mywork-new using
2484cherry-pick, and possibly modifying them as you go using commit --amend.
2485The gitlink:git-gui[1] command may also help as it allows you to
2486individually select diff hunks for inclusion in the index (by
2487right-clicking on the diff hunk and choosing "Stage Hunk for Commit").
2488
2489Another technique is to use git-format-patch to create a series of
2490patches, then reset the state to before the patches:
2491
2492-------------------------------------------------
2493$ git format-patch origin
2494$ git reset --hard origin
2495-------------------------------------------------
2496
2497Then modify, reorder, or eliminate patches as preferred before applying
2498them again with gitlink:git-am[1].
2499
2500[[patch-series-tools]]
2501Other tools
2502-----------
2503
2504There are numerous other tools, such as StGIT, which exist for the
2505purpose of maintaining a patch series.  These are outside of the scope of
2506this manual.
2507
2508[[problems-with-rewriting-history]]
2509Problems with rewriting history
2510-------------------------------
2511
2512The primary problem with rewriting the history of a branch has to do
2513with merging.  Suppose somebody fetches your branch and merges it into
2514their branch, with a result something like this:
2515
2516................................................
2517 o--o--O--o--o--o <-- origin
2518        \        \
2519         t--t--t--m <-- their branch:
2520................................................
2521
2522Then suppose you modify the last three commits:
2523
2524................................................
2525         o--o--o <-- new head of origin
2526        /
2527 o--o--O--o--o--o <-- old head of origin
2528................................................
2529
2530If we examined all this history together in one repository, it will
2531look like:
2532
2533................................................
2534         o--o--o <-- new head of origin
2535        /
2536 o--o--O--o--o--o <-- old head of origin
2537        \        \
2538         t--t--t--m <-- their branch:
2539................................................
2540
2541Git has no way of knowing that the new head is an updated version of
2542the old head; it treats this situation exactly the same as it would if
2543two developers had independently done the work on the old and new heads
2544in parallel.  At this point, if someone attempts to merge the new head
2545in to their branch, git will attempt to merge together the two (old and
2546new) lines of development, instead of trying to replace the old by the
2547new.  The results are likely to be unexpected.
2548
2549You may still choose to publish branches whose history is rewritten,
2550and it may be useful for others to be able to fetch those branches in
2551order to examine or test them, but they should not attempt to pull such
2552branches into their own work.
2553
2554For true distributed development that supports proper merging,
2555published branches should never be rewritten.
2556
2557[[advanced-branch-management]]
2558Advanced branch management
2559==========================
2560
2561[[fetching-individual-branches]]
2562Fetching individual branches
2563----------------------------
2564
2565Instead of using gitlink:git-remote[1], you can also choose just
2566to update one branch at a time, and to store it locally under an
2567arbitrary name:
2568
2569-------------------------------------------------
2570$ git fetch origin todo:my-todo-work
2571-------------------------------------------------
2572
2573The first argument, "origin", just tells git to fetch from the
2574repository you originally cloned from.  The second argument tells git
2575to fetch the branch named "todo" from the remote repository, and to
2576store it locally under the name refs/heads/my-todo-work.
2577
2578You can also fetch branches from other repositories; so
2579
2580-------------------------------------------------
2581$ git fetch git://example.com/proj.git master:example-master
2582-------------------------------------------------
2583
2584will create a new branch named "example-master" and store in it the
2585branch named "master" from the repository at the given URL.  If you
2586already have a branch named example-master, it will attempt to
2587<<fast-forwards,fast-forward>> to the commit given by example.com's
2588master branch.  In more detail:
2589
2590[[fetch-fast-forwards]]
2591git fetch and fast-forwards
2592---------------------------
2593
2594In the previous example, when updating an existing branch, "git
2595fetch" checks to make sure that the most recent commit on the remote
2596branch is a descendant of the most recent commit on your copy of the
2597branch before updating your copy of the branch to point at the new
2598commit.  Git calls this process a <<fast-forwards,fast forward>>.
2599
2600A fast forward looks something like this:
2601
2602................................................
2603 o--o--o--o <-- old head of the branch
2604           \
2605            o--o--o <-- new head of the branch
2606................................................
2607
2608
2609In some cases it is possible that the new head will *not* actually be
2610a descendant of the old head.  For example, the developer may have
2611realized she made a serious mistake, and decided to backtrack,
2612resulting in a situation like:
2613
2614................................................
2615 o--o--o--o--a--b <-- old head of the branch
2616           \
2617            o--o--o <-- new head of the branch
2618................................................
2619
2620In this case, "git fetch" will fail, and print out a warning.
2621
2622In that case, you can still force git to update to the new head, as
2623described in the following section.  However, note that in the
2624situation above this may mean losing the commits labeled "a" and "b",
2625unless you've already created a reference of your own pointing to
2626them.
2627
2628[[forcing-fetch]]
2629Forcing git fetch to do non-fast-forward updates
2630------------------------------------------------
2631
2632If git fetch fails because the new head of a branch is not a
2633descendant of the old head, you may force the update with:
2634
2635-------------------------------------------------
2636$ git fetch git://example.com/proj.git +master:refs/remotes/example/master
2637-------------------------------------------------
2638
2639Note the addition of the "+" sign.  Alternatively, you can use the "-f"
2640flag to force updates of all the fetched branches, as in:
2641
2642-------------------------------------------------
2643$ git fetch -f origin
2644-------------------------------------------------
2645
2646Be aware that commits that the old version of example/master pointed at
2647may be lost, as we saw in the previous section.
2648
2649[[remote-branch-configuration]]
2650Configuring remote branches
2651---------------------------
2652
2653We saw above that "origin" is just a shortcut to refer to the
2654repository that you originally cloned from.  This information is
2655stored in git configuration variables, which you can see using
2656gitlink:git-config[1]:
2657
2658-------------------------------------------------
2659$ git config -l
2660core.repositoryformatversion=0
2661core.filemode=true
2662core.logallrefupdates=true
2663remote.origin.url=git://git.kernel.org/pub/scm/git/git.git
2664remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*
2665branch.master.remote=origin
2666branch.master.merge=refs/heads/master
2667-------------------------------------------------
2668
2669If there are other repositories that you also use frequently, you can
2670create similar configuration options to save typing; for example,
2671after
2672
2673-------------------------------------------------
2674$ git config remote.example.url git://example.com/proj.git
2675-------------------------------------------------
2676
2677then the following two commands will do the same thing:
2678
2679-------------------------------------------------
2680$ git fetch git://example.com/proj.git master:refs/remotes/example/master
2681$ git fetch example master:refs/remotes/example/master
2682-------------------------------------------------
2683
2684Even better, if you add one more option:
2685
2686-------------------------------------------------
2687$ git config remote.example.fetch master:refs/remotes/example/master
2688-------------------------------------------------
2689
2690then the following commands will all do the same thing:
2691
2692-------------------------------------------------
2693$ git fetch git://example.com/proj.git master:refs/remotes/example/master
2694$ git fetch example master:refs/remotes/example/master
2695$ git fetch example
2696-------------------------------------------------
2697
2698You can also add a "+" to force the update each time:
2699
2700-------------------------------------------------
2701$ git config remote.example.fetch +master:ref/remotes/example/master
2702-------------------------------------------------
2703
2704Don't do this unless you're sure you won't mind "git fetch" possibly
2705throwing away commits on mybranch.
2706
2707Also note that all of the above configuration can be performed by
2708directly editing the file .git/config instead of using
2709gitlink:git-config[1].
2710
2711See gitlink:git-config[1] for more details on the configuration
2712options mentioned above.
2713
2714
2715[[git-concepts]]
2716Git concepts
2717============
2718
2719Git is built on a small number of simple but powerful ideas.  While it
2720is possible to get things done without understanding them, you will find
2721git much more intuitive if you do.
2722
2723We start with the most important, the  <<def_object_database,object
2724database>> and the <<def_index,index>>.
2725
2726[[the-object-database]]
2727The Object Database
2728-------------------
2729
2730
2731We already saw in <<understanding-commits>> that all commits are stored
2732under a 40-digit "object name".  In fact, all the information needed to
2733represent the history of a project is stored in objects with such names.
2734In each case the name is calculated by taking the SHA1 hash of the
2735contents of the object.  The SHA1 hash is a cryptographic hash function.
2736What that means to us is that it is impossible to find two different
2737objects with the same name.  This has a number of advantages; among
2738others:
2739
2740- Git can quickly determine whether two objects are identical or not,
2741  just by comparing names.
2742- Since object names are computed the same way in ever repository, the
2743  same content stored in two repositories will always be stored under
2744  the same name.
2745- Git can detect errors when it reads an object, by checking that the
2746  object's name is still the SHA1 hash of its contents.
2747
2748(See <<object-details>> for the details of the object formatting and
2749SHA1 calculation.)
2750
2751There are four different types of objects: "blob", "tree", "commit", and
2752"tag".
2753
2754- A <<def_blob_object,"blob" object>> is used to store file data.
2755- A <<def_tree_object,"tree" object>> is an object that ties one or more
2756  "blob" objects into a directory structure. In addition, a tree object
2757  can refer to other tree objects, thus creating a directory hierarchy.
2758- A <<def_commit_object,"commit" object>> ties such directory hierarchies
2759  together into a <<def_DAG,directed acyclic graph>> of revisions - each
2760  commit contains the object name of exactly one tree designating the
2761  directory hierarchy at the time of the commit. In addition, a commit
2762  refers to "parent" commit objects that describe the history of how we
2763  arrived at that directory hierarchy.
2764- A <<def_tag_object,"tag" object>> symbolically identifies and can be
2765  used to sign other objects. It contains the object name and type of
2766  another object, a symbolic name (of course!) and, optionally, a
2767  signature.
2768
2769The object types in some more detail:
2770
2771[[commit-object]]
2772Commit Object
2773~~~~~~~~~~~~~
2774
2775The "commit" object links a physical state of a tree with a description
2776of how we got there and why.  Use the --pretty=raw option to
2777gitlink:git-show[1] or gitlink:git-log[1] to examine your favorite
2778commit:
2779
2780------------------------------------------------
2781$ git show -s --pretty=raw 2be7fcb476
2782commit 2be7fcb4764f2dbcee52635b91fedb1b3dcf7ab4
2783tree fb3a8bdd0ceddd019615af4d57a53f43d8cee2bf
2784parent 257a84d9d02e90447b149af58b271c19405edb6a
2785author Dave Watson <dwatson@mimvista.com> 1187576872 -0400
2786committer Junio C Hamano <gitster@pobox.com> 1187591163 -0700
2787
2788    Fix misspelling of 'suppress' in docs
2789
2790    Signed-off-by: Junio C Hamano <gitster@pobox.com>
2791------------------------------------------------
2792
2793As you can see, a commit is defined by:
2794
2795- a tree: The SHA1 name of a tree object (as defined below), representing
2796  the contents of a directory at a certain point in time.
2797- parent(s): The SHA1 name of some number of commits which represent the
2798  immediately prevoius step(s) in the history of the project.  The
2799  example above has one parent; merge commits may have more than
2800  one.  A commit with no parents is called a "root" commit, and
2801  represents the initial revision of a project.  Each project must have
2802  at least one root.  A project can also have multiple roots, though
2803  that isn't common (or necessarily a good idea).
2804- an author: The name of the person responsible for this change, together
2805  with its date.
2806- a committer: The name of the person who actually created the commit,
2807  with the date it was done.  This may be different from the author, for
2808  example, if the author was someone who wrote a patch and emailed it
2809  to the person who used it to create the commit.
2810- a comment describing this commit.
2811
2812Note that a commit does not itself contain any information about what
2813actually changed; all changes are calculated by comparing the contents
2814of the tree referred to by this commit with the trees associated with
2815its parents.  In particular, git does not attempt to record file renames
2816explicitly, though it can identify cases where the existence of the same
2817file data at changing paths suggests a rename.  (See, for example, the
2818-M option to gitlink:git-diff[1]).
2819
2820A commit is usually created by gitlink:git-commit[1], which creates a
2821commit whose parent is normally the current HEAD, and whose tree is
2822taken from the content currently stored in the index.
2823
2824[[tree-object]]
2825Tree Object
2826~~~~~~~~~~~
2827
2828The ever-versatile gitlink:git-show[1] command can also be used to
2829examine tree objects, but gitlink:git-ls-tree[1] will give you more
2830details:
2831
2832------------------------------------------------
2833$ git ls-tree fb3a8bdd0ce
2834100644 blob 63c918c667fa005ff12ad89437f2fdc80926e21c    .gitignore
2835100644 blob 5529b198e8d14decbe4ad99db3f7fb632de0439d    .mailmap
2836100644 blob 6ff87c4664981e4397625791c8ea3bbb5f2279a3    COPYING
2837040000 tree 2fb783e477100ce076f6bf57e4a6f026013dc745    Documentation
2838100755 blob 3c0032cec592a765692234f1cba47dfdcc3a9200    GIT-VERSION-GEN
2839100644 blob 289b046a443c0647624607d471289b2c7dcd470b    INSTALL
2840100644 blob 4eb463797adc693dc168b926b6932ff53f17d0b1    Makefile
2841100644 blob 548142c327a6790ff8821d67c2ee1eff7a656b52    README
2842...
2843------------------------------------------------
2844
2845As you can see, a tree object contains a list of entries, each with a
2846mode, object type, SHA1 name, and name, sorted by name.  It represents
2847the contents of a single directory tree.
2848
2849The object type may be a blob, representing the contents of a file, or
2850another tree, representing the contents of a subdirectory.  Since trees
2851and blobs, like all other objects, are named by the SHA1 hash of their
2852contents, two trees have the same SHA1 name if and only if their
2853contents (including, recursively, the contents of all subdirectories)
2854are identical.  This allows git to quickly determine the differences
2855between two related tree objects, since it can ignore any entries with
2856identical object names.
2857
2858(Note: in the presence of submodules, trees may also have commits as
2859entries.   See gitlink:git-submodule[1] and gitlink:gitmodules.txt[1]
2860for partial documentation.)
2861
2862Note that the files all have mode 644 or 755: git actually only pays
2863attention to the executable bit.
2864
2865[[blob-object]]
2866Blob Object
2867~~~~~~~~~~~
2868
2869You can use gitlink:git-show[1] to examine the contents of a blob; take,
2870for example, the blob in the entry for "COPYING" from the tree above:
2871
2872------------------------------------------------
2873$ git show 6ff87c4664
2874
2875 Note that the only valid version of the GPL as far as this project
2876 is concerned is _this_ particular version of the license (ie v2, not
2877 v2.2 or v3.x or whatever), unless explicitly otherwise stated.
2878...
2879------------------------------------------------
2880
2881A "blob" object is nothing but a binary blob of data.  It doesn't refer
2882to anything else or have attributes of any kind.
2883
2884Since the blob is entirely defined by its data, if two files in a
2885directory tree (or in multiple different versions of the repository)
2886have the same contents, they will share the same blob object. The object
2887is totally independent of its location in the directory tree, and
2888renaming a file does not change the object that file is associated with.
2889
2890Note that any tree or blob object can be examined using
2891gitlink:git-show[1] with the <revision>:<path> syntax.  This can
2892sometimes be useful for browsing the contents of a tree that is not
2893currently checked out.
2894
2895[[trust]]
2896Trust
2897~~~~~
2898
2899If you receive the SHA1 name of a blob from one source, and its contents
2900from another (possibly untrusted) source, you can still trust that those
2901contents are correct as long as the SHA1 name agrees.  This is because
2902the SHA1 is designed so that it is infeasible to find different contents
2903that produce the same hash.
2904
2905Similarly, you need only trust the SHA1 name of a top-level tree object
2906to trust the contents of the entire directory that it refers to, and if
2907you receive the SHA1 name of a commit from a trusted source, then you
2908can easily verify the entire history of commits reachable through
2909parents of that commit, and all of those contents of the trees referred
2910to by those commits.
2911
2912So to introduce some real trust in the system, the only thing you need
2913to do is to digitally sign just 'one' special note, which includes the
2914name of a top-level commit.  Your digital signature shows others
2915that you trust that commit, and the immutability of the history of
2916commits tells others that they can trust the whole history.
2917
2918In other words, you can easily validate a whole archive by just
2919sending out a single email that tells the people the name (SHA1 hash)
2920of the top commit, and digitally sign that email using something
2921like GPG/PGP.
2922
2923To assist in this, git also provides the tag object...
2924
2925[[tag-object]]
2926Tag Object
2927~~~~~~~~~~
2928
2929A tag object contains an object, object type, tag name, the name of the
2930person ("tagger") who created the tag, and a message, which may contain
2931a signature, as can be seen using the gitlink:git-cat-file[1]:
2932
2933------------------------------------------------
2934$ git cat-file tag v1.5.0
2935object 437b1b20df4b356c9342dac8d38849f24ef44f27
2936type commit
2937tag v1.5.0
2938tagger Junio C Hamano <junkio@cox.net> 1171411200 +0000
2939
2940GIT 1.5.0
2941-----BEGIN PGP SIGNATURE-----
2942Version: GnuPG v1.4.6 (GNU/Linux)
2943
2944iD8DBQBF0lGqwMbZpPMRm5oRAuRiAJ9ohBLd7s2kqjkKlq1qqC57SbnmzQCdG4ui
2945nLE/L9aUXdWeTFPron96DLA=
2946=2E+0
2947-----END PGP SIGNATURE-----
2948------------------------------------------------
2949
2950See the gitlink:git-tag[1] command to learn how to create and verify tag
2951objects.  (Note that gitlink:git-tag[1] can also be used to create
2952"lightweight tags", which are not tag objects at all, but just simple
2953references whose names begin with "refs/tags/").
2954
2955[[pack-files]]
2956How git stores objects efficiently: pack files
2957~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2958
2959Newly created objects are initially created in a file named after the
2960object's SHA1 hash (stored in .git/objects).
2961
2962Unfortunately this system becomes inefficient once a project has a
2963lot of objects.  Try this on an old project:
2964
2965------------------------------------------------
2966$ git count-objects
29676930 objects, 47620 kilobytes
2968------------------------------------------------
2969
2970The first number is the number of objects which are kept in
2971individual files.  The second is the amount of space taken up by
2972those "loose" objects.
2973
2974You can save space and make git faster by moving these loose objects in
2975to a "pack file", which stores a group of objects in an efficient
2976compressed format; the details of how pack files are formatted can be
2977found in link:technical/pack-format.txt[technical/pack-format.txt].
2978
2979To put the loose objects into a pack, just run git repack:
2980
2981------------------------------------------------
2982$ git repack
2983Generating pack...
2984Done counting 6020 objects.
2985Deltifying 6020 objects.
2986 100% (6020/6020) done
2987Writing 6020 objects.
2988 100% (6020/6020) done
2989Total 6020, written 6020 (delta 4070), reused 0 (delta 0)
2990Pack pack-3e54ad29d5b2e05838c75df582c65257b8d08e1c created.
2991------------------------------------------------
2992
2993You can then run
2994
2995------------------------------------------------
2996$ git prune
2997------------------------------------------------
2998
2999to remove any of the "loose" objects that are now contained in the
3000pack.  This will also remove any unreferenced objects (which may be
3001created when, for example, you use "git reset" to remove a commit).
3002You can verify that the loose objects are gone by looking at the
3003.git/objects directory or by running
3004
3005------------------------------------------------
3006$ git count-objects
30070 objects, 0 kilobytes
3008------------------------------------------------
3009
3010Although the object files are gone, any commands that refer to those
3011objects will work exactly as they did before.
3012
3013The gitlink:git-gc[1] command performs packing, pruning, and more for
3014you, so is normally the only high-level command you need.
3015
3016[[dangling-objects]]
3017Dangling objects
3018~~~~~~~~~~~~~~~~
3019
3020The gitlink:git-fsck[1] command will sometimes complain about dangling
3021objects.  They are not a problem.
3022
3023The most common cause of dangling objects is that you've rebased a
3024branch, or you have pulled from somebody else who rebased a branch--see
3025<<cleaning-up-history>>.  In that case, the old head of the original
3026branch still exists, as does everything it pointed to. The branch
3027pointer itself just doesn't, since you replaced it with another one.
3028
3029There are also other situations that cause dangling objects. For
3030example, a "dangling blob" may arise because you did a "git add" of a
3031file, but then, before you actually committed it and made it part of the
3032bigger picture, you changed something else in that file and committed
3033that *updated* thing - the old state that you added originally ends up
3034not being pointed to by any commit or tree, so it's now a dangling blob
3035object.
3036
3037Similarly, when the "recursive" merge strategy runs, and finds that
3038there are criss-cross merges and thus more than one merge base (which is
3039fairly unusual, but it does happen), it will generate one temporary
3040midway tree (or possibly even more, if you had lots of criss-crossing
3041merges and more than two merge bases) as a temporary internal merge
3042base, and again, those are real objects, but the end result will not end
3043up pointing to them, so they end up "dangling" in your repository.
3044
3045Generally, dangling objects aren't anything to worry about. They can
3046even be very useful: if you screw something up, the dangling objects can
3047be how you recover your old tree (say, you did a rebase, and realized
3048that you really didn't want to - you can look at what dangling objects
3049you have, and decide to reset your head to some old dangling state).
3050
3051For commits, you can just use:
3052
3053------------------------------------------------
3054$ gitk <dangling-commit-sha-goes-here> --not --all
3055------------------------------------------------
3056
3057This asks for all the history reachable from the given commit but not
3058from any branch, tag, or other reference.  If you decide it's something
3059you want, you can always create a new reference to it, e.g.,
3060
3061------------------------------------------------
3062$ git branch recovered-branch <dangling-commit-sha-goes-here>
3063------------------------------------------------
3064
3065For blobs and trees, you can't do the same, but you can still examine
3066them.  You can just do
3067
3068------------------------------------------------
3069$ git show <dangling-blob/tree-sha-goes-here>
3070------------------------------------------------
3071
3072to show what the contents of the blob were (or, for a tree, basically
3073what the "ls" for that directory was), and that may give you some idea
3074of what the operation was that left that dangling object.
3075
3076Usually, dangling blobs and trees aren't very interesting. They're
3077almost always the result of either being a half-way mergebase (the blob
3078will often even have the conflict markers from a merge in it, if you
3079have had conflicting merges that you fixed up by hand), or simply
3080because you interrupted a "git fetch" with ^C or something like that,
3081leaving _some_ of the new objects in the object database, but just
3082dangling and useless.
3083
3084Anyway, once you are sure that you're not interested in any dangling
3085state, you can just prune all unreachable objects:
3086
3087------------------------------------------------
3088$ git prune
3089------------------------------------------------
3090
3091and they'll be gone. But you should only run "git prune" on a quiescent
3092repository - it's kind of like doing a filesystem fsck recovery: you
3093don't want to do that while the filesystem is mounted.
3094
3095(The same is true of "git-fsck" itself, btw - but since
3096git-fsck never actually *changes* the repository, it just reports
3097on what it found, git-fsck itself is never "dangerous" to run.
3098Running it while somebody is actually changing the repository can cause
3099confusing and scary messages, but it won't actually do anything bad. In
3100contrast, running "git prune" while somebody is actively changing the
3101repository is a *BAD* idea).
3102
3103[[the-index]]
3104The index
3105-----------
3106
3107The index is a binary file (generally kept in .git/index) containing a
3108sorted list of path names, each with permissions and the SHA1 of a blob
3109object; gitlink:git-ls-files[1] can show you the contents of the index:
3110
3111-------------------------------------------------
3112$ git ls-files --stage
3113100644 63c918c667fa005ff12ad89437f2fdc80926e21c 0       .gitignore
3114100644 5529b198e8d14decbe4ad99db3f7fb632de0439d 0       .mailmap
3115100644 6ff87c4664981e4397625791c8ea3bbb5f2279a3 0       COPYING
3116100644 a37b2152bd26be2c2289e1f57a292534a51a93c7 0       Documentation/.gitignore
3117100644 fbefe9a45b00a54b58d94d06eca48b03d40a50e0 0       Documentation/Makefile
3118...
3119100644 2511aef8d89ab52be5ec6a5e46236b4b6bcd07ea 0       xdiff/xtypes.h
3120100644 2ade97b2574a9f77e7ae4002a4e07a6a38e46d07 0       xdiff/xutils.c
3121100644 d5de8292e05e7c36c4b68857c1cf9855e3d2f70a 0       xdiff/xutils.h
3122-------------------------------------------------
3123
3124Note that in older documentation you may see the index called the
3125"current directory cache" or just the "cache".  It has three important
3126properties:
3127
31281. The index contains all the information necessary to generate a single
3129(uniquely determined) tree object.
3130+
3131For example, running gitlink:git-commit[1] generates this tree object
3132from the index, stores it in the object database, and uses it as the
3133tree object associated with the new commit.
3134
31352. The index enables fast comparisons between the tree object it defines
3136and the working tree.
3137+
3138It does this by storing some additional data for each entry (such as
3139the last modified time).  This data is not displayed above, and is not
3140stored in the created tree object, but it can be used to determine
3141quickly which files in the working directory differ from what was
3142stored in the index, and thus save git from having to read all of the
3143data from such files to look for changes.
3144
31453. It can efficiently represent information about merge conflicts
3146between different tree objects, allowing each pathname to be
3147associated with sufficient information about the trees involved that
3148you can create a three-way merge between them.
3149+
3150We saw in <<conflict-resolution>> that during a merge the index can
3151store multiple versions of a single file (called "stages").  The third
3152column in the gitlink:git-ls-files[1] output above is the stage
3153number, and will take on values other than 0 for files with merge
3154conflicts.
3155
3156The index is thus a sort of temporary staging area, which is filled with
3157a tree which you are in the process of working on.
3158
3159If you blow the index away entirely, you generally haven't lost any
3160information as long as you have the name of the tree that it described.
3161
3162[[low-level-operations]]
3163Low-level git operations
3164========================
3165
3166Many of the higher-level commands were originally implemented as shell
3167scripts using a smaller core of low-level git commands.  These can still
3168be useful when doing unusual things with git, or just as a way to
3169understand its inner workings.
3170
3171[[object-manipulation]]
3172Object access and manipulation
3173------------------------------
3174
3175The gitlink:git-cat-file[1] command can show the contents of any object,
3176though the higher-level gitlink:git-show[1] is usually more useful.
3177
3178The gitlink:git-commit-tree[1] command allows constructing commits with
3179arbitrary parents and trees.
3180
3181A tree can be created with gitlink:git-write-tree[1] and its data can be
3182accessed by gitlink:git-ls-tree[1].  Two trees can be compared with
3183gitlink:git-diff-tree[1].
3184
3185A tag is created with gitlink:git-mktag[1], and the signature can be
3186verified by gitlink:git-verify-tag[1], though it is normally simpler to
3187use gitlink:git-tag[1] for both.
3188
3189[[the-workflow]]
3190The Workflow
3191------------
3192
3193High-level operations such as gitlink:git-commit[1],
3194gitlink:git-checkout[1] and git-reset[1] work by moving data between the
3195working tree, the index, and the object database.  Git provides
3196low-level operations which perform each of these steps individually.
3197
3198Generally, all "git" operations work on the index file. Some operations
3199work *purely* on the index file (showing the current state of the
3200index), but most operations move data between the index file and either
3201the database or the working directory. Thus there are four main
3202combinations:
3203
3204[[working-directory-to-index]]
3205working directory -> index
3206~~~~~~~~~~~~~~~~~~~~~~~~~~
3207
3208The gitlink:git-update-index[1] command updates the index with
3209information from the working directory.  You generally update the
3210index information by just specifying the filename you want to update,
3211like so:
3212
3213-------------------------------------------------
3214$ git update-index filename
3215-------------------------------------------------
3216
3217but to avoid common mistakes with filename globbing etc, the command
3218will not normally add totally new entries or remove old entries,
3219i.e. it will normally just update existing cache entries.
3220
3221To tell git that yes, you really do realize that certain files no
3222longer exist, or that new files should be added, you
3223should use the `--remove` and `--add` flags respectively.
3224
3225NOTE! A `--remove` flag does 'not' mean that subsequent filenames will
3226necessarily be removed: if the files still exist in your directory
3227structure, the index will be updated with their new status, not
3228removed. The only thing `--remove` means is that update-cache will be
3229considering a removed file to be a valid thing, and if the file really
3230does not exist any more, it will update the index accordingly.
3231
3232As a special case, you can also do `git-update-index --refresh`, which
3233will refresh the "stat" information of each index to match the current
3234stat information. It will 'not' update the object status itself, and
3235it will only update the fields that are used to quickly test whether
3236an object still matches its old backing store object.
3237
3238The previously introduced gitlink:git-add[1] is just a wrapper for
3239gitlink:git-update-index[1].
3240
3241[[index-to-object-database]]
3242index -> object database
3243~~~~~~~~~~~~~~~~~~~~~~~~
3244
3245You write your current index file to a "tree" object with the program
3246
3247-------------------------------------------------
3248$ git write-tree
3249-------------------------------------------------
3250
3251that doesn't come with any options - it will just write out the
3252current index into the set of tree objects that describe that state,
3253and it will return the name of the resulting top-level tree. You can
3254use that tree to re-generate the index at any time by going in the
3255other direction:
3256
3257[[object-database-to-index]]
3258object database -> index
3259~~~~~~~~~~~~~~~~~~~~~~~~
3260
3261You read a "tree" file from the object database, and use that to
3262populate (and overwrite - don't do this if your index contains any
3263unsaved state that you might want to restore later!) your current
3264index.  Normal operation is just
3265
3266-------------------------------------------------
3267$ git-read-tree <sha1 of tree>
3268-------------------------------------------------
3269
3270and your index file will now be equivalent to the tree that you saved
3271earlier. However, that is only your 'index' file: your working
3272directory contents have not been modified.
3273
3274[[index-to-working-directory]]
3275index -> working directory
3276~~~~~~~~~~~~~~~~~~~~~~~~~~
3277
3278You update your working directory from the index by "checking out"
3279files. This is not a very common operation, since normally you'd just
3280keep your files updated, and rather than write to your working
3281directory, you'd tell the index files about the changes in your
3282working directory (i.e. `git-update-index`).
3283
3284However, if you decide to jump to a new version, or check out somebody
3285else's version, or just restore a previous tree, you'd populate your
3286index file with read-tree, and then you need to check out the result
3287with
3288
3289-------------------------------------------------
3290$ git-checkout-index filename
3291-------------------------------------------------
3292
3293or, if you want to check out all of the index, use `-a`.
3294
3295NOTE! git-checkout-index normally refuses to overwrite old files, so
3296if you have an old version of the tree already checked out, you will
3297need to use the "-f" flag ('before' the "-a" flag or the filename) to
3298'force' the checkout.
3299
3300
3301Finally, there are a few odds and ends which are not purely moving
3302from one representation to the other:
3303
3304[[tying-it-all-together]]
3305Tying it all together
3306~~~~~~~~~~~~~~~~~~~~~
3307
3308To commit a tree you have instantiated with "git-write-tree", you'd
3309create a "commit" object that refers to that tree and the history
3310behind it - most notably the "parent" commits that preceded it in
3311history.
3312
3313Normally a "commit" has one parent: the previous state of the tree
3314before a certain change was made. However, sometimes it can have two
3315or more parent commits, in which case we call it a "merge", due to the
3316fact that such a commit brings together ("merges") two or more
3317previous states represented by other commits.
3318
3319In other words, while a "tree" represents a particular directory state
3320of a working directory, a "commit" represents that state in "time",
3321and explains how we got there.
3322
3323You create a commit object by giving it the tree that describes the
3324state at the time of the commit, and a list of parents:
3325
3326-------------------------------------------------
3327$ git-commit-tree <tree> -p <parent> [-p <parent2> ..]
3328-------------------------------------------------
3329
3330and then giving the reason for the commit on stdin (either through
3331redirection from a pipe or file, or by just typing it at the tty).
3332
3333git-commit-tree will return the name of the object that represents
3334that commit, and you should save it away for later use. Normally,
3335you'd commit a new `HEAD` state, and while git doesn't care where you
3336save the note about that state, in practice we tend to just write the
3337result to the file pointed at by `.git/HEAD`, so that we can always see
3338what the last committed state was.
3339
3340Here is an ASCII art by Jon Loeliger that illustrates how
3341various pieces fit together.
3342
3343------------
3344
3345                     commit-tree
3346                      commit obj
3347                       +----+
3348                       |    |
3349                       |    |
3350                       V    V
3351                    +-----------+
3352                    | Object DB |
3353                    |  Backing  |
3354                    |   Store   |
3355                    +-----------+
3356                       ^
3357           write-tree  |     |
3358             tree obj  |     |
3359                       |     |  read-tree
3360                       |     |  tree obj
3361                             V
3362                    +-----------+
3363                    |   Index   |
3364                    |  "cache"  |
3365                    +-----------+
3366         update-index  ^
3367             blob obj  |     |
3368                       |     |
3369    checkout-index -u  |     |  checkout-index
3370             stat      |     |  blob obj
3371                             V
3372                    +-----------+
3373                    |  Working  |
3374                    | Directory |
3375                    +-----------+
3376
3377------------
3378
3379
3380[[examining-the-data]]
3381Examining the data
3382------------------
3383
3384You can examine the data represented in the object database and the
3385index with various helper tools. For every object, you can use
3386gitlink:git-cat-file[1] to examine details about the
3387object:
3388
3389-------------------------------------------------
3390$ git-cat-file -t <objectname>
3391-------------------------------------------------
3392
3393shows the type of the object, and once you have the type (which is
3394usually implicit in where you find the object), you can use
3395
3396-------------------------------------------------
3397$ git-cat-file blob|tree|commit|tag <objectname>
3398-------------------------------------------------
3399
3400to show its contents. NOTE! Trees have binary content, and as a result
3401there is a special helper for showing that content, called
3402`git-ls-tree`, which turns the binary content into a more easily
3403readable form.
3404
3405It's especially instructive to look at "commit" objects, since those
3406tend to be small and fairly self-explanatory. In particular, if you
3407follow the convention of having the top commit name in `.git/HEAD`,
3408you can do
3409
3410-------------------------------------------------
3411$ git-cat-file commit HEAD
3412-------------------------------------------------
3413
3414to see what the top commit was.
3415
3416[[merging-multiple-trees]]
3417Merging multiple trees
3418----------------------
3419
3420Git helps you do a three-way merge, which you can expand to n-way by
3421repeating the merge procedure arbitrary times until you finally
3422"commit" the state.  The normal situation is that you'd only do one
3423three-way merge (two parents), and commit it, but if you like to, you
3424can do multiple parents in one go.
3425
3426To do a three-way merge, you need the two sets of "commit" objects
3427that you want to merge, use those to find the closest common parent (a
3428third "commit" object), and then use those commit objects to find the
3429state of the directory ("tree" object) at these points.
3430
3431To get the "base" for the merge, you first look up the common parent
3432of two commits with
3433
3434-------------------------------------------------
3435$ git-merge-base <commit1> <commit2>
3436-------------------------------------------------
3437
3438which will return you the commit they are both based on.  You should
3439now look up the "tree" objects of those commits, which you can easily
3440do with (for example)
3441
3442-------------------------------------------------
3443$ git-cat-file commit <commitname> | head -1
3444-------------------------------------------------
3445
3446since the tree object information is always the first line in a commit
3447object.
3448
3449Once you know the three trees you are going to merge (the one "original"
3450tree, aka the common tree, and the two "result" trees, aka the branches
3451you want to merge), you do a "merge" read into the index. This will
3452complain if it has to throw away your old index contents, so you should
3453make sure that you've committed those - in fact you would normally
3454always do a merge against your last commit (which should thus match what
3455you have in your current index anyway).
3456
3457To do the merge, do
3458
3459-------------------------------------------------
3460$ git-read-tree -m -u <origtree> <yourtree> <targettree>
3461-------------------------------------------------
3462
3463which will do all trivial merge operations for you directly in the
3464index file, and you can just write the result out with
3465`git-write-tree`.
3466
3467
3468[[merging-multiple-trees-2]]
3469Merging multiple trees, continued
3470---------------------------------
3471
3472Sadly, many merges aren't trivial. If there are files that have
3473been added.moved or removed, or if both branches have modified the
3474same file, you will be left with an index tree that contains "merge
3475entries" in it. Such an index tree can 'NOT' be written out to a tree
3476object, and you will have to resolve any such merge clashes using
3477other tools before you can write out the result.
3478
3479You can examine such index state with `git-ls-files --unmerged`
3480command.  An example:
3481
3482------------------------------------------------
3483$ git-read-tree -m $orig HEAD $target
3484$ git-ls-files --unmerged
3485100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1       hello.c
3486100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2       hello.c
3487100644 cc44c73eb783565da5831b4d820c962954019b69 3       hello.c
3488------------------------------------------------
3489
3490Each line of the `git-ls-files --unmerged` output begins with
3491the blob mode bits, blob SHA1, 'stage number', and the
3492filename.  The 'stage number' is git's way to say which tree it
3493came from: stage 1 corresponds to `$orig` tree, stage 2 `HEAD`
3494tree, and stage3 `$target` tree.
3495
3496Earlier we said that trivial merges are done inside
3497`git-read-tree -m`.  For example, if the file did not change
3498from `$orig` to `HEAD` nor `$target`, or if the file changed
3499from `$orig` to `HEAD` and `$orig` to `$target` the same way,
3500obviously the final outcome is what is in `HEAD`.  What the
3501above example shows is that file `hello.c` was changed from
3502`$orig` to `HEAD` and `$orig` to `$target` in a different way.
3503You could resolve this by running your favorite 3-way merge
3504program, e.g.  `diff3`, `merge`, or git's own merge-file, on
3505the blob objects from these three stages yourself, like this:
3506
3507------------------------------------------------
3508$ git-cat-file blob 263414f... >hello.c~1
3509$ git-cat-file blob 06fa6a2... >hello.c~2
3510$ git-cat-file blob cc44c73... >hello.c~3
3511$ git merge-file hello.c~2 hello.c~1 hello.c~3
3512------------------------------------------------
3513
3514This would leave the merge result in `hello.c~2` file, along
3515with conflict markers if there are conflicts.  After verifying
3516the merge result makes sense, you can tell git what the final
3517merge result for this file is by:
3518
3519-------------------------------------------------
3520$ mv -f hello.c~2 hello.c
3521$ git-update-index hello.c
3522-------------------------------------------------
3523
3524When a path is in unmerged state, running `git-update-index` for
3525that path tells git to mark the path resolved.
3526
3527The above is the description of a git merge at the lowest level,
3528to help you understand what conceptually happens under the hood.
3529In practice, nobody, not even git itself, uses three `git-cat-file`
3530for this.  There is `git-merge-index` program that extracts the
3531stages to temporary files and calls a "merge" script on it:
3532
3533-------------------------------------------------
3534$ git-merge-index git-merge-one-file hello.c
3535-------------------------------------------------
3536
3537and that is what higher level `git merge -s resolve` is implemented with.
3538
3539[[hacking-git]]
3540Hacking git
3541===========
3542
3543This chapter covers internal details of the git implementation which
3544probably only git developers need to understand.
3545
3546[[object-details]]
3547Object storage format
3548---------------------
3549
3550All objects have a statically determined "type" which identifies the
3551format of the object (i.e. how it is used, and how it can refer to other
3552objects).  There are currently four different object types: "blob",
3553"tree", "commit", and "tag".
3554
3555Regardless of object type, all objects share the following
3556characteristics: they are all deflated with zlib, and have a header
3557that not only specifies their type, but also provides size information
3558about the data in the object.  It's worth noting that the SHA1 hash
3559that is used to name the object is the hash of the original data
3560plus this header, so `sha1sum` 'file' does not match the object name
3561for 'file'.
3562(Historical note: in the dawn of the age of git the hash
3563was the sha1 of the 'compressed' object.)
3564
3565As a result, the general consistency of an object can always be tested
3566independently of the contents or the type of the object: all objects can
3567be validated by verifying that (a) their hashes match the content of the
3568file and (b) the object successfully inflates to a stream of bytes that
3569forms a sequence of <ascii type without space> {plus} <space> {plus} <ascii decimal
3570size> {plus} <byte\0> {plus} <binary object data>.
3571
3572The structured objects can further have their structure and
3573connectivity to other objects verified. This is generally done with
3574the `git-fsck` program, which generates a full dependency graph
3575of all objects, and verifies their internal consistency (in addition
3576to just verifying their superficial consistency through the hash).
3577
3578[[birdview-on-the-source-code]]
3579A birds-eye view of Git's source code
3580-------------------------------------
3581
3582It is not always easy for new developers to find their way through Git's
3583source code.  This section gives you a little guidance to show where to
3584start.
3585
3586A good place to start is with the contents of the initial commit, with:
3587
3588----------------------------------------------------
3589$ git checkout e83c5163
3590----------------------------------------------------
3591
3592The initial revision lays the foundation for almost everything git has
3593today, but is small enough to read in one sitting.
3594
3595Note that terminology has changed since that revision.  For example, the
3596README in that revision uses the word "changeset" to describe what we
3597now call a <<def_commit_object,commit>>.
3598
3599Also, we do not call it "cache" any more, but "index", however, the
3600file is still called `cache.h`.  Remark: Not much reason to change it now,
3601especially since there is no good single name for it anyway, because it is
3602basically _the_ header file which is included by _all_ of Git's C sources.
3603
3604If you grasp the ideas in that initial commit, you should check out a
3605more recent version and skim `cache.h`, `object.h` and `commit.h`.
3606
3607In the early days, Git (in the tradition of UNIX) was a bunch of programs
3608which were extremely simple, and which you used in scripts, piping the
3609output of one into another. This turned out to be good for initial
3610development, since it was easier to test new things.  However, recently
3611many of these parts have become builtins, and some of the core has been
3612"libified", i.e. put into libgit.a for performance, portability reasons,
3613and to avoid code duplication.
3614
3615By now, you know what the index is (and find the corresponding data
3616structures in `cache.h`), and that there are just a couple of object types
3617(blobs, trees, commits and tags) which inherit their common structure from
3618`struct object`, which is their first member (and thus, you can cast e.g.
3619`(struct object *)commit` to achieve the _same_ as `&commit->object`, i.e.
3620get at the object name and flags).
3621
3622Now is a good point to take a break to let this information sink in.
3623
3624Next step: get familiar with the object naming.  Read <<naming-commits>>.
3625There are quite a few ways to name an object (and not only revisions!).
3626All of these are handled in `sha1_name.c`. Just have a quick look at
3627the function `get_sha1()`. A lot of the special handling is done by
3628functions like `get_sha1_basic()` or the likes.
3629
3630This is just to get you into the groove for the most libified part of Git:
3631the revision walker.
3632
3633Basically, the initial version of `git log` was a shell script:
3634
3635----------------------------------------------------------------
3636$ git-rev-list --pretty $(git-rev-parse --default HEAD "$@") | \
3637        LESS=-S ${PAGER:-less}
3638----------------------------------------------------------------
3639
3640What does this mean?
3641
3642`git-rev-list` is the original version of the revision walker, which
3643_always_ printed a list of revisions to stdout.  It is still functional,
3644and needs to, since most new Git programs start out as scripts using
3645`git-rev-list`.
3646
3647`git-rev-parse` is not as important any more; it was only used to filter out
3648options that were relevant for the different plumbing commands that were
3649called by the script.
3650
3651Most of what `git-rev-list` did is contained in `revision.c` and
3652`revision.h`.  It wraps the options in a struct named `rev_info`, which
3653controls how and what revisions are walked, and more.
3654
3655The original job of `git-rev-parse` is now taken by the function
3656`setup_revisions()`, which parses the revisions and the common command line
3657options for the revision walker. This information is stored in the struct
3658`rev_info` for later consumption. You can do your own command line option
3659parsing after calling `setup_revisions()`. After that, you have to call
3660`prepare_revision_walk()` for initialization, and then you can get the
3661commits one by one with the function `get_revision()`.
3662
3663If you are interested in more details of the revision walking process,
3664just have a look at the first implementation of `cmd_log()`; call
3665`git-show v1.3.0~155^2~4` and scroll down to that function (note that you
3666no longer need to call `setup_pager()` directly).
3667
3668Nowadays, `git log` is a builtin, which means that it is _contained_ in the
3669command `git`.  The source side of a builtin is
3670
3671- a function called `cmd_<bla>`, typically defined in `builtin-<bla>.c`,
3672  and declared in `builtin.h`,
3673
3674- an entry in the `commands[]` array in `git.c`, and
3675
3676- an entry in `BUILTIN_OBJECTS` in the `Makefile`.
3677
3678Sometimes, more than one builtin is contained in one source file.  For
3679example, `cmd_whatchanged()` and `cmd_log()` both reside in `builtin-log.c`,
3680since they share quite a bit of code.  In that case, the commands which are
3681_not_ named like the `.c` file in which they live have to be listed in
3682`BUILT_INS` in the `Makefile`.
3683
3684`git log` looks more complicated in C than it does in the original script,
3685but that allows for a much greater flexibility and performance.
3686
3687Here again it is a good point to take a pause.
3688
3689Lesson three is: study the code.  Really, it is the best way to learn about
3690the organization of Git (after you know the basic concepts).
3691
3692So, think about something which you are interested in, say, "how can I
3693access a blob just knowing the object name of it?".  The first step is to
3694find a Git command with which you can do it.  In this example, it is either
3695`git show` or `git cat-file`.
3696
3697For the sake of clarity, let's stay with `git cat-file`, because it
3698
3699- is plumbing, and
3700
3701- was around even in the initial commit (it literally went only through
3702  some 20 revisions as `cat-file.c`, was renamed to `builtin-cat-file.c`
3703  when made a builtin, and then saw less than 10 versions).
3704
3705So, look into `builtin-cat-file.c`, search for `cmd_cat_file()` and look what
3706it does.
3707
3708------------------------------------------------------------------
3709        git_config(git_default_config);
3710        if (argc != 3)
3711                usage("git-cat-file [-t|-s|-e|-p|<type>] <sha1>");
3712        if (get_sha1(argv[2], sha1))
3713                die("Not a valid object name %s", argv[2]);
3714------------------------------------------------------------------
3715
3716Let's skip over the obvious details; the only really interesting part
3717here is the call to `get_sha1()`.  It tries to interpret `argv[2]` as an
3718object name, and if it refers to an object which is present in the current
3719repository, it writes the resulting SHA-1 into the variable `sha1`.
3720
3721Two things are interesting here:
3722
3723- `get_sha1()` returns 0 on _success_.  This might surprise some new
3724  Git hackers, but there is a long tradition in UNIX to return different
3725  negative numbers in case of different errors -- and 0 on success.
3726
3727- the variable `sha1` in the function signature of `get_sha1()` is `unsigned
3728  char \*`, but is actually expected to be a pointer to `unsigned
3729  char[20]`.  This variable will contain the 160-bit SHA-1 of the given
3730  commit.  Note that whenever a SHA-1 is passed as `unsigned char \*`, it
3731  is the binary representation, as opposed to the ASCII representation in
3732  hex characters, which is passed as `char *`.
3733
3734You will see both of these things throughout the code.
3735
3736Now, for the meat:
3737
3738-----------------------------------------------------------------------------
3739        case 0:
3740                buf = read_object_with_reference(sha1, argv[1], &size, NULL);
3741-----------------------------------------------------------------------------
3742
3743This is how you read a blob (actually, not only a blob, but any type of
3744object).  To know how the function `read_object_with_reference()` actually
3745works, find the source code for it (something like `git grep
3746read_object_with | grep ":[a-z]"` in the git repository), and read
3747the source.
3748
3749To find out how the result can be used, just read on in `cmd_cat_file()`:
3750
3751-----------------------------------
3752        write_or_die(1, buf, size);
3753-----------------------------------
3754
3755Sometimes, you do not know where to look for a feature.  In many such cases,
3756it helps to search through the output of `git log`, and then `git show` the
3757corresponding commit.
3758
3759Example: If you know that there was some test case for `git bundle`, but
3760do not remember where it was (yes, you _could_ `git grep bundle t/`, but that
3761does not illustrate the point!):
3762
3763------------------------
3764$ git log --no-merges t/
3765------------------------
3766
3767In the pager (`less`), just search for "bundle", go a few lines back,
3768and see that it is in commit 18449ab0...  Now just copy this object name,
3769and paste it into the command line
3770
3771-------------------
3772$ git show 18449ab0
3773-------------------
3774
3775Voila.
3776
3777Another example: Find out what to do in order to make some script a
3778builtin:
3779
3780-------------------------------------------------
3781$ git log --no-merges --diff-filter=A builtin-*.c
3782-------------------------------------------------
3783
3784You see, Git is actually the best tool to find out about the source of Git
3785itself!
3786
3787[[glossary]]
3788include::glossary.txt[]
3789
3790[[git-quick-start]]
3791Appendix A: Git Quick Reference
3792===============================
3793
3794This is a quick summary of the major commands; the previous chapters
3795explain how these work in more detail.
3796
3797[[quick-creating-a-new-repository]]
3798Creating a new repository
3799-------------------------
3800
3801From a tarball:
3802
3803-----------------------------------------------
3804$ tar xzf project.tar.gz
3805$ cd project
3806$ git init
3807Initialized empty Git repository in .git/
3808$ git add .
3809$ git commit
3810-----------------------------------------------
3811
3812From a remote repository:
3813
3814-----------------------------------------------
3815$ git clone git://example.com/pub/project.git
3816$ cd project
3817-----------------------------------------------
3818
3819[[managing-branches]]
3820Managing branches
3821-----------------
3822
3823-----------------------------------------------
3824$ git branch         # list all local branches in this repo
3825$ git checkout test  # switch working directory to branch "test"
3826$ git branch new     # create branch "new" starting at current HEAD
3827$ git branch -d new  # delete branch "new"
3828-----------------------------------------------
3829
3830Instead of basing new branch on current HEAD (the default), use:
3831
3832-----------------------------------------------
3833$ git branch new test    # branch named "test"
3834$ git branch new v2.6.15 # tag named v2.6.15
3835$ git branch new HEAD^   # commit before the most recent
3836$ git branch new HEAD^^  # commit before that
3837$ git branch new test~10 # ten commits before tip of branch "test"
3838-----------------------------------------------
3839
3840Create and switch to a new branch at the same time:
3841
3842-----------------------------------------------
3843$ git checkout -b new v2.6.15
3844-----------------------------------------------
3845
3846Update and examine branches from the repository you cloned from:
3847
3848-----------------------------------------------
3849$ git fetch             # update
3850$ git branch -r         # list
3851  origin/master
3852  origin/next
3853  ...
3854$ git checkout -b masterwork origin/master
3855-----------------------------------------------
3856
3857Fetch a branch from a different repository, and give it a new
3858name in your repository:
3859
3860-----------------------------------------------
3861$ git fetch git://example.com/project.git theirbranch:mybranch
3862$ git fetch git://example.com/project.git v2.6.15:mybranch
3863-----------------------------------------------
3864
3865Keep a list of repositories you work with regularly:
3866
3867-----------------------------------------------
3868$ git remote add example git://example.com/project.git
3869$ git remote                    # list remote repositories
3870example
3871origin
3872$ git remote show example       # get details
3873* remote example
3874  URL: git://example.com/project.git
3875  Tracked remote branches
3876    master next ...
3877$ git fetch example             # update branches from example
3878$ git branch -r                 # list all remote branches
3879-----------------------------------------------
3880
3881
3882[[exploring-history]]
3883Exploring history
3884-----------------
3885
3886-----------------------------------------------
3887$ gitk                      # visualize and browse history
3888$ git log                   # list all commits
3889$ git log src/              # ...modifying src/
3890$ git log v2.6.15..v2.6.16  # ...in v2.6.16, not in v2.6.15
3891$ git log master..test      # ...in branch test, not in branch master
3892$ git log test..master      # ...in branch master, but not in test
3893$ git log test...master     # ...in one branch, not in both
3894$ git log -S'foo()'         # ...where difference contain "foo()"
3895$ git log --since="2 weeks ago"
3896$ git log -p                # show patches as well
3897$ git show                  # most recent commit
3898$ git diff v2.6.15..v2.6.16 # diff between two tagged versions
3899$ git diff v2.6.15..HEAD    # diff with current head
3900$ git grep "foo()"          # search working directory for "foo()"
3901$ git grep v2.6.15 "foo()"  # search old tree for "foo()"
3902$ git show v2.6.15:a.txt    # look at old version of a.txt
3903-----------------------------------------------
3904
3905Search for regressions:
3906
3907-----------------------------------------------
3908$ git bisect start
3909$ git bisect bad                # current version is bad
3910$ git bisect good v2.6.13-rc2   # last known good revision
3911Bisecting: 675 revisions left to test after this
3912                                # test here, then:
3913$ git bisect good               # if this revision is good, or
3914$ git bisect bad                # if this revision is bad.
3915                                # repeat until done.
3916-----------------------------------------------
3917
3918[[making-changes]]
3919Making changes
3920--------------
3921
3922Make sure git knows who to blame:
3923
3924------------------------------------------------
3925$ cat >>~/.gitconfig <<\EOF
3926[user]
3927        name = Your Name Comes Here
3928        email = you@yourdomain.example.com
3929EOF
3930------------------------------------------------
3931
3932Select file contents to include in the next commit, then make the
3933commit:
3934
3935-----------------------------------------------
3936$ git add a.txt    # updated file
3937$ git add b.txt    # new file
3938$ git rm c.txt     # old file
3939$ git commit
3940-----------------------------------------------
3941
3942Or, prepare and create the commit in one step:
3943
3944-----------------------------------------------
3945$ git commit d.txt # use latest content only of d.txt
3946$ git commit -a    # use latest content of all tracked files
3947-----------------------------------------------
3948
3949[[merging]]
3950Merging
3951-------
3952
3953-----------------------------------------------
3954$ git merge test   # merge branch "test" into the current branch
3955$ git pull git://example.com/project.git master
3956                   # fetch and merge in remote branch
3957$ git pull . test  # equivalent to git merge test
3958-----------------------------------------------
3959
3960[[sharing-your-changes]]
3961Sharing your changes
3962--------------------
3963
3964Importing or exporting patches:
3965
3966-----------------------------------------------
3967$ git format-patch origin..HEAD # format a patch for each commit
3968                                # in HEAD but not in origin
3969$ git am mbox # import patches from the mailbox "mbox"
3970-----------------------------------------------
3971
3972Fetch a branch in a different git repository, then merge into the
3973current branch:
3974
3975-----------------------------------------------
3976$ git pull git://example.com/project.git theirbranch
3977-----------------------------------------------
3978
3979Store the fetched branch into a local branch before merging into the
3980current branch:
3981
3982-----------------------------------------------
3983$ git pull git://example.com/project.git theirbranch:mybranch
3984-----------------------------------------------
3985
3986After creating commits on a local branch, update the remote
3987branch with your commits:
3988
3989-----------------------------------------------
3990$ git push ssh://example.com/project.git mybranch:theirbranch
3991-----------------------------------------------
3992
3993When remote and local branch are both named "test":
3994
3995-----------------------------------------------
3996$ git push ssh://example.com/project.git test
3997-----------------------------------------------
3998
3999Shortcut version for a frequently used remote repository:
4000
4001-----------------------------------------------
4002$ git remote add example ssh://example.com/project.git
4003$ git push example test
4004-----------------------------------------------
4005
4006[[repository-maintenance]]
4007Repository maintenance
4008----------------------
4009
4010Check for corruption:
4011
4012-----------------------------------------------
4013$ git fsck
4014-----------------------------------------------
4015
4016Recompress, remove unused cruft:
4017
4018-----------------------------------------------
4019$ git gc
4020-----------------------------------------------
4021
4022
4023[[todo]]
4024Appendix B: Notes and todo list for this manual
4025===============================================
4026
4027This is a work in progress.
4028
4029The basic requirements:
4030
4031- It must be readable in order, from beginning to end, by someone
4032  intelligent with a basic grasp of the UNIX command line, but without
4033  any special knowledge of git.  If necessary, any other prerequisites
4034  should be specifically mentioned as they arise.
4035- Whenever possible, section headings should clearly describe the task
4036  they explain how to do, in language that requires no more knowledge
4037  than necessary: for example, "importing patches into a project" rather
4038  than "the git-am command"
4039
4040Think about how to create a clear chapter dependency graph that will
4041allow people to get to important topics without necessarily reading
4042everything in between.
4043
4044Scan Documentation/ for other stuff left out; in particular:
4045
4046- howto's
4047- some of technical/?
4048- hooks
4049- list of commands in gitlink:git[1]
4050
4051Scan email archives for other stuff left out
4052
4053Scan man pages to see if any assume more background than this manual
4054provides.
4055
4056Simplify beginning by suggesting disconnected head instead of
4057temporary branch creation?
4058
4059Add more good examples.  Entire sections of just cookbook examples
4060might be a good idea; maybe make an "advanced examples" section a
4061standard end-of-chapter section?
4062
4063Include cross-references to the glossary, where appropriate.
4064
4065Document shallow clones?  See draft 1.5.0 release notes for some
4066documentation.
4067
4068Add a section on working with other version control systems, including
4069CVS, Subversion, and just imports of series of release tarballs.
4070
4071More details on gitweb?
4072
4073Write a chapter on using plumbing and writing scripts.
4074
4075Alternates, clone -reference, etc.
4076
4077git unpack-objects -r for recovery
4078
4079submodules