Documentation / user-manual.txton commit State the effect of filter-branch on graft explicitly (c6d8f76)
   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 <repo>", 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 linkgit: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 ("git"
  60or "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,
  62called the <<def_working_tree,working tree>>, together with a special
  63top-level directory named ".git", which contains all the information
  64about the history of the project.
  65
  66[[how-to-check-out]]
  67How to check out a different version of a project
  68-------------------------------------------------
  69
  70Git is best thought of as a tool for storing the history of a collection
  71of files.  It stores the history as a compressed collection of
  72interrelated snapshots of the project's contents.  In git each such
  73version is called a <<def_commit,commit>>.
  74
  75Those snapshots aren't necessarily all arranged in a single line from
  76oldest to newest; instead, work may simultaneously proceed along
  77parallel lines of development, called <<def_branch,branches>>, which may
  78merge and diverge.
  79
  80A single git repository can track development on multiple branches.  It
  81does this by keeping a list of <<def_head,heads>> which reference the
  82latest commit on each branch; the linkgit:git-branch[1] command shows
  83you the list of branch heads:
  84
  85------------------------------------------------
  86$ git branch
  87* master
  88------------------------------------------------
  89
  90A freshly cloned repository contains a single branch head, by default
  91named "master", with the working directory initialized to the state of
  92the project referred to by that branch head.
  93
  94Most projects also use <<def_tag,tags>>.  Tags, like heads, are
  95references into the project's history, and can be listed using the
  96linkgit:git-tag[1] command:
  97
  98------------------------------------------------
  99$ git tag -l
 100v2.6.11
 101v2.6.11-tree
 102v2.6.12
 103v2.6.12-rc2
 104v2.6.12-rc3
 105v2.6.12-rc4
 106v2.6.12-rc5
 107v2.6.12-rc6
 108v2.6.13
 109...
 110------------------------------------------------
 111
 112Tags are expected to always point at the same version of a project,
 113while heads are expected to advance as development progresses.
 114
 115Create a new branch head pointing to one of these versions and check it
 116out using linkgit:git-checkout[1]:
 117
 118------------------------------------------------
 119$ git checkout -b new v2.6.13
 120------------------------------------------------
 121
 122The working directory then reflects the contents that the project had
 123when it was tagged v2.6.13, and linkgit:git-branch[1] shows two
 124branches, with an asterisk marking the currently checked-out branch:
 125
 126------------------------------------------------
 127$ git branch
 128  master
 129* new
 130------------------------------------------------
 131
 132If you decide that you'd rather see version 2.6.17, you can modify
 133the current branch to point at v2.6.17 instead, with
 134
 135------------------------------------------------
 136$ git reset --hard v2.6.17
 137------------------------------------------------
 138
 139Note that if the current branch head was your only reference to a
 140particular point in history, then resetting that branch may leave you
 141with no way to find the history it used to point to; so use this command
 142carefully.
 143
 144[[understanding-commits]]
 145Understanding History: Commits
 146------------------------------
 147
 148Every change in the history of a project is represented by a commit.
 149The linkgit:git-show[1] command shows the most recent commit on the
 150current branch:
 151
 152------------------------------------------------
 153$ git show
 154commit 17cf781661e6d38f737f15f53ab552f1e95960d7
 155Author: Linus Torvalds <torvalds@ppc970.osdl.org.(none)>
 156Date:   Tue Apr 19 14:11:06 2005 -0700
 157
 158    Remove duplicate getenv(DB_ENVIRONMENT) call
 159
 160    Noted by Tony Luck.
 161
 162diff --git a/init-db.c b/init-db.c
 163index 65898fa..b002dc6 100644
 164--- a/init-db.c
 165+++ b/init-db.c
 166@@ -7,7 +7,7 @@
 167 
 168 int main(int argc, char **argv)
 169 {
 170-       char *sha1_dir = getenv(DB_ENVIRONMENT), *path;
 171+       char *sha1_dir, *path;
 172        int len, i;
 173 
 174        if (mkdir(".git", 0755) < 0) {
 175------------------------------------------------
 176
 177As you can see, a commit shows who made the latest change, what they
 178did, and why.
 179
 180Every commit has a 40-hexdigit id, sometimes called the "object name" or the
 181"SHA1 id", shown on the first line of the "git-show" output.  You can usually
 182refer to a commit by a shorter name, such as a tag or a branch name, but this
 183longer name can also be useful.  Most importantly, it is a globally unique
 184name for this commit: so if you tell somebody else the object name (for
 185example in email), then you are guaranteed that name will refer to the same
 186commit in their repository that it does in yours (assuming their repository
 187has that commit at all).  Since the object name is computed as a hash over the
 188contents of the commit, you are guaranteed that the commit can never change
 189without its name also changing.
 190
 191In fact, in <<git-concepts>> we shall see that everything stored in git
 192history, including file data and directory contents, is stored in an object
 193with a name that is a hash of its contents.
 194
 195[[understanding-reachability]]
 196Understanding history: commits, parents, and reachability
 197~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 198
 199Every commit (except the very first commit in a project) also has a
 200parent commit which shows what happened before this commit.
 201Following the chain of parents will eventually take you back to the
 202beginning of the project.
 203
 204However, the commits do not form a simple list; git allows lines of
 205development to diverge and then reconverge, and the point where two
 206lines of development reconverge is called a "merge".  The commit
 207representing a merge can therefore have more than one parent, with
 208each parent representing the most recent commit on one of the lines
 209of development leading to that point.
 210
 211The best way to see how this works is using the linkgit:gitk[1]
 212command; running gitk now on a git repository and looking for merge
 213commits will help understand how the git organizes history.
 214
 215In the following, we say that commit X is "reachable" from commit Y
 216if commit X is an ancestor of commit Y.  Equivalently, you could say
 217that Y is a descendant of X, or that there is a chain of parents
 218leading from commit Y to commit X.
 219
 220[[history-diagrams]]
 221Understanding history: History diagrams
 222~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 223
 224We will sometimes represent git history using diagrams like the one
 225below.  Commits are shown as "o", and the links between them with
 226lines drawn with - / and \.  Time goes left to right:
 227
 228
 229................................................
 230         o--o--o <-- Branch A
 231        /
 232 o--o--o <-- master
 233        \
 234         o--o--o <-- Branch B
 235................................................
 236
 237If we need to talk about a particular commit, the character "o" may
 238be replaced with another letter or number.
 239
 240[[what-is-a-branch]]
 241Understanding history: What is a branch?
 242~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 243
 244When we need to be precise, we will use the word "branch" to mean a line
 245of development, and "branch head" (or just "head") to mean a reference
 246to the most recent commit on a branch.  In the example above, the branch
 247head named "A" is a pointer to one particular commit, but we refer to
 248the line of three commits leading up to that point as all being part of
 249"branch A".
 250
 251However, when no confusion will result, we often just use the term
 252"branch" both for branches and for branch heads.
 253
 254[[manipulating-branches]]
 255Manipulating branches
 256---------------------
 257
 258Creating, deleting, and modifying branches is quick and easy; here's
 259a summary of the commands:
 260
 261git branch::
 262        list all branches
 263git branch <branch>::
 264        create a new branch named <branch>, referencing the same
 265        point in history as the current branch
 266git branch <branch> <start-point>::
 267        create a new branch named <branch>, referencing
 268        <start-point>, which may be specified any way you like,
 269        including using a branch name or a tag name
 270git branch -d <branch>::
 271        delete the branch <branch>; if the branch you are deleting
 272        points to a commit which is not reachable from the current
 273        branch, this command will fail with a warning.
 274git branch -D <branch>::
 275        even if the branch points to a commit not reachable
 276        from the current branch, you may know that that commit
 277        is still reachable from some other branch or tag.  In that
 278        case it is safe to use this command to force git to delete
 279        the branch.
 280git checkout <branch>::
 281        make the current branch <branch>, updating the working
 282        directory to reflect the version referenced by <branch>
 283git checkout -b <new> <start-point>::
 284        create a new branch <new> referencing <start-point>, and
 285        check it out.
 286
 287The special symbol "HEAD" can always be used to refer to the current
 288branch.  In fact, git uses a file named "HEAD" in the .git directory to
 289remember which branch is current:
 290
 291------------------------------------------------
 292$ cat .git/HEAD
 293ref: refs/heads/master
 294------------------------------------------------
 295
 296[[detached-head]]
 297Examining an old version without creating a new branch
 298------------------------------------------------------
 299
 300The git-checkout command normally expects a branch head, but will also
 301accept an arbitrary commit; for example, you can check out the commit
 302referenced by a tag:
 303
 304------------------------------------------------
 305$ git checkout v2.6.17
 306Note: moving to "v2.6.17" which isn't a local branch
 307If you want to create a new branch from this checkout, you may do so
 308(now or later) by using -b with the checkout command again. Example:
 309  git checkout -b <new_branch_name>
 310HEAD is now at 427abfa... Linux v2.6.17
 311------------------------------------------------
 312
 313The HEAD then refers to the SHA1 of the commit instead of to a branch,
 314and git branch shows that you are no longer on a branch:
 315
 316------------------------------------------------
 317$ cat .git/HEAD
 318427abfa28afedffadfca9dd8b067eb6d36bac53f
 319$ git branch
 320* (no branch)
 321  master
 322------------------------------------------------
 323
 324In this case we say that the HEAD is "detached".
 325
 326This is an easy way to check out a particular version without having to
 327make up a name for the new branch.   You can still create a new branch
 328(or tag) for this version later if you decide to.
 329
 330[[examining-remote-branches]]
 331Examining branches from a remote repository
 332-------------------------------------------
 333
 334The "master" branch that was created at the time you cloned is a copy
 335of the HEAD in the repository that you cloned from.  That repository
 336may also have had other branches, though, and your local repository
 337keeps branches which track each of those remote branches, which you
 338can view using the "-r" option to linkgit:git-branch[1]:
 339
 340------------------------------------------------
 341$ git branch -r
 342  origin/HEAD
 343  origin/html
 344  origin/maint
 345  origin/man
 346  origin/master
 347  origin/next
 348  origin/pu
 349  origin/todo
 350------------------------------------------------
 351
 352You cannot check out these remote-tracking branches, but you can
 353examine them on a branch of your own, just as you would a tag:
 354
 355------------------------------------------------
 356$ git checkout -b my-todo-copy origin/todo
 357------------------------------------------------
 358
 359Note that the name "origin" is just the name that git uses by default
 360to refer to the repository that you cloned from.
 361
 362[[how-git-stores-references]]
 363Naming branches, tags, and other references
 364-------------------------------------------
 365
 366Branches, remote-tracking branches, and tags are all references to
 367commits.  All references are named with a slash-separated path name
 368starting with "refs"; the names we've been using so far are actually
 369shorthand:
 370
 371        - The branch "test" is short for "refs/heads/test".
 372        - The tag "v2.6.18" is short for "refs/tags/v2.6.18".
 373        - "origin/master" is short for "refs/remotes/origin/master".
 374
 375The full name is occasionally useful if, for example, there ever
 376exists a tag and a branch with the same name.
 377
 378(Newly created refs are actually stored in the .git/refs directory,
 379under the path given by their name.  However, for efficiency reasons
 380they may also be packed together in a single file; see
 381linkgit:git-pack-refs[1]).
 382
 383As another useful shortcut, the "HEAD" of a repository can be referred
 384to just using the name of that repository.  So, for example, "origin"
 385is usually a shortcut for the HEAD branch in the repository "origin".
 386
 387For the complete list of paths which git checks for references, and
 388the order it uses to decide which to choose when there are multiple
 389references with the same shorthand name, see the "SPECIFYING
 390REVISIONS" section of linkgit:git-rev-parse[1].
 391
 392[[Updating-a-repository-With-git-fetch]]
 393Updating a repository with git-fetch
 394------------------------------------
 395
 396Eventually the developer cloned from will do additional work in her
 397repository, creating new commits and advancing the branches to point
 398at the new commits.
 399
 400The command "git fetch", with no arguments, will update all of the
 401remote-tracking branches to the latest version found in her
 402repository.  It will not touch any of your own branches--not even the
 403"master" branch that was created for you on clone.
 404
 405[[fetching-branches]]
 406Fetching branches from other repositories
 407-----------------------------------------
 408
 409You can also track branches from repositories other than the one you
 410cloned from, using linkgit:git-remote[1]:
 411
 412-------------------------------------------------
 413$ git remote add linux-nfs git://linux-nfs.org/pub/nfs-2.6.git
 414$ git fetch linux-nfs
 415* refs/remotes/linux-nfs/master: storing branch 'master' ...
 416  commit: bf81b46
 417-------------------------------------------------
 418
 419New remote-tracking branches will be stored under the shorthand name
 420that you gave "git-remote add", in this case linux-nfs:
 421
 422-------------------------------------------------
 423$ git branch -r
 424linux-nfs/master
 425origin/master
 426-------------------------------------------------
 427
 428If you run "git fetch <remote>" later, the tracking branches for the
 429named <remote> will be updated.
 430
 431If you examine the file .git/config, you will see that git has added
 432a new stanza:
 433
 434-------------------------------------------------
 435$ cat .git/config
 436...
 437[remote "linux-nfs"]
 438        url = git://linux-nfs.org/pub/nfs-2.6.git
 439        fetch = +refs/heads/*:refs/remotes/linux-nfs/*
 440...
 441-------------------------------------------------
 442
 443This is what causes git to track the remote's branches; you may modify
 444or delete these configuration options by editing .git/config with a
 445text editor.  (See the "CONFIGURATION FILE" section of
 446linkgit:git-config[1] for details.)
 447
 448[[exploring-git-history]]
 449Exploring git history
 450=====================
 451
 452Git is best thought of as a tool for storing the history of a
 453collection of files.  It does this by storing compressed snapshots of
 454the contents of a file hierarchy, together with "commits" which show
 455the relationships between these snapshots.
 456
 457Git provides extremely flexible and fast tools for exploring the
 458history of a project.
 459
 460We start with one specialized tool that is useful for finding the
 461commit that introduced a bug into a project.
 462
 463[[using-bisect]]
 464How to use bisect to find a regression
 465--------------------------------------
 466
 467Suppose version 2.6.18 of your project worked, but the version at
 468"master" crashes.  Sometimes the best way to find the cause of such a
 469regression is to perform a brute-force search through the project's
 470history to find the particular commit that caused the problem.  The
 471linkgit:git-bisect[1] command can help you do this:
 472
 473-------------------------------------------------
 474$ git bisect start
 475$ git bisect good v2.6.18
 476$ git bisect bad master
 477Bisecting: 3537 revisions left to test after this
 478[65934a9a028b88e83e2b0f8b36618fe503349f8e] BLOCK: Make USB storage depend on SCSI rather than selecting it [try #6]
 479-------------------------------------------------
 480
 481If you run "git branch" at this point, you'll see that git has
 482temporarily moved you in "(no branch)". HEAD is now detached from any
 483branch and points directly to a commit (with commit id 65934...) that
 484is reachable from "master" but not from v2.6.18. Compile and test it,
 485and see whether it crashes. Assume it does crash. Then:
 486
 487-------------------------------------------------
 488$ git bisect bad
 489Bisecting: 1769 revisions left to test after this
 490[7eff82c8b1511017ae605f0c99ac275a7e21b867] i2c-core: Drop useless bitmaskings
 491-------------------------------------------------
 492
 493checks out an older version.  Continue like this, telling git at each
 494stage whether the version it gives you is good or bad, and notice
 495that the number of revisions left to test is cut approximately in
 496half each time.
 497
 498After about 13 tests (in this case), it will output the commit id of
 499the guilty commit.  You can then examine the commit with
 500linkgit:git-show[1], find out who wrote it, and mail them your bug
 501report with the commit id.  Finally, run
 502
 503-------------------------------------------------
 504$ git bisect reset
 505-------------------------------------------------
 506
 507to return you to the branch you were on before.
 508
 509Note that the version which git-bisect checks out for you at each
 510point is just a suggestion, and you're free to try a different
 511version if you think it would be a good idea.  For example,
 512occasionally you may land on a commit that broke something unrelated;
 513run
 514
 515-------------------------------------------------
 516$ git bisect visualize
 517-------------------------------------------------
 518
 519which will run gitk and label the commit it chose with a marker that
 520says "bisect".  Choose a safe-looking commit nearby, note its commit
 521id, and check it out with:
 522
 523-------------------------------------------------
 524$ git reset --hard fb47ddb2db...
 525-------------------------------------------------
 526
 527then test, run "bisect good" or "bisect bad" as appropriate, and
 528continue.
 529
 530Instead of "git bisect visualize" and then "git reset --hard
 531fb47ddb2db...", you might just want to tell git that you want to skip
 532the current commit:
 533
 534-------------------------------------------------
 535$ git bisect skip
 536-------------------------------------------------
 537
 538In this case, though, git may not eventually be able to tell the first
 539bad one between some first skipped commits and a later bad commit.
 540
 541There are also ways to automate the bisecting process if you have a
 542test script that can tell a good from a bad commit. See
 543linkgit:git-bisect[1] for more information about this and other "git
 544bisect" features.
 545
 546[[naming-commits]]
 547Naming commits
 548--------------
 549
 550We have seen several ways of naming commits already:
 551
 552        - 40-hexdigit object name
 553        - branch name: refers to the commit at the head of the given
 554          branch
 555        - tag name: refers to the commit pointed to by the given tag
 556          (we've seen branches and tags are special cases of
 557          <<how-git-stores-references,references>>).
 558        - HEAD: refers to the head of the current branch
 559
 560There are many more; see the "SPECIFYING REVISIONS" section of the
 561linkgit:git-rev-parse[1] man page for the complete list of ways to
 562name revisions.  Some examples:
 563
 564-------------------------------------------------
 565$ git show fb47ddb2 # the first few characters of the object name
 566                    # are usually enough to specify it uniquely
 567$ git show HEAD^    # the parent of the HEAD commit
 568$ git show HEAD^^   # the grandparent
 569$ git show HEAD~4   # the great-great-grandparent
 570-------------------------------------------------
 571
 572Recall that merge commits may have more than one parent; by default,
 573^ and ~ follow the first parent listed in the commit, but you can
 574also choose:
 575
 576-------------------------------------------------
 577$ git show HEAD^1   # show the first parent of HEAD
 578$ git show HEAD^2   # show the second parent of HEAD
 579-------------------------------------------------
 580
 581In addition to HEAD, there are several other special names for
 582commits:
 583
 584Merges (to be discussed later), as well as operations such as
 585git-reset, which change the currently checked-out commit, generally
 586set ORIG_HEAD to the value HEAD had before the current operation.
 587
 588The git-fetch operation always stores the head of the last fetched
 589branch in FETCH_HEAD.  For example, if you run git fetch without
 590specifying a local branch as the target of the operation
 591
 592-------------------------------------------------
 593$ git fetch git://example.com/proj.git theirbranch
 594-------------------------------------------------
 595
 596the fetched commits will still be available from FETCH_HEAD.
 597
 598When we discuss merges we'll also see the special name MERGE_HEAD,
 599which refers to the other branch that we're merging in to the current
 600branch.
 601
 602The linkgit:git-rev-parse[1] command is a low-level command that is
 603occasionally useful for translating some name for a commit to the object
 604name for that commit:
 605
 606-------------------------------------------------
 607$ git rev-parse origin
 608e05db0fd4f31dde7005f075a84f96b360d05984b
 609-------------------------------------------------
 610
 611[[creating-tags]]
 612Creating tags
 613-------------
 614
 615We can also create a tag to refer to a particular commit; after
 616running
 617
 618-------------------------------------------------
 619$ git tag stable-1 1b2e1d63ff
 620-------------------------------------------------
 621
 622You can use stable-1 to refer to the commit 1b2e1d63ff.
 623
 624This creates a "lightweight" tag.  If you would also like to include a
 625comment with the tag, and possibly sign it cryptographically, then you
 626should create a tag object instead; see the linkgit:git-tag[1] man page
 627for details.
 628
 629[[browsing-revisions]]
 630Browsing revisions
 631------------------
 632
 633The linkgit:git-log[1] command can show lists of commits.  On its
 634own, it shows all commits reachable from the parent commit; but you
 635can also make more specific requests:
 636
 637-------------------------------------------------
 638$ git log v2.5..        # commits since (not reachable from) v2.5
 639$ git log test..master  # commits reachable from master but not test
 640$ git log master..test  # ...reachable from test but not master
 641$ git log master...test # ...reachable from either test or master,
 642                        #    but not both
 643$ git log --since="2 weeks ago" # commits from the last 2 weeks
 644$ git log Makefile      # commits which modify Makefile
 645$ git log fs/           # ... which modify any file under fs/
 646$ git log -S'foo()'     # commits which add or remove any file data
 647                        # matching the string 'foo()'
 648-------------------------------------------------
 649
 650And of course you can combine all of these; the following finds
 651commits since v2.5 which touch the Makefile or any file under fs:
 652
 653-------------------------------------------------
 654$ git log v2.5.. Makefile fs/
 655-------------------------------------------------
 656
 657You can also ask git log to show patches:
 658
 659-------------------------------------------------
 660$ git log -p
 661-------------------------------------------------
 662
 663See the "--pretty" option in the linkgit:git-log[1] man page for more
 664display options.
 665
 666Note that git log starts with the most recent commit and works
 667backwards through the parents; however, since git history can contain
 668multiple independent lines of development, the particular order that
 669commits are listed in may be somewhat arbitrary.
 670
 671[[generating-diffs]]
 672Generating diffs
 673----------------
 674
 675You can generate diffs between any two versions using
 676linkgit:git-diff[1]:
 677
 678-------------------------------------------------
 679$ git diff master..test
 680-------------------------------------------------
 681
 682That will produce the diff between the tips of the two branches.  If
 683you'd prefer to find the diff from their common ancestor to test, you
 684can use three dots instead of two:
 685
 686-------------------------------------------------
 687$ git diff master...test
 688-------------------------------------------------
 689
 690Sometimes what you want instead is a set of patches; for this you can
 691use linkgit:git-format-patch[1]:
 692
 693-------------------------------------------------
 694$ git format-patch master..test
 695-------------------------------------------------
 696
 697will generate a file with a patch for each commit reachable from test
 698but not from master.
 699
 700[[viewing-old-file-versions]]
 701Viewing old file versions
 702-------------------------
 703
 704You can always view an old version of a file by just checking out the
 705correct revision first.  But sometimes it is more convenient to be
 706able to view an old version of a single file without checking
 707anything out; this command does that:
 708
 709-------------------------------------------------
 710$ git show v2.5:fs/locks.c
 711-------------------------------------------------
 712
 713Before the colon may be anything that names a commit, and after it
 714may be any path to a file tracked by git.
 715
 716[[history-examples]]
 717Examples
 718--------
 719
 720[[counting-commits-on-a-branch]]
 721Counting the number of commits on a branch
 722~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 723
 724Suppose you want to know how many commits you've made on "mybranch"
 725since it diverged from "origin":
 726
 727-------------------------------------------------
 728$ git log --pretty=oneline origin..mybranch | wc -l
 729-------------------------------------------------
 730
 731Alternatively, you may often see this sort of thing done with the
 732lower-level command linkgit:git-rev-list[1], which just lists the SHA1's
 733of all the given commits:
 734
 735-------------------------------------------------
 736$ git rev-list origin..mybranch | wc -l
 737-------------------------------------------------
 738
 739[[checking-for-equal-branches]]
 740Check whether two branches point at the same history
 741~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 742
 743Suppose you want to check whether two branches point at the same point
 744in history.
 745
 746-------------------------------------------------
 747$ git diff origin..master
 748-------------------------------------------------
 749
 750will tell you whether the contents of the project are the same at the
 751two branches; in theory, however, it's possible that the same project
 752contents could have been arrived at by two different historical
 753routes.  You could compare the object names:
 754
 755-------------------------------------------------
 756$ git rev-list origin
 757e05db0fd4f31dde7005f075a84f96b360d05984b
 758$ git rev-list master
 759e05db0fd4f31dde7005f075a84f96b360d05984b
 760-------------------------------------------------
 761
 762Or you could recall that the ... operator selects all commits
 763contained reachable from either one reference or the other but not
 764both: so
 765
 766-------------------------------------------------
 767$ git log origin...master
 768-------------------------------------------------
 769
 770will return no commits when the two branches are equal.
 771
 772[[finding-tagged-descendants]]
 773Find first tagged version including a given fix
 774~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 775
 776Suppose you know that the commit e05db0fd fixed a certain problem.
 777You'd like to find the earliest tagged release that contains that
 778fix.
 779
 780Of course, there may be more than one answer--if the history branched
 781after commit e05db0fd, then there could be multiple "earliest" tagged
 782releases.
 783
 784You could just visually inspect the commits since e05db0fd:
 785
 786-------------------------------------------------
 787$ gitk e05db0fd..
 788-------------------------------------------------
 789
 790Or you can use linkgit:git-name-rev[1], which will give the commit a
 791name based on any tag it finds pointing to one of the commit's
 792descendants:
 793
 794-------------------------------------------------
 795$ git name-rev --tags e05db0fd
 796e05db0fd tags/v1.5.0-rc1^0~23
 797-------------------------------------------------
 798
 799The linkgit:git-describe[1] command does the opposite, naming the
 800revision using a tag on which the given commit is based:
 801
 802-------------------------------------------------
 803$ git describe e05db0fd
 804v1.5.0-rc0-260-ge05db0f
 805-------------------------------------------------
 806
 807but that may sometimes help you guess which tags might come after the
 808given commit.
 809
 810If you just want to verify whether a given tagged version contains a
 811given commit, you could use linkgit:git-merge-base[1]:
 812
 813-------------------------------------------------
 814$ git merge-base e05db0fd v1.5.0-rc1
 815e05db0fd4f31dde7005f075a84f96b360d05984b
 816-------------------------------------------------
 817
 818The merge-base command finds a common ancestor of the given commits,
 819and always returns one or the other in the case where one is a
 820descendant of the other; so the above output shows that e05db0fd
 821actually is an ancestor of v1.5.0-rc1.
 822
 823Alternatively, note that
 824
 825-------------------------------------------------
 826$ git log v1.5.0-rc1..e05db0fd
 827-------------------------------------------------
 828
 829will produce empty output if and only if v1.5.0-rc1 includes e05db0fd,
 830because it outputs only commits that are not reachable from v1.5.0-rc1.
 831
 832As yet another alternative, the linkgit:git-show-branch[1] command lists
 833the commits reachable from its arguments with a display on the left-hand
 834side that indicates which arguments that commit is reachable from.  So,
 835you can run something like
 836
 837-------------------------------------------------
 838$ git show-branch e05db0fd v1.5.0-rc0 v1.5.0-rc1 v1.5.0-rc2
 839! [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
 840available
 841 ! [v1.5.0-rc0] GIT v1.5.0 preview
 842  ! [v1.5.0-rc1] GIT v1.5.0-rc1
 843   ! [v1.5.0-rc2] GIT v1.5.0-rc2
 844...
 845-------------------------------------------------
 846
 847then search for a line that looks like
 848
 849-------------------------------------------------
 850+ ++ [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if
 851available
 852-------------------------------------------------
 853
 854Which shows that e05db0fd is reachable from itself, from v1.5.0-rc1, and
 855from v1.5.0-rc2, but not from v1.5.0-rc0.
 856
 857[[showing-commits-unique-to-a-branch]]
 858Showing commits unique to a given branch
 859~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 860
 861Suppose you would like to see all the commits reachable from the branch
 862head named "master" but not from any other head in your repository.
 863
 864We can list all the heads in this repository with
 865linkgit:git-show-ref[1]:
 866
 867-------------------------------------------------
 868$ git show-ref --heads
 869bf62196b5e363d73353a9dcf094c59595f3153b7 refs/heads/core-tutorial
 870db768d5504c1bb46f63ee9d6e1772bd047e05bf9 refs/heads/maint
 871a07157ac624b2524a059a3414e99f6f44bebc1e7 refs/heads/master
 87224dbc180ea14dc1aebe09f14c8ecf32010690627 refs/heads/tutorial-2
 8731e87486ae06626c2f31eaa63d26fc0fd646c8af2 refs/heads/tutorial-fixes
 874-------------------------------------------------
 875
 876We can get just the branch-head names, and remove "master", with
 877the help of the standard utilities cut and grep:
 878
 879-------------------------------------------------
 880$ git show-ref --heads | cut -d' ' -f2 | grep -v '^refs/heads/master'
 881refs/heads/core-tutorial
 882refs/heads/maint
 883refs/heads/tutorial-2
 884refs/heads/tutorial-fixes
 885-------------------------------------------------
 886
 887And then we can ask to see all the commits reachable from master
 888but not from these other heads:
 889
 890-------------------------------------------------
 891$ gitk master --not $( git show-ref --heads | cut -d' ' -f2 |
 892                                grep -v '^refs/heads/master' )
 893-------------------------------------------------
 894
 895Obviously, endless variations are possible; for example, to see all
 896commits reachable from some head but not from any tag in the repository:
 897
 898-------------------------------------------------
 899$ gitk $( git show-ref --heads ) --not  $( git show-ref --tags )
 900-------------------------------------------------
 901
 902(See linkgit:git-rev-parse[1] for explanations of commit-selecting
 903syntax such as `--not`.)
 904
 905[[making-a-release]]
 906Creating a changelog and tarball for a software release
 907~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 908
 909The linkgit:git-archive[1] command can create a tar or zip archive from
 910any version of a project; for example:
 911
 912-------------------------------------------------
 913$ git archive --format=tar --prefix=project/ HEAD | gzip >latest.tar.gz
 914-------------------------------------------------
 915
 916will use HEAD to produce a tar archive in which each filename is
 917preceded by "project/".
 918
 919If you're releasing a new version of a software project, you may want
 920to simultaneously make a changelog to include in the release
 921announcement.
 922
 923Linus Torvalds, for example, makes new kernel releases by tagging them,
 924then running:
 925
 926-------------------------------------------------
 927$ release-script 2.6.12 2.6.13-rc6 2.6.13-rc7
 928-------------------------------------------------
 929
 930where release-script is a shell script that looks like:
 931
 932-------------------------------------------------
 933#!/bin/sh
 934stable="$1"
 935last="$2"
 936new="$3"
 937echo "# git tag v$new"
 938echo "git archive --prefix=linux-$new/ v$new | gzip -9 > ../linux-$new.tar.gz"
 939echo "git diff v$stable v$new | gzip -9 > ../patch-$new.gz"
 940echo "git log --no-merges v$new ^v$last > ../ChangeLog-$new"
 941echo "git shortlog --no-merges v$new ^v$last > ../ShortLog"
 942echo "git diff --stat --summary -M v$last v$new > ../diffstat-$new"
 943-------------------------------------------------
 944
 945and then he just cut-and-pastes the output commands after verifying that
 946they look OK.
 947
 948[[Finding-comments-With-given-Content]]
 949Finding commits referencing a file with given content
 950~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 951
 952Somebody hands you a copy of a file, and asks which commits modified a
 953file such that it contained the given content either before or after the
 954commit.  You can find out with this:
 955
 956-------------------------------------------------
 957$  git log --raw --abbrev=40 --pretty=oneline |
 958        grep -B 1 `git hash-object filename`
 959-------------------------------------------------
 960
 961Figuring out why this works is left as an exercise to the (advanced)
 962student.  The linkgit:git-log[1], linkgit:git-diff-tree[1], and
 963linkgit:git-hash-object[1] man pages may prove helpful.
 964
 965[[Developing-With-git]]
 966Developing with git
 967===================
 968
 969[[telling-git-your-name]]
 970Telling git your name
 971---------------------
 972
 973Before creating any commits, you should introduce yourself to git.  The
 974easiest way to do so is to make sure the following lines appear in a
 975file named .gitconfig in your home directory:
 976
 977------------------------------------------------
 978[user]
 979        name = Your Name Comes Here
 980        email = you@yourdomain.example.com
 981------------------------------------------------
 982
 983(See the "CONFIGURATION FILE" section of linkgit:git-config[1] for
 984details on the configuration file.)
 985
 986
 987[[creating-a-new-repository]]
 988Creating a new repository
 989-------------------------
 990
 991Creating a new repository from scratch is very easy:
 992
 993-------------------------------------------------
 994$ mkdir project
 995$ cd project
 996$ git init
 997-------------------------------------------------
 998
 999If you have some initial content (say, a tarball):
1000
1001-------------------------------------------------
1002$ tar -xzvf project.tar.gz
1003$ cd project
1004$ git init
1005$ git add . # include everything below ./ in the first commit:
1006$ git commit
1007-------------------------------------------------
1008
1009[[how-to-make-a-commit]]
1010How to make a commit
1011--------------------
1012
1013Creating a new commit takes three steps:
1014
1015        1. Making some changes to the working directory using your
1016           favorite editor.
1017        2. Telling git about your changes.
1018        3. Creating the commit using the content you told git about
1019           in step 2.
1020
1021In practice, you can interleave and repeat steps 1 and 2 as many
1022times as you want: in order to keep track of what you want committed
1023at step 3, git maintains a snapshot of the tree's contents in a
1024special staging area called "the index."
1025
1026At the beginning, the content of the index will be identical to
1027that of the HEAD.  The command "git diff --cached", which shows
1028the difference between the HEAD and the index, should therefore
1029produce no output at that point.
1030
1031Modifying the index is easy:
1032
1033To update the index with the new contents of a modified file, use
1034
1035-------------------------------------------------
1036$ git add path/to/file
1037-------------------------------------------------
1038
1039To add the contents of a new file to the index, use
1040
1041-------------------------------------------------
1042$ git add path/to/file
1043-------------------------------------------------
1044
1045To remove a file from the index and from the working tree,
1046
1047-------------------------------------------------
1048$ git rm path/to/file
1049-------------------------------------------------
1050
1051After each step you can verify that
1052
1053-------------------------------------------------
1054$ git diff --cached
1055-------------------------------------------------
1056
1057always shows the difference between the HEAD and the index file--this
1058is what you'd commit if you created the commit now--and that
1059
1060-------------------------------------------------
1061$ git diff
1062-------------------------------------------------
1063
1064shows the difference between the working tree and the index file.
1065
1066Note that "git-add" always adds just the current contents of a file
1067to the index; further changes to the same file will be ignored unless
1068you run git-add on the file again.
1069
1070When you're ready, just run
1071
1072-------------------------------------------------
1073$ git commit
1074-------------------------------------------------
1075
1076and git will prompt you for a commit message and then create the new
1077commit.  Check to make sure it looks like what you expected with
1078
1079-------------------------------------------------
1080$ git show
1081-------------------------------------------------
1082
1083As a special shortcut,
1084
1085-------------------------------------------------
1086$ git commit -a
1087-------------------------------------------------
1088
1089will update the index with any files that you've modified or removed
1090and create a commit, all in one step.
1091
1092A number of commands are useful for keeping track of what you're
1093about to commit:
1094
1095-------------------------------------------------
1096$ git diff --cached # difference between HEAD and the index; what
1097                    # would be committed if you ran "commit" now.
1098$ git diff          # difference between the index file and your
1099                    # working directory; changes that would not
1100                    # be included if you ran "commit" now.
1101$ git diff HEAD     # difference between HEAD and working tree; what
1102                    # would be committed if you ran "commit -a" now.
1103$ git status        # a brief per-file summary of the above.
1104-------------------------------------------------
1105
1106You can also use linkgit:git-gui[1] to create commits, view changes in
1107the index and the working tree files, and individually select diff hunks
1108for inclusion in the index (by right-clicking on the diff hunk and
1109choosing "Stage Hunk For Commit").
1110
1111[[creating-good-commit-messages]]
1112Creating good commit messages
1113-----------------------------
1114
1115Though not required, it's a good idea to begin the commit message
1116with a single short (less than 50 character) line summarizing the
1117change, followed by a blank line and then a more thorough
1118description.  Tools that turn commits into email, for example, use
1119the first line on the Subject line and the rest of the commit in the
1120body.
1121
1122[[ignoring-files]]
1123Ignoring files
1124--------------
1125
1126A project will often generate files that you do 'not' want to track with git.
1127This typically includes files generated by a build process or temporary
1128backup files made by your editor. Of course, 'not' tracking files with git
1129is just a matter of 'not' calling "`git-add`" on them. But it quickly becomes
1130annoying to have these untracked files lying around; e.g. they make
1131"`git add .`" practically useless, and they keep showing up in the output of
1132"`git status`".
1133
1134You can tell git to ignore certain files by creating a file called .gitignore
1135in the top level of your working directory, with contents such as:
1136
1137-------------------------------------------------
1138# Lines starting with '#' are considered comments.
1139# Ignore any file named foo.txt.
1140foo.txt
1141# Ignore (generated) html files,
1142*.html
1143# except foo.html which is maintained by hand.
1144!foo.html
1145# Ignore objects and archives.
1146*.[oa]
1147-------------------------------------------------
1148
1149See linkgit:gitignore[5] for a detailed explanation of the syntax.  You can
1150also place .gitignore files in other directories in your working tree, and they
1151will apply to those directories and their subdirectories.  The `.gitignore`
1152files can be added to your repository like any other files (just run `git add
1153.gitignore` and `git commit`, as usual), which is convenient when the exclude
1154patterns (such as patterns matching build output files) would also make sense
1155for other users who clone your repository.
1156
1157If you wish the exclude patterns to affect only certain repositories
1158(instead of every repository for a given project), you may instead put
1159them in a file in your repository named .git/info/exclude, or in any file
1160specified by the `core.excludesfile` configuration variable.  Some git
1161commands can also take exclude patterns directly on the command line.
1162See linkgit:gitignore[5] for the details.
1163
1164[[how-to-merge]]
1165How to merge
1166------------
1167
1168You can rejoin two diverging branches of development using
1169linkgit:git-merge[1]:
1170
1171-------------------------------------------------
1172$ git merge branchname
1173-------------------------------------------------
1174
1175merges the development in the branch "branchname" into the current
1176branch.  If there are conflicts--for example, if the same file is
1177modified in two different ways in the remote branch and the local
1178branch--then you are warned; the output may look something like this:
1179
1180-------------------------------------------------
1181$ git merge next
1182 100% (4/4) done
1183Auto-merged file.txt
1184CONFLICT (content): Merge conflict in file.txt
1185Automatic merge failed; fix conflicts and then commit the result.
1186-------------------------------------------------
1187
1188Conflict markers are left in the problematic files, and after
1189you resolve the conflicts manually, you can update the index
1190with the contents and run git commit, as you normally would when
1191creating a new file.
1192
1193If you examine the resulting commit using gitk, you will see that it
1194has two parents, one pointing to the top of the current branch, and
1195one to the top of the other branch.
1196
1197[[resolving-a-merge]]
1198Resolving a merge
1199-----------------
1200
1201When a merge isn't resolved automatically, git leaves the index and
1202the working tree in a special state that gives you all the
1203information you need to help resolve the merge.
1204
1205Files with conflicts are marked specially in the index, so until you
1206resolve the problem and update the index, linkgit:git-commit[1] will
1207fail:
1208
1209-------------------------------------------------
1210$ git commit
1211file.txt: needs merge
1212-------------------------------------------------
1213
1214Also, linkgit:git-status[1] will list those files as "unmerged", and the
1215files with conflicts will have conflict markers added, like this:
1216
1217-------------------------------------------------
1218<<<<<<< HEAD:file.txt
1219Hello world
1220=======
1221Goodbye
1222>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1223-------------------------------------------------
1224
1225All you need to do is edit the files to resolve the conflicts, and then
1226
1227-------------------------------------------------
1228$ git add file.txt
1229$ git commit
1230-------------------------------------------------
1231
1232Note that the commit message will already be filled in for you with
1233some information about the merge.  Normally you can just use this
1234default message unchanged, but you may add additional commentary of
1235your own if desired.
1236
1237The above is all you need to know to resolve a simple merge.  But git
1238also provides more information to help resolve conflicts:
1239
1240[[conflict-resolution]]
1241Getting conflict-resolution help during a merge
1242~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1243
1244All of the changes that git was able to merge automatically are
1245already added to the index file, so linkgit:git-diff[1] shows only
1246the conflicts.  It uses an unusual syntax:
1247
1248-------------------------------------------------
1249$ git diff
1250diff --cc file.txt
1251index 802992c,2b60207..0000000
1252--- a/file.txt
1253+++ b/file.txt
1254@@@ -1,1 -1,1 +1,5 @@@
1255++<<<<<<< HEAD:file.txt
1256 +Hello world
1257++=======
1258+ Goodbye
1259++>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt
1260-------------------------------------------------
1261
1262Recall that the commit which will be committed after we resolve this
1263conflict will have two parents instead of the usual one: one parent
1264will be HEAD, the tip of the current branch; the other will be the
1265tip of the other branch, which is stored temporarily in MERGE_HEAD.
1266
1267During the merge, the index holds three versions of each file.  Each of
1268these three "file stages" represents a different version of the file:
1269
1270-------------------------------------------------
1271$ git show :1:file.txt  # the file in a common ancestor of both branches
1272$ git show :2:file.txt  # the version from HEAD.
1273$ git show :3:file.txt  # the version from MERGE_HEAD.
1274-------------------------------------------------
1275
1276When you ask linkgit:git-diff[1] to show the conflicts, it runs a
1277three-way diff between the conflicted merge results in the work tree with
1278stages 2 and 3 to show only hunks whose contents come from both sides,
1279mixed (in other words, when a hunk's merge results come only from stage 2,
1280that part is not conflicting and is not shown.  Same for stage 3).
1281
1282The diff above shows the differences between the working-tree version of
1283file.txt and the stage 2 and stage 3 versions.  So instead of preceding
1284each line by a single "+" or "-", it now uses two columns: the first
1285column is used for differences between the first parent and the working
1286directory copy, and the second for differences between the second parent
1287and the working directory copy.  (See the "COMBINED DIFF FORMAT" section
1288of linkgit:git-diff-files[1] for a details of the format.)
1289
1290After resolving the conflict in the obvious way (but before updating the
1291index), the diff will look like:
1292
1293-------------------------------------------------
1294$ git diff
1295diff --cc file.txt
1296index 802992c,2b60207..0000000
1297--- a/file.txt
1298+++ b/file.txt
1299@@@ -1,1 -1,1 +1,1 @@@
1300- Hello world
1301 -Goodbye
1302++Goodbye world
1303-------------------------------------------------
1304
1305This shows that our resolved version deleted "Hello world" from the
1306first parent, deleted "Goodbye" from the second parent, and added
1307"Goodbye world", which was previously absent from both.
1308
1309Some special diff options allow diffing the working directory against
1310any of these stages:
1311
1312-------------------------------------------------
1313$ git diff -1 file.txt          # diff against stage 1
1314$ git diff --base file.txt      # same as the above
1315$ git diff -2 file.txt          # diff against stage 2
1316$ git diff --ours file.txt      # same as the above
1317$ git diff -3 file.txt          # diff against stage 3
1318$ git diff --theirs file.txt    # same as the above.
1319-------------------------------------------------
1320
1321The linkgit:git-log[1] and linkgit:gitk[1] commands also provide special help
1322for merges:
1323
1324-------------------------------------------------
1325$ git log --merge
1326$ gitk --merge
1327-------------------------------------------------
1328
1329These will display all commits which exist only on HEAD or on
1330MERGE_HEAD, and which touch an unmerged file.
1331
1332You may also use linkgit:git-mergetool[1], which lets you merge the
1333unmerged files using external tools such as emacs or kdiff3.
1334
1335Each time you resolve the conflicts in a file and update the index:
1336
1337-------------------------------------------------
1338$ git add file.txt
1339-------------------------------------------------
1340
1341the different stages of that file will be "collapsed", after which
1342git-diff will (by default) no longer show diffs for that file.
1343
1344[[undoing-a-merge]]
1345Undoing a merge
1346---------------
1347
1348If you get stuck and decide to just give up and throw the whole mess
1349away, you can always return to the pre-merge state with
1350
1351-------------------------------------------------
1352$ git reset --hard HEAD
1353-------------------------------------------------
1354
1355Or, if you've already committed the merge that you want to throw away,
1356
1357-------------------------------------------------
1358$ git reset --hard ORIG_HEAD
1359-------------------------------------------------
1360
1361However, this last command can be dangerous in some cases--never
1362throw away a commit you have already committed if that commit may
1363itself have been merged into another branch, as doing so may confuse
1364further merges.
1365
1366[[fast-forwards]]
1367Fast-forward merges
1368-------------------
1369
1370There is one special case not mentioned above, which is treated
1371differently.  Normally, a merge results in a merge commit, with two
1372parents, one pointing at each of the two lines of development that
1373were merged.
1374
1375However, if the current branch is a descendant of the other--so every
1376commit present in the one is already contained in the other--then git
1377just performs a "fast forward"; the head of the current branch is moved
1378forward to point at the head of the merged-in branch, without any new
1379commits being created.
1380
1381[[fixing-mistakes]]
1382Fixing mistakes
1383---------------
1384
1385If you've messed up the working tree, but haven't yet committed your
1386mistake, you can return the entire working tree to the last committed
1387state with
1388
1389-------------------------------------------------
1390$ git reset --hard HEAD
1391-------------------------------------------------
1392
1393If you make a commit that you later wish you hadn't, there are two
1394fundamentally different ways to fix the problem:
1395
1396        1. You can create a new commit that undoes whatever was done
1397        by the old commit.  This is the correct thing if your
1398        mistake has already been made public.
1399
1400        2. You can go back and modify the old commit.  You should
1401        never do this if you have already made the history public;
1402        git does not normally expect the "history" of a project to
1403        change, and cannot correctly perform repeated merges from
1404        a branch that has had its history changed.
1405
1406[[reverting-a-commit]]
1407Fixing a mistake with a new commit
1408~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1409
1410Creating a new commit that reverts an earlier change is very easy;
1411just pass the linkgit:git-revert[1] command a reference to the bad
1412commit; for example, to revert the most recent commit:
1413
1414-------------------------------------------------
1415$ git revert HEAD
1416-------------------------------------------------
1417
1418This will create a new commit which undoes the change in HEAD.  You
1419will be given a chance to edit the commit message for the new commit.
1420
1421You can also revert an earlier change, for example, the next-to-last:
1422
1423-------------------------------------------------
1424$ git revert HEAD^
1425-------------------------------------------------
1426
1427In this case git will attempt to undo the old change while leaving
1428intact any changes made since then.  If more recent changes overlap
1429with the changes to be reverted, then you will be asked to fix
1430conflicts manually, just as in the case of <<resolving-a-merge,
1431resolving a merge>>.
1432
1433[[fixing-a-mistake-by-rewriting-history]]
1434Fixing a mistake by rewriting history
1435~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1436
1437If the problematic commit is the most recent commit, and you have not
1438yet made that commit public, then you may just
1439<<undoing-a-merge,destroy it using git-reset>>.
1440
1441Alternatively, you
1442can edit the working directory and update the index to fix your
1443mistake, just as if you were going to <<how-to-make-a-commit,create a
1444new commit>>, then run
1445
1446-------------------------------------------------
1447$ git commit --amend
1448-------------------------------------------------
1449
1450which will replace the old commit by a new commit incorporating your
1451changes, giving you a chance to edit the old commit message first.
1452
1453Again, you should never do this to a commit that may already have
1454been merged into another branch; use linkgit:git-revert[1] instead in
1455that case.
1456
1457It is also possible to replace commits further back in the history, but
1458this is an advanced topic to be left for
1459<<cleaning-up-history,another chapter>>.
1460
1461[[checkout-of-path]]
1462Checking out an old version of a file
1463~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1464
1465In the process of undoing a previous bad change, you may find it
1466useful to check out an older version of a particular file using
1467linkgit:git-checkout[1].  We've used git-checkout before to switch
1468branches, but it has quite different behavior if it is given a path
1469name: the command
1470
1471-------------------------------------------------
1472$ git checkout HEAD^ path/to/file
1473-------------------------------------------------
1474
1475replaces path/to/file by the contents it had in the commit HEAD^, and
1476also updates the index to match.  It does not change branches.
1477
1478If you just want to look at an old version of the file, without
1479modifying the working directory, you can do that with
1480linkgit:git-show[1]:
1481
1482-------------------------------------------------
1483$ git show HEAD^:path/to/file
1484-------------------------------------------------
1485
1486which will display the given version of the file.
1487
1488[[interrupted-work]]
1489Temporarily setting aside work in progress
1490~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1491
1492While you are in the middle of working on something complicated, you
1493find an unrelated but obvious and trivial bug.  You would like to fix it
1494before continuing.  You can use linkgit:git-stash[1] to save the current
1495state of your work, and after fixing the bug (or, optionally after doing
1496so on a different branch and then coming back), unstash the
1497work-in-progress changes.
1498
1499------------------------------------------------
1500$ git stash save "work in progress for foo feature"
1501------------------------------------------------
1502
1503This command will save your changes away to the `stash`, and
1504reset your working tree and the index to match the tip of your
1505current branch.  Then you can make your fix as usual.
1506
1507------------------------------------------------
1508... edit and test ...
1509$ git commit -a -m "blorpl: typofix"
1510------------------------------------------------
1511
1512After that, you can go back to what you were working on with
1513`git stash apply`:
1514
1515------------------------------------------------
1516$ git stash apply
1517------------------------------------------------
1518
1519
1520[[ensuring-good-performance]]
1521Ensuring good performance
1522-------------------------
1523
1524On large repositories, git depends on compression to keep the history
1525information from taking up too much space on disk or in memory.
1526
1527This compression is not performed automatically.  Therefore you
1528should occasionally run linkgit:git-gc[1]:
1529
1530-------------------------------------------------
1531$ git gc
1532-------------------------------------------------
1533
1534to recompress the archive.  This can be very time-consuming, so
1535you may prefer to run git-gc when you are not doing other work.
1536
1537
1538[[ensuring-reliability]]
1539Ensuring reliability
1540--------------------
1541
1542[[checking-for-corruption]]
1543Checking the repository for corruption
1544~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1545
1546The linkgit:git-fsck[1] command runs a number of self-consistency checks
1547on the repository, and reports on any problems.  This may take some
1548time.  The most common warning by far is about "dangling" objects:
1549
1550-------------------------------------------------
1551$ git fsck
1552dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1553dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1554dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1555dangling blob 218761f9d90712d37a9c5e36f406f92202db07eb
1556dangling commit bf093535a34a4d35731aa2bd90fe6b176302f14f
1557dangling commit 8e4bec7f2ddaa268bef999853c25755452100f8e
1558dangling tree d50bb86186bf27b681d25af89d3b5b68382e4085
1559dangling tree b24c2473f1fd3d91352a624795be026d64c8841f
1560...
1561-------------------------------------------------
1562
1563Dangling objects are not a problem.  At worst they may take up a little
1564extra disk space.  They can sometimes provide a last-resort method for
1565recovering lost work--see <<dangling-objects>> for details.
1566
1567[[recovering-lost-changes]]
1568Recovering lost changes
1569~~~~~~~~~~~~~~~~~~~~~~~
1570
1571[[reflogs]]
1572Reflogs
1573^^^^^^^
1574
1575Say you modify a branch with `linkgit:git-reset[1] --hard`, and then
1576realize that the branch was the only reference you had to that point in
1577history.
1578
1579Fortunately, git also keeps a log, called a "reflog", of all the
1580previous values of each branch.  So in this case you can still find the
1581old history using, for example,
1582
1583-------------------------------------------------
1584$ git log master@{1}
1585-------------------------------------------------
1586
1587This lists the commits reachable from the previous version of the
1588"master" branch head.  This syntax can be used with any git command
1589that accepts a commit, not just with git log.  Some other examples:
1590
1591-------------------------------------------------
1592$ git show master@{2}           # See where the branch pointed 2,
1593$ git show master@{3}           # 3, ... changes ago.
1594$ gitk master@{yesterday}       # See where it pointed yesterday,
1595$ gitk master@{"1 week ago"}    # ... or last week
1596$ git log --walk-reflogs master # show reflog entries for master
1597-------------------------------------------------
1598
1599A separate reflog is kept for the HEAD, so
1600
1601-------------------------------------------------
1602$ git show HEAD@{"1 week ago"}
1603-------------------------------------------------
1604
1605will show what HEAD pointed to one week ago, not what the current branch
1606pointed to one week ago.  This allows you to see the history of what
1607you've checked out.
1608
1609The reflogs are kept by default for 30 days, after which they may be
1610pruned.  See linkgit:git-reflog[1] and linkgit:git-gc[1] to learn
1611how to control this pruning, and see the "SPECIFYING REVISIONS"
1612section of linkgit:git-rev-parse[1] for details.
1613
1614Note that the reflog history is very different from normal git history.
1615While normal history is shared by every repository that works on the
1616same project, the reflog history is not shared: it tells you only about
1617how the branches in your local repository have changed over time.
1618
1619[[dangling-object-recovery]]
1620Examining dangling objects
1621^^^^^^^^^^^^^^^^^^^^^^^^^^
1622
1623In some situations the reflog may not be able to save you.  For example,
1624suppose you delete a branch, then realize you need the history it
1625contained.  The reflog is also deleted; however, if you have not yet
1626pruned the repository, then you may still be able to find the lost
1627commits in the dangling objects that git-fsck reports.  See
1628<<dangling-objects>> for the details.
1629
1630-------------------------------------------------
1631$ git fsck
1632dangling commit 7281251ddd2a61e38657c827739c57015671a6b3
1633dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a63
1634dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b5
1635...
1636-------------------------------------------------
1637
1638You can examine
1639one of those dangling commits with, for example,
1640
1641------------------------------------------------
1642$ gitk 7281251ddd --not --all
1643------------------------------------------------
1644
1645which does what it sounds like: it says that you want to see the commit
1646history that is described by the dangling commit(s), but not the
1647history that is described by all your existing branches and tags.  Thus
1648you get exactly the history reachable from that commit that is lost.
1649(And notice that it might not be just one commit: we only report the
1650"tip of the line" as being dangling, but there might be a whole deep
1651and complex commit history that was dropped.)
1652
1653If you decide you want the history back, you can always create a new
1654reference pointing to it, for example, a new branch:
1655
1656------------------------------------------------
1657$ git branch recovered-branch 7281251ddd
1658------------------------------------------------
1659
1660Other types of dangling objects (blobs and trees) are also possible, and
1661dangling objects can arise in other situations.
1662
1663
1664[[sharing-development]]
1665Sharing development with others
1666===============================
1667
1668[[getting-updates-With-git-pull]]
1669Getting updates with git-pull
1670-----------------------------
1671
1672After you clone a repository and make a few changes of your own, you
1673may wish to check the original repository for updates and merge them
1674into your own work.
1675
1676We have already seen <<Updating-a-repository-With-git-fetch,how to
1677keep remote tracking branches up to date>> with linkgit:git-fetch[1],
1678and how to merge two branches.  So you can merge in changes from the
1679original repository's master branch with:
1680
1681-------------------------------------------------
1682$ git fetch
1683$ git merge origin/master
1684-------------------------------------------------
1685
1686However, the linkgit:git-pull[1] command provides a way to do this in
1687one step:
1688
1689-------------------------------------------------
1690$ git pull origin master
1691-------------------------------------------------
1692
1693In fact, if you have "master" checked out, then by default "git pull"
1694merges from the HEAD branch of the origin repository.  So often you can
1695accomplish the above with just a simple
1696
1697-------------------------------------------------
1698$ git pull
1699-------------------------------------------------
1700
1701More generally, a branch that is created from a remote branch will pull
1702by default from that branch.  See the descriptions of the
1703branch.<name>.remote and branch.<name>.merge options in
1704linkgit:git-config[1], and the discussion of the `--track` option in
1705linkgit:git-checkout[1], to learn how to control these defaults.
1706
1707In addition to saving you keystrokes, "git pull" also helps you by
1708producing a default commit message documenting the branch and
1709repository that you pulled from.
1710
1711(But note that no such commit will be created in the case of a
1712<<fast-forwards,fast forward>>; instead, your branch will just be
1713updated to point to the latest commit from the upstream branch.)
1714
1715The git-pull command can also be given "." as the "remote" repository,
1716in which case it just merges in a branch from the current repository; so
1717the commands
1718
1719-------------------------------------------------
1720$ git pull . branch
1721$ git merge branch
1722-------------------------------------------------
1723
1724are roughly equivalent.  The former is actually very commonly used.
1725
1726[[submitting-patches]]
1727Submitting patches to a project
1728-------------------------------
1729
1730If you just have a few changes, the simplest way to submit them may
1731just be to send them as patches in email:
1732
1733First, use linkgit:git-format-patch[1]; for example:
1734
1735-------------------------------------------------
1736$ git format-patch origin
1737-------------------------------------------------
1738
1739will produce a numbered series of files in the current directory, one
1740for each patch in the current branch but not in origin/HEAD.
1741
1742You can then import these into your mail client and send them by
1743hand.  However, if you have a lot to send at once, you may prefer to
1744use the linkgit:git-send-email[1] script to automate the process.
1745Consult the mailing list for your project first to determine how they
1746prefer such patches be handled.
1747
1748[[importing-patches]]
1749Importing patches to a project
1750------------------------------
1751
1752Git also provides a tool called linkgit:git-am[1] (am stands for
1753"apply mailbox"), for importing such an emailed series of patches.
1754Just save all of the patch-containing messages, in order, into a
1755single mailbox file, say "patches.mbox", then run
1756
1757-------------------------------------------------
1758$ git am -3 patches.mbox
1759-------------------------------------------------
1760
1761Git will apply each patch in order; if any conflicts are found, it
1762will stop, and you can fix the conflicts as described in
1763"<<resolving-a-merge,Resolving a merge>>".  (The "-3" option tells
1764git to perform a merge; if you would prefer it just to abort and
1765leave your tree and index untouched, you may omit that option.)
1766
1767Once the index is updated with the results of the conflict
1768resolution, instead of creating a new commit, just run
1769
1770-------------------------------------------------
1771$ git am --resolved
1772-------------------------------------------------
1773
1774and git will create the commit for you and continue applying the
1775remaining patches from the mailbox.
1776
1777The final result will be a series of commits, one for each patch in
1778the original mailbox, with authorship and commit log message each
1779taken from the message containing each patch.
1780
1781[[public-repositories]]
1782Public git repositories
1783-----------------------
1784
1785Another way to submit changes to a project is to tell the maintainer
1786of that project to pull the changes from your repository using
1787linkgit:git-pull[1].  In the section "<<getting-updates-With-git-pull,
1788Getting updates with git-pull>>" we described this as a way to get
1789updates from the "main" repository, but it works just as well in the
1790other direction.
1791
1792If you and the maintainer both have accounts on the same machine, then
1793you can just pull changes from each other's repositories directly;
1794commands that accept repository URLs as arguments will also accept a
1795local directory name:
1796
1797-------------------------------------------------
1798$ git clone /path/to/repository
1799$ git pull /path/to/other/repository
1800-------------------------------------------------
1801
1802or an ssh URL:
1803
1804-------------------------------------------------
1805$ git clone ssh://yourhost/~you/repository
1806-------------------------------------------------
1807
1808For projects with few developers, or for synchronizing a few private
1809repositories, this may be all you need.
1810
1811However, the more common way to do this is to maintain a separate public
1812repository (usually on a different host) for others to pull changes
1813from.  This is usually more convenient, and allows you to cleanly
1814separate private work in progress from publicly visible work.
1815
1816You will continue to do your day-to-day work in your personal
1817repository, but periodically "push" changes from your personal
1818repository into your public repository, allowing other developers to
1819pull from that repository.  So the flow of changes, in a situation
1820where there is one other developer with a public repository, looks
1821like this:
1822
1823                        you push
1824  your personal repo ------------------> your public repo
1825        ^                                     |
1826        |                                     |
1827        | you pull                            | they pull
1828        |                                     |
1829        |                                     |
1830        |               they push             V
1831  their public repo <------------------- their repo
1832
1833We explain how to do this in the following sections.
1834
1835[[setting-up-a-public-repository]]
1836Setting up a public repository
1837~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1838
1839Assume your personal repository is in the directory ~/proj.  We
1840first create a new clone of the repository and tell git-daemon that it
1841is meant to be public:
1842
1843-------------------------------------------------
1844$ git clone --bare ~/proj proj.git
1845$ touch proj.git/git-daemon-export-ok
1846-------------------------------------------------
1847
1848The resulting directory proj.git contains a "bare" git repository--it is
1849just the contents of the ".git" directory, without any files checked out
1850around it.
1851
1852Next, copy proj.git to the server where you plan to host the
1853public repository.  You can use scp, rsync, or whatever is most
1854convenient.
1855
1856[[exporting-via-git]]
1857Exporting a git repository via the git protocol
1858~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1859
1860This is the preferred method.
1861
1862If someone else administers the server, they should tell you what
1863directory to put the repository in, and what git:// URL it will appear
1864at.  You can then skip to the section
1865"<<pushing-changes-to-a-public-repository,Pushing changes to a public
1866repository>>", below.
1867
1868Otherwise, all you need to do is start linkgit:git-daemon[1]; it will
1869listen on port 9418.  By default, it will allow access to any directory
1870that looks like a git directory and contains the magic file
1871git-daemon-export-ok.  Passing some directory paths as git-daemon
1872arguments will further restrict the exports to those paths.
1873
1874You can also run git-daemon as an inetd service; see the
1875linkgit:git-daemon[1] man page for details.  (See especially the
1876examples section.)
1877
1878[[exporting-via-http]]
1879Exporting a git repository via http
1880~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1881
1882The git protocol gives better performance and reliability, but on a
1883host with a web server set up, http exports may be simpler to set up.
1884
1885All you need to do is place the newly created bare git repository in
1886a directory that is exported by the web server, and make some
1887adjustments to give web clients some extra information they need:
1888
1889-------------------------------------------------
1890$ mv proj.git /home/you/public_html/proj.git
1891$ cd proj.git
1892$ git --bare update-server-info
1893$ mv hooks/post-update.sample hooks/post-update
1894-------------------------------------------------
1895
1896(For an explanation of the last two lines, see
1897linkgit:git-update-server-info[1] and linkgit:githooks[5].)
1898
1899Advertise the URL of proj.git.  Anybody else should then be able to
1900clone or pull from that URL, for example with a command line like:
1901
1902-------------------------------------------------
1903$ git clone http://yourserver.com/~you/proj.git
1904-------------------------------------------------
1905
1906(See also
1907link:howto/setup-git-server-over-http.txt[setup-git-server-over-http]
1908for a slightly more sophisticated setup using WebDAV which also
1909allows pushing over http.)
1910
1911[[pushing-changes-to-a-public-repository]]
1912Pushing changes to a public repository
1913~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1914
1915Note that the two techniques outlined above (exporting via
1916<<exporting-via-http,http>> or <<exporting-via-git,git>>) allow other
1917maintainers to fetch your latest changes, but they do not allow write
1918access, which you will need to update the public repository with the
1919latest changes created in your private repository.
1920
1921The simplest way to do this is using linkgit:git-push[1] and ssh; to
1922update the remote branch named "master" with the latest state of your
1923branch named "master", run
1924
1925-------------------------------------------------
1926$ git push ssh://yourserver.com/~you/proj.git master:master
1927-------------------------------------------------
1928
1929or just
1930
1931-------------------------------------------------
1932$ git push ssh://yourserver.com/~you/proj.git master
1933-------------------------------------------------
1934
1935As with git-fetch, git-push will complain if this does not result in a
1936<<fast-forwards,fast forward>>; see the following section for details on
1937handling this case.
1938
1939Note that the target of a "push" is normally a
1940<<def_bare_repository,bare>> repository.  You can also push to a
1941repository that has a checked-out working tree, but the working tree
1942will not be updated by the push.  This may lead to unexpected results if
1943the branch you push to is the currently checked-out branch!
1944
1945As with git-fetch, you may also set up configuration options to
1946save typing; so, for example, after
1947
1948-------------------------------------------------
1949$ cat >>.git/config <<EOF
1950[remote "public-repo"]
1951        url = ssh://yourserver.com/~you/proj.git
1952EOF
1953-------------------------------------------------
1954
1955you should be able to perform the above push with just
1956
1957-------------------------------------------------
1958$ git push public-repo master
1959-------------------------------------------------
1960
1961See the explanations of the remote.<name>.url, branch.<name>.remote,
1962and remote.<name>.push options in linkgit:git-config[1] for
1963details.
1964
1965[[forcing-push]]
1966What to do when a push fails
1967~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1968
1969If a push would not result in a <<fast-forwards,fast forward>> of the
1970remote branch, then it will fail with an error like:
1971
1972-------------------------------------------------
1973error: remote 'refs/heads/master' is not an ancestor of
1974 local  'refs/heads/master'.
1975 Maybe you are not up-to-date and need to pull first?
1976error: failed to push to 'ssh://yourserver.com/~you/proj.git'
1977-------------------------------------------------
1978
1979This can happen, for example, if you:
1980
1981        - use `git-reset --hard` to remove already-published commits, or
1982        - use `git-commit --amend` to replace already-published commits
1983          (as in <<fixing-a-mistake-by-rewriting-history>>), or
1984        - use `git-rebase` to rebase any already-published commits (as
1985          in <<using-git-rebase>>).
1986
1987You may force git-push to perform the update anyway by preceding the
1988branch name with a plus sign:
1989
1990-------------------------------------------------
1991$ git push ssh://yourserver.com/~you/proj.git +master
1992-------------------------------------------------
1993
1994Normally whenever a branch head in a public repository is modified, it
1995is modified to point to a descendant of the commit that it pointed to
1996before.  By forcing a push in this situation, you break that convention.
1997(See <<problems-With-rewriting-history>>.)
1998
1999Nevertheless, this is a common practice for people that need a simple
2000way to publish a work-in-progress patch series, and it is an acceptable
2001compromise as long as you warn other developers that this is how you
2002intend to manage the branch.
2003
2004It's also possible for a push to fail in this way when other people have
2005the right to push to the same repository.  In that case, the correct
2006solution is to retry the push after first updating your work: either by a
2007pull, or by a fetch followed by a rebase; see the
2008<<setting-up-a-shared-repository,next section>> and
2009linkgit:gitcvs-migration[7] for more.
2010
2011[[setting-up-a-shared-repository]]
2012Setting up a shared repository
2013~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2014
2015Another way to collaborate is by using a model similar to that
2016commonly used in CVS, where several developers with special rights
2017all push to and pull from a single shared repository.  See
2018linkgit:gitcvs-migration[7] for instructions on how to
2019set this up.
2020
2021However, while there is nothing wrong with git's support for shared
2022repositories, this mode of operation is not generally recommended,
2023simply because the mode of collaboration that git supports--by
2024exchanging patches and pulling from public repositories--has so many
2025advantages over the central shared repository:
2026
2027        - Git's ability to quickly import and merge patches allows a
2028          single maintainer to process incoming changes even at very
2029          high rates.  And when that becomes too much, git-pull provides
2030          an easy way for that maintainer to delegate this job to other
2031          maintainers while still allowing optional review of incoming
2032          changes.
2033        - Since every developer's repository has the same complete copy
2034          of the project history, no repository is special, and it is
2035          trivial for another developer to take over maintenance of a
2036          project, either by mutual agreement, or because a maintainer
2037          becomes unresponsive or difficult to work with.
2038        - The lack of a central group of "committers" means there is
2039          less need for formal decisions about who is "in" and who is
2040          "out".
2041
2042[[setting-up-gitweb]]
2043Allowing web browsing of a repository
2044~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2045
2046The gitweb cgi script provides users an easy way to browse your
2047project's files and history without having to install git; see the file
2048gitweb/INSTALL in the git source tree for instructions on setting it up.
2049
2050[[sharing-development-examples]]
2051Examples
2052--------
2053
2054[[maintaining-topic-branches]]
2055Maintaining topic branches for a Linux subsystem maintainer
2056~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2057
2058This describes how Tony Luck uses git in his role as maintainer of the
2059IA64 architecture for the Linux kernel.
2060
2061He uses two public branches:
2062
2063 - A "test" tree into which patches are initially placed so that they
2064   can get some exposure when integrated with other ongoing development.
2065   This tree is available to Andrew for pulling into -mm whenever he
2066   wants.
2067
2068 - A "release" tree into which tested patches are moved for final sanity
2069   checking, and as a vehicle to send them upstream to Linus (by sending
2070   him a "please pull" request.)
2071
2072He also uses a set of temporary branches ("topic branches"), each
2073containing a logical grouping of patches.
2074
2075To set this up, first create your work tree by cloning Linus's public
2076tree:
2077
2078-------------------------------------------------
2079$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git work
2080$ cd work
2081-------------------------------------------------
2082
2083Linus's tree will be stored in the remote branch named origin/master,
2084and can be updated using linkgit:git-fetch[1]; you can track other
2085public trees using linkgit:git-remote[1] to set up a "remote" and
2086linkgit:git-fetch[1] to keep them up-to-date; see
2087<<repositories-and-branches>>.
2088
2089Now create the branches in which you are going to work; these start out
2090at the current tip of origin/master branch, and should be set up (using
2091the --track option to linkgit:git-branch[1]) to merge changes in from
2092Linus by default.
2093
2094-------------------------------------------------
2095$ git branch --track test origin/master
2096$ git branch --track release origin/master
2097-------------------------------------------------
2098
2099These can be easily kept up to date using linkgit:git-pull[1].
2100
2101-------------------------------------------------
2102$ git checkout test && git pull
2103$ git checkout release && git pull
2104-------------------------------------------------
2105
2106Important note!  If you have any local changes in these branches, then
2107this merge will create a commit object in the history (with no local
2108changes git will simply do a "Fast forward" merge).  Many people dislike
2109the "noise" that this creates in the Linux history, so you should avoid
2110doing this capriciously in the "release" branch, as these noisy commits
2111will become part of the permanent history when you ask Linus to pull
2112from the release branch.
2113
2114A few configuration variables (see linkgit:git-config[1]) can
2115make it easy to push both branches to your public tree.  (See
2116<<setting-up-a-public-repository>>.)
2117
2118-------------------------------------------------
2119$ cat >> .git/config <<EOF
2120[remote "mytree"]
2121        url =  master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6.git
2122        push = release
2123        push = test
2124EOF
2125-------------------------------------------------
2126
2127Then you can push both the test and release trees using
2128linkgit:git-push[1]:
2129
2130-------------------------------------------------
2131$ git push mytree
2132-------------------------------------------------
2133
2134or push just one of the test and release branches using:
2135
2136-------------------------------------------------
2137$ git push mytree test
2138-------------------------------------------------
2139
2140or
2141
2142-------------------------------------------------
2143$ git push mytree release
2144-------------------------------------------------
2145
2146Now to apply some patches from the community.  Think of a short
2147snappy name for a branch to hold this patch (or related group of
2148patches), and create a new branch from the current tip of Linus's
2149branch:
2150
2151-------------------------------------------------
2152$ git checkout -b speed-up-spinlocks origin
2153-------------------------------------------------
2154
2155Now you apply the patch(es), run some tests, and commit the change(s).  If
2156the patch is a multi-part series, then you should apply each as a separate
2157commit to this branch.
2158
2159-------------------------------------------------
2160$ ... patch ... test  ... commit [ ... patch ... test ... commit ]*
2161-------------------------------------------------
2162
2163When you are happy with the state of this change, you can pull it into the
2164"test" branch in preparation to make it public:
2165
2166-------------------------------------------------
2167$ git checkout test && git pull . speed-up-spinlocks
2168-------------------------------------------------
2169
2170It is unlikely that you would have any conflicts here ... but you might if you
2171spent a while on this step and had also pulled new versions from upstream.
2172
2173Some time later when enough time has passed and testing done, you can pull the
2174same branch into the "release" tree ready to go upstream.  This is where you
2175see the value of keeping each patch (or patch series) in its own branch.  It
2176means that the patches can be moved into the "release" tree in any order.
2177
2178-------------------------------------------------
2179$ git checkout release && git pull . speed-up-spinlocks
2180-------------------------------------------------
2181
2182After a while, you will have a number of branches, and despite the
2183well chosen names you picked for each of them, you may forget what
2184they are for, or what status they are in.  To get a reminder of what
2185changes are in a specific branch, use:
2186
2187-------------------------------------------------
2188$ git log linux..branchname | git shortlog
2189-------------------------------------------------
2190
2191To see whether it has already been merged into the test or release branches,
2192use:
2193
2194-------------------------------------------------
2195$ git log test..branchname
2196-------------------------------------------------
2197
2198or
2199
2200-------------------------------------------------
2201$ git log release..branchname
2202-------------------------------------------------
2203
2204(If this branch has not yet been merged, you will see some log entries.
2205If it has been merged, then there will be no output.)
2206
2207Once a patch completes the great cycle (moving from test to release,
2208then pulled by Linus, and finally coming back into your local
2209"origin/master" branch), the branch for this change is no longer needed.
2210You detect this when the output from:
2211
2212-------------------------------------------------
2213$ git log origin..branchname
2214-------------------------------------------------
2215
2216is empty.  At this point the branch can be deleted:
2217
2218-------------------------------------------------
2219$ git branch -d branchname
2220-------------------------------------------------
2221
2222Some changes are so trivial that it is not necessary to create a separate
2223branch and then merge into each of the test and release branches.  For
2224these changes, just apply directly to the "release" branch, and then
2225merge that into the "test" branch.
2226
2227To create diffstat and shortlog summaries of changes to include in a "please
2228pull" request to Linus you can use:
2229
2230-------------------------------------------------
2231$ git diff --stat origin..release
2232-------------------------------------------------
2233
2234and
2235
2236-------------------------------------------------
2237$ git log -p origin..release | git shortlog
2238-------------------------------------------------
2239
2240Here are some of the scripts that simplify all this even further.
2241
2242-------------------------------------------------
2243==== update script ====
2244# Update a branch in my GIT tree.  If the branch to be updated
2245# is origin, then pull from kernel.org.  Otherwise merge
2246# origin/master branch into test|release branch
2247
2248case "$1" in
2249test|release)
2250        git checkout $1 && git pull . origin
2251        ;;
2252origin)
2253        before=$(git rev-parse refs/remotes/origin/master)
2254        git fetch origin
2255        after=$(git rev-parse refs/remotes/origin/master)
2256        if [ $before != $after ]
2257        then
2258                git log $before..$after | git shortlog
2259        fi
2260        ;;
2261*)
2262        echo "Usage: $0 origin|test|release" 1>&2
2263        exit 1
2264        ;;
2265esac
2266-------------------------------------------------
2267
2268-------------------------------------------------
2269==== merge script ====
2270# Merge a branch into either the test or release branch
2271
2272pname=$0
2273
2274usage()
2275{
2276        echo "Usage: $pname branch test|release" 1>&2
2277        exit 1
2278}
2279
2280git show-ref -q --verify -- refs/heads/"$1" || {
2281        echo "Can't see branch <$1>" 1>&2
2282        usage
2283}
2284
2285case "$2" in
2286test|release)
2287        if [ $(git log $2..$1 | wc -c) -eq 0 ]
2288        then
2289                echo $1 already merged into $2 1>&2
2290                exit 1
2291        fi
2292        git checkout $2 && git pull . $1
2293        ;;
2294*)
2295        usage
2296        ;;
2297esac
2298-------------------------------------------------
2299
2300-------------------------------------------------
2301==== status script ====
2302# report on status of my ia64 GIT tree
2303
2304gb=$(tput setab 2)
2305rb=$(tput setab 1)
2306restore=$(tput setab 9)
2307
2308if [ `git rev-list test..release | wc -c` -gt 0 ]
2309then
2310        echo $rb Warning: commits in release that are not in test $restore
2311        git log test..release
2312fi
2313
2314for branch in `git show-ref --heads | sed 's|^.*/||'`
2315do
2316        if [ $branch = test -o $branch = release ]
2317        then
2318                continue
2319        fi
2320
2321        echo -n $gb ======= $branch ====== $restore " "
2322        status=
2323        for ref in test release origin/master
2324        do
2325                if [ `git rev-list $ref..$branch | wc -c` -gt 0 ]
2326                then
2327                        status=$status${ref:0:1}
2328                fi
2329        done
2330        case $status in
2331        trl)
2332                echo $rb Need to pull into test $restore
2333                ;;
2334        rl)
2335                echo "In test"
2336                ;;
2337        l)
2338                echo "Waiting for linus"
2339                ;;
2340        "")
2341                echo $rb All done $restore
2342                ;;
2343        *)
2344                echo $rb "<$status>" $restore
2345                ;;
2346        esac
2347        git log origin/master..$branch | git shortlog
2348done
2349-------------------------------------------------
2350
2351
2352[[cleaning-up-history]]
2353Rewriting history and maintaining patch series
2354==============================================
2355
2356Normally commits are only added to a project, never taken away or
2357replaced.  Git is designed with this assumption, and violating it will
2358cause git's merge machinery (for example) to do the wrong thing.
2359
2360However, there is a situation in which it can be useful to violate this
2361assumption.
2362
2363[[patch-series]]
2364Creating the perfect patch series
2365---------------------------------
2366
2367Suppose you are a contributor to a large project, and you want to add a
2368complicated feature, and to present it to the other developers in a way
2369that makes it easy for them to read your changes, verify that they are
2370correct, and understand why you made each change.
2371
2372If you present all of your changes as a single patch (or commit), they
2373may find that it is too much to digest all at once.
2374
2375If you present them with the entire history of your work, complete with
2376mistakes, corrections, and dead ends, they may be overwhelmed.
2377
2378So the ideal is usually to produce a series of patches such that:
2379
2380        1. Each patch can be applied in order.
2381
2382        2. Each patch includes a single logical change, together with a
2383           message explaining the change.
2384
2385        3. No patch introduces a regression: after applying any initial
2386           part of the series, the resulting project still compiles and
2387           works, and has no bugs that it didn't have before.
2388
2389        4. The complete series produces the same end result as your own
2390           (probably much messier!) development process did.
2391
2392We will introduce some tools that can help you do this, explain how to
2393use them, and then explain some of the problems that can arise because
2394you are rewriting history.
2395
2396[[using-git-rebase]]
2397Keeping a patch series up to date using git-rebase
2398--------------------------------------------------
2399
2400Suppose that you create a branch "mywork" on a remote-tracking branch
2401"origin", and create some commits on top of it:
2402
2403-------------------------------------------------
2404$ git checkout -b mywork origin
2405$ vi file.txt
2406$ git commit
2407$ vi otherfile.txt
2408$ git commit
2409...
2410-------------------------------------------------
2411
2412You have performed no merges into mywork, so it is just a simple linear
2413sequence of patches on top of "origin":
2414
2415................................................
2416 o--o--o <-- origin
2417        \
2418         o--o--o <-- mywork
2419................................................
2420
2421Some more interesting work has been done in the upstream project, and
2422"origin" has advanced:
2423
2424................................................
2425 o--o--O--o--o--o <-- origin
2426        \
2427         a--b--c <-- mywork
2428................................................
2429
2430At this point, you could use "pull" to merge your changes back in;
2431the result would create a new merge commit, like this:
2432
2433................................................
2434 o--o--O--o--o--o <-- origin
2435        \        \
2436         a--b--c--m <-- mywork
2437................................................
2438
2439However, if you prefer to keep the history in mywork a simple series of
2440commits without any merges, you may instead choose to use
2441linkgit:git-rebase[1]:
2442
2443-------------------------------------------------
2444$ git checkout mywork
2445$ git rebase origin
2446-------------------------------------------------
2447
2448This will remove each of your commits from mywork, temporarily saving
2449them as patches (in a directory named ".git/rebase-apply"), update mywork to
2450point at the latest version of origin, then apply each of the saved
2451patches to the new mywork.  The result will look like:
2452
2453
2454................................................
2455 o--o--O--o--o--o <-- origin
2456                 \
2457                  a'--b'--c' <-- mywork
2458................................................
2459
2460In the process, it may discover conflicts.  In that case it will stop
2461and allow you to fix the conflicts; after fixing conflicts, use "git-add"
2462to update the index with those contents, and then, instead of
2463running git-commit, just run
2464
2465-------------------------------------------------
2466$ git rebase --continue
2467-------------------------------------------------
2468
2469and git will continue applying the rest of the patches.
2470
2471At any point you may use the `--abort` option to abort this process and
2472return mywork to the state it had before you started the rebase:
2473
2474-------------------------------------------------
2475$ git rebase --abort
2476-------------------------------------------------
2477
2478[[rewriting-one-commit]]
2479Rewriting a single commit
2480-------------------------
2481
2482We saw in <<fixing-a-mistake-by-rewriting-history>> that you can replace the
2483most recent commit using
2484
2485-------------------------------------------------
2486$ git commit --amend
2487-------------------------------------------------
2488
2489which will replace the old commit by a new commit incorporating your
2490changes, giving you a chance to edit the old commit message first.
2491
2492You can also use a combination of this and linkgit:git-rebase[1] to
2493replace a commit further back in your history and recreate the
2494intervening changes on top of it.  First, tag the problematic commit
2495with
2496
2497-------------------------------------------------
2498$ git tag bad mywork~5
2499-------------------------------------------------
2500
2501(Either gitk or git-log may be useful for finding the commit.)
2502
2503Then check out that commit, edit it, and rebase the rest of the series
2504on top of it (note that we could check out the commit on a temporary
2505branch, but instead we're using a <<detached-head,detached head>>):
2506
2507-------------------------------------------------
2508$ git checkout bad
2509$ # make changes here and update the index
2510$ git commit --amend
2511$ git rebase --onto HEAD bad mywork
2512-------------------------------------------------
2513
2514When you're done, you'll be left with mywork checked out, with the top
2515patches on mywork reapplied on top of your modified commit.  You can
2516then clean up with
2517
2518-------------------------------------------------
2519$ git tag -d bad
2520-------------------------------------------------
2521
2522Note that the immutable nature of git history means that you haven't really
2523"modified" existing commits; instead, you have replaced the old commits with
2524new commits having new object names.
2525
2526[[reordering-patch-series]]
2527Reordering or selecting from a patch series
2528-------------------------------------------
2529
2530Given one existing commit, the linkgit:git-cherry-pick[1] command
2531allows you to apply the change introduced by that commit and create a
2532new commit that records it.  So, for example, if "mywork" points to a
2533series of patches on top of "origin", you might do something like:
2534
2535-------------------------------------------------
2536$ git checkout -b mywork-new origin
2537$ gitk origin..mywork &
2538-------------------------------------------------
2539
2540and browse through the list of patches in the mywork branch using gitk,
2541applying them (possibly in a different order) to mywork-new using
2542cherry-pick, and possibly modifying them as you go using `commit --amend`.
2543The linkgit:git-gui[1] command may also help as it allows you to
2544individually select diff hunks for inclusion in the index (by
2545right-clicking on the diff hunk and choosing "Stage Hunk for Commit").
2546
2547Another technique is to use git-format-patch to create a series of
2548patches, then reset the state to before the patches:
2549
2550-------------------------------------------------
2551$ git format-patch origin
2552$ git reset --hard origin
2553-------------------------------------------------
2554
2555Then modify, reorder, or eliminate patches as preferred before applying
2556them again with linkgit:git-am[1].
2557
2558[[patch-series-tools]]
2559Other tools
2560-----------
2561
2562There are numerous other tools, such as StGIT, which exist for the
2563purpose of maintaining a patch series.  These are outside of the scope of
2564this manual.
2565
2566[[problems-With-rewriting-history]]
2567Problems with rewriting history
2568-------------------------------
2569
2570The primary problem with rewriting the history of a branch has to do
2571with merging.  Suppose somebody fetches your branch and merges it into
2572their branch, with a result something like this:
2573
2574................................................
2575 o--o--O--o--o--o <-- origin
2576        \        \
2577         t--t--t--m <-- their branch:
2578................................................
2579
2580Then suppose you modify the last three commits:
2581
2582................................................
2583         o--o--o <-- new head of origin
2584        /
2585 o--o--O--o--o--o <-- old head of origin
2586................................................
2587
2588If we examined all this history together in one repository, it will
2589look like:
2590
2591................................................
2592         o--o--o <-- new head of origin
2593        /
2594 o--o--O--o--o--o <-- old head of origin
2595        \        \
2596         t--t--t--m <-- their branch:
2597................................................
2598
2599Git has no way of knowing that the new head is an updated version of
2600the old head; it treats this situation exactly the same as it would if
2601two developers had independently done the work on the old and new heads
2602in parallel.  At this point, if someone attempts to merge the new head
2603in to their branch, git will attempt to merge together the two (old and
2604new) lines of development, instead of trying to replace the old by the
2605new.  The results are likely to be unexpected.
2606
2607You may still choose to publish branches whose history is rewritten,
2608and it may be useful for others to be able to fetch those branches in
2609order to examine or test them, but they should not attempt to pull such
2610branches into their own work.
2611
2612For true distributed development that supports proper merging,
2613published branches should never be rewritten.
2614
2615[[bisect-merges]]
2616Why bisecting merge commits can be harder than bisecting linear history
2617-----------------------------------------------------------------------
2618
2619The linkgit:git-bisect[1] command correctly handles history that
2620includes merge commits.  However, when the commit that it finds is a
2621merge commit, the user may need to work harder than usual to figure out
2622why that commit introduced a problem.
2623
2624Imagine this history:
2625
2626................................................
2627      ---Z---o---X---...---o---A---C---D
2628          \                       /
2629           o---o---Y---...---o---B
2630................................................
2631
2632Suppose that on the upper line of development, the meaning of one
2633of the functions that exists at Z is changed at commit X.  The
2634commits from Z leading to A change both the function's
2635implementation and all calling sites that exist at Z, as well
2636as new calling sites they add, to be consistent.  There is no
2637bug at A.
2638
2639Suppose that in the meantime on the lower line of development somebody
2640adds a new calling site for that function at commit Y.  The
2641commits from Z leading to B all assume the old semantics of that
2642function and the callers and the callee are consistent with each
2643other.  There is no bug at B, either.
2644
2645Suppose further that the two development lines merge cleanly at C,
2646so no conflict resolution is required.
2647
2648Nevertheless, the code at C is broken, because the callers added
2649on the lower line of development have not been converted to the new
2650semantics introduced on the upper line of development.  So if all
2651you know is that D is bad, that Z is good, and that
2652linkgit:git-bisect[1] identifies C as the culprit, how will you
2653figure out that the problem is due to this change in semantics?
2654
2655When the result of a git-bisect is a non-merge commit, you should
2656normally be able to discover the problem by examining just that commit.
2657Developers can make this easy by breaking their changes into small
2658self-contained commits.  That won't help in the case above, however,
2659because the problem isn't obvious from examination of any single
2660commit; instead, a global view of the development is required.  To
2661make matters worse, the change in semantics in the problematic
2662function may be just one small part of the changes in the upper
2663line of development.
2664
2665On the other hand, if instead of merging at C you had rebased the
2666history between Z to B on top of A, you would have gotten this
2667linear history:
2668
2669................................................................
2670    ---Z---o---X--...---o---A---o---o---Y*--...---o---B*--D*
2671................................................................
2672
2673Bisecting between Z and D* would hit a single culprit commit Y*,
2674and understanding why Y* was broken would probably be easier.
2675
2676Partly for this reason, many experienced git users, even when
2677working on an otherwise merge-heavy project, keep the history
2678linear by rebasing against the latest upstream version before
2679publishing.
2680
2681[[advanced-branch-management]]
2682Advanced branch management
2683==========================
2684
2685[[fetching-individual-branches]]
2686Fetching individual branches
2687----------------------------
2688
2689Instead of using linkgit:git-remote[1], you can also choose just
2690to update one branch at a time, and to store it locally under an
2691arbitrary name:
2692
2693-------------------------------------------------
2694$ git fetch origin todo:my-todo-work
2695-------------------------------------------------
2696
2697The first argument, "origin", just tells git to fetch from the
2698repository you originally cloned from.  The second argument tells git
2699to fetch the branch named "todo" from the remote repository, and to
2700store it locally under the name refs/heads/my-todo-work.
2701
2702You can also fetch branches from other repositories; so
2703
2704-------------------------------------------------
2705$ git fetch git://example.com/proj.git master:example-master
2706-------------------------------------------------
2707
2708will create a new branch named "example-master" and store in it the
2709branch named "master" from the repository at the given URL.  If you
2710already have a branch named example-master, it will attempt to
2711<<fast-forwards,fast-forward>> to the commit given by example.com's
2712master branch.  In more detail:
2713
2714[[fetch-fast-forwards]]
2715git fetch and fast-forwards
2716---------------------------
2717
2718In the previous example, when updating an existing branch, "git-fetch"
2719checks to make sure that the most recent commit on the remote
2720branch is a descendant of the most recent commit on your copy of the
2721branch before updating your copy of the branch to point at the new
2722commit.  Git calls this process a <<fast-forwards,fast forward>>.
2723
2724A fast forward looks something like this:
2725
2726................................................
2727 o--o--o--o <-- old head of the branch
2728           \
2729            o--o--o <-- new head of the branch
2730................................................
2731
2732
2733In some cases it is possible that the new head will *not* actually be
2734a descendant of the old head.  For example, the developer may have
2735realized she made a serious mistake, and decided to backtrack,
2736resulting in a situation like:
2737
2738................................................
2739 o--o--o--o--a--b <-- old head of the branch
2740           \
2741            o--o--o <-- new head of the branch
2742................................................
2743
2744In this case, "git-fetch" will fail, and print out a warning.
2745
2746In that case, you can still force git to update to the new head, as
2747described in the following section.  However, note that in the
2748situation above this may mean losing the commits labeled "a" and "b",
2749unless you've already created a reference of your own pointing to
2750them.
2751
2752[[forcing-fetch]]
2753Forcing git-fetch to do non-fast-forward updates
2754------------------------------------------------
2755
2756If git fetch fails because the new head of a branch is not a
2757descendant of the old head, you may force the update with:
2758
2759-------------------------------------------------
2760$ git fetch git://example.com/proj.git +master:refs/remotes/example/master
2761-------------------------------------------------
2762
2763Note the addition of the "+" sign.  Alternatively, you can use the "-f"
2764flag to force updates of all the fetched branches, as in:
2765
2766-------------------------------------------------
2767$ git fetch -f origin
2768-------------------------------------------------
2769
2770Be aware that commits that the old version of example/master pointed at
2771may be lost, as we saw in the previous section.
2772
2773[[remote-branch-configuration]]
2774Configuring remote branches
2775---------------------------
2776
2777We saw above that "origin" is just a shortcut to refer to the
2778repository that you originally cloned from.  This information is
2779stored in git configuration variables, which you can see using
2780linkgit:git-config[1]:
2781
2782-------------------------------------------------
2783$ git config -l
2784core.repositoryformatversion=0
2785core.filemode=true
2786core.logallrefupdates=true
2787remote.origin.url=git://git.kernel.org/pub/scm/git/git.git
2788remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*
2789branch.master.remote=origin
2790branch.master.merge=refs/heads/master
2791-------------------------------------------------
2792
2793If there are other repositories that you also use frequently, you can
2794create similar configuration options to save typing; for example,
2795after
2796
2797-------------------------------------------------
2798$ git config remote.example.url git://example.com/proj.git
2799-------------------------------------------------
2800
2801then the following two commands will do the same thing:
2802
2803-------------------------------------------------
2804$ git fetch git://example.com/proj.git master:refs/remotes/example/master
2805$ git fetch example master:refs/remotes/example/master
2806-------------------------------------------------
2807
2808Even better, if you add one more option:
2809
2810-------------------------------------------------
2811$ git config remote.example.fetch master:refs/remotes/example/master
2812-------------------------------------------------
2813
2814then the following commands will all do the same thing:
2815
2816-------------------------------------------------
2817$ git fetch git://example.com/proj.git master:refs/remotes/example/master
2818$ git fetch example master:refs/remotes/example/master
2819$ git fetch example
2820-------------------------------------------------
2821
2822You can also add a "+" to force the update each time:
2823
2824-------------------------------------------------
2825$ git config remote.example.fetch +master:ref/remotes/example/master
2826-------------------------------------------------
2827
2828Don't do this unless you're sure you won't mind "git fetch" possibly
2829throwing away commits on 'example/master'.
2830
2831Also note that all of the above configuration can be performed by
2832directly editing the file .git/config instead of using
2833linkgit:git-config[1].
2834
2835See linkgit:git-config[1] for more details on the configuration
2836options mentioned above.
2837
2838
2839[[git-concepts]]
2840Git concepts
2841============
2842
2843Git is built on a small number of simple but powerful ideas.  While it
2844is possible to get things done without understanding them, you will find
2845git much more intuitive if you do.
2846
2847We start with the most important, the  <<def_object_database,object
2848database>> and the <<def_index,index>>.
2849
2850[[the-object-database]]
2851The Object Database
2852-------------------
2853
2854
2855We already saw in <<understanding-commits>> that all commits are stored
2856under a 40-digit "object name".  In fact, all the information needed to
2857represent the history of a project is stored in objects with such names.
2858In each case the name is calculated by taking the SHA1 hash of the
2859contents of the object.  The SHA1 hash is a cryptographic hash function.
2860What that means to us is that it is impossible to find two different
2861objects with the same name.  This has a number of advantages; among
2862others:
2863
2864- Git can quickly determine whether two objects are identical or not,
2865  just by comparing names.
2866- Since object names are computed the same way in every repository, the
2867  same content stored in two repositories will always be stored under
2868  the same name.
2869- Git can detect errors when it reads an object, by checking that the
2870  object's name is still the SHA1 hash of its contents.
2871
2872(See <<object-details>> for the details of the object formatting and
2873SHA1 calculation.)
2874
2875There are four different types of objects: "blob", "tree", "commit", and
2876"tag".
2877
2878- A <<def_blob_object,"blob" object>> is used to store file data.
2879- A <<def_tree_object,"tree" object>> ties one or more
2880  "blob" objects into a directory structure. In addition, a tree object
2881  can refer to other tree objects, thus creating a directory hierarchy.
2882- A <<def_commit_object,"commit" object>> ties such directory hierarchies
2883  together into a <<def_DAG,directed acyclic graph>> of revisions--each
2884  commit contains the object name of exactly one tree designating the
2885  directory hierarchy at the time of the commit. In addition, a commit
2886  refers to "parent" commit objects that describe the history of how we
2887  arrived at that directory hierarchy.
2888- A <<def_tag_object,"tag" object>> symbolically identifies and can be
2889  used to sign other objects. It contains the object name and type of
2890  another object, a symbolic name (of course!) and, optionally, a
2891  signature.
2892
2893The object types in some more detail:
2894
2895[[commit-object]]
2896Commit Object
2897~~~~~~~~~~~~~
2898
2899The "commit" object links a physical state of a tree with a description
2900of how we got there and why.  Use the --pretty=raw option to
2901linkgit:git-show[1] or linkgit:git-log[1] to examine your favorite
2902commit:
2903
2904------------------------------------------------
2905$ git show -s --pretty=raw 2be7fcb476
2906commit 2be7fcb4764f2dbcee52635b91fedb1b3dcf7ab4
2907tree fb3a8bdd0ceddd019615af4d57a53f43d8cee2bf
2908parent 257a84d9d02e90447b149af58b271c19405edb6a
2909author Dave Watson <dwatson@mimvista.com> 1187576872 -0400
2910committer Junio C Hamano <gitster@pobox.com> 1187591163 -0700
2911
2912    Fix misspelling of 'suppress' in docs
2913
2914    Signed-off-by: Junio C Hamano <gitster@pobox.com>
2915------------------------------------------------
2916
2917As you can see, a commit is defined by:
2918
2919- a tree: The SHA1 name of a tree object (as defined below), representing
2920  the contents of a directory at a certain point in time.
2921- parent(s): The SHA1 name of some number of commits which represent the
2922  immediately previous step(s) in the history of the project.  The
2923  example above has one parent; merge commits may have more than
2924  one.  A commit with no parents is called a "root" commit, and
2925  represents the initial revision of a project.  Each project must have
2926  at least one root.  A project can also have multiple roots, though
2927  that isn't common (or necessarily a good idea).
2928- an author: The name of the person responsible for this change, together
2929  with its date.
2930- a committer: The name of the person who actually created the commit,
2931  with the date it was done.  This may be different from the author, for
2932  example, if the author was someone who wrote a patch and emailed it
2933  to the person who used it to create the commit.
2934- a comment describing this commit.
2935
2936Note that a commit does not itself contain any information about what
2937actually changed; all changes are calculated by comparing the contents
2938of the tree referred to by this commit with the trees associated with
2939its parents.  In particular, git does not attempt to record file renames
2940explicitly, though it can identify cases where the existence of the same
2941file data at changing paths suggests a rename.  (See, for example, the
2942-M option to linkgit:git-diff[1]).
2943
2944A commit is usually created by linkgit:git-commit[1], which creates a
2945commit whose parent is normally the current HEAD, and whose tree is
2946taken from the content currently stored in the index.
2947
2948[[tree-object]]
2949Tree Object
2950~~~~~~~~~~~
2951
2952The ever-versatile linkgit:git-show[1] command can also be used to
2953examine tree objects, but linkgit:git-ls-tree[1] will give you more
2954details:
2955
2956------------------------------------------------
2957$ git ls-tree fb3a8bdd0ce
2958100644 blob 63c918c667fa005ff12ad89437f2fdc80926e21c    .gitignore
2959100644 blob 5529b198e8d14decbe4ad99db3f7fb632de0439d    .mailmap
2960100644 blob 6ff87c4664981e4397625791c8ea3bbb5f2279a3    COPYING
2961040000 tree 2fb783e477100ce076f6bf57e4a6f026013dc745    Documentation
2962100755 blob 3c0032cec592a765692234f1cba47dfdcc3a9200    GIT-VERSION-GEN
2963100644 blob 289b046a443c0647624607d471289b2c7dcd470b    INSTALL
2964100644 blob 4eb463797adc693dc168b926b6932ff53f17d0b1    Makefile
2965100644 blob 548142c327a6790ff8821d67c2ee1eff7a656b52    README
2966...
2967------------------------------------------------
2968
2969As you can see, a tree object contains a list of entries, each with a
2970mode, object type, SHA1 name, and name, sorted by name.  It represents
2971the contents of a single directory tree.
2972
2973The object type may be a blob, representing the contents of a file, or
2974another tree, representing the contents of a subdirectory.  Since trees
2975and blobs, like all other objects, are named by the SHA1 hash of their
2976contents, two trees have the same SHA1 name if and only if their
2977contents (including, recursively, the contents of all subdirectories)
2978are identical.  This allows git to quickly determine the differences
2979between two related tree objects, since it can ignore any entries with
2980identical object names.
2981
2982(Note: in the presence of submodules, trees may also have commits as
2983entries.  See <<submodules>> for documentation.)
2984
2985Note that the files all have mode 644 or 755: git actually only pays
2986attention to the executable bit.
2987
2988[[blob-object]]
2989Blob Object
2990~~~~~~~~~~~
2991
2992You can use linkgit:git-show[1] to examine the contents of a blob; take,
2993for example, the blob in the entry for "COPYING" from the tree above:
2994
2995------------------------------------------------
2996$ git show 6ff87c4664
2997
2998 Note that the only valid version of the GPL as far as this project
2999 is concerned is _this_ particular version of the license (ie v2, not
3000 v2.2 or v3.x or whatever), unless explicitly otherwise stated.
3001...
3002------------------------------------------------
3003
3004A "blob" object is nothing but a binary blob of data.  It doesn't refer
3005to anything else or have attributes of any kind.
3006
3007Since the blob is entirely defined by its data, if two files in a
3008directory tree (or in multiple different versions of the repository)
3009have the same contents, they will share the same blob object. The object
3010is totally independent of its location in the directory tree, and
3011renaming a file does not change the object that file is associated with.
3012
3013Note that any tree or blob object can be examined using
3014linkgit:git-show[1] with the <revision>:<path> syntax.  This can
3015sometimes be useful for browsing the contents of a tree that is not
3016currently checked out.
3017
3018[[trust]]
3019Trust
3020~~~~~
3021
3022If you receive the SHA1 name of a blob from one source, and its contents
3023from another (possibly untrusted) source, you can still trust that those
3024contents are correct as long as the SHA1 name agrees.  This is because
3025the SHA1 is designed so that it is infeasible to find different contents
3026that produce the same hash.
3027
3028Similarly, you need only trust the SHA1 name of a top-level tree object
3029to trust the contents of the entire directory that it refers to, and if
3030you receive the SHA1 name of a commit from a trusted source, then you
3031can easily verify the entire history of commits reachable through
3032parents of that commit, and all of those contents of the trees referred
3033to by those commits.
3034
3035So to introduce some real trust in the system, the only thing you need
3036to do is to digitally sign just 'one' special note, which includes the
3037name of a top-level commit.  Your digital signature shows others
3038that you trust that commit, and the immutability of the history of
3039commits tells others that they can trust the whole history.
3040
3041In other words, you can easily validate a whole archive by just
3042sending out a single email that tells the people the name (SHA1 hash)
3043of the top commit, and digitally sign that email using something
3044like GPG/PGP.
3045
3046To assist in this, git also provides the tag object...
3047
3048[[tag-object]]
3049Tag Object
3050~~~~~~~~~~
3051
3052A tag object contains an object, object type, tag name, the name of the
3053person ("tagger") who created the tag, and a message, which may contain
3054a signature, as can be seen using linkgit:git-cat-file[1]:
3055
3056------------------------------------------------
3057$ git cat-file tag v1.5.0
3058object 437b1b20df4b356c9342dac8d38849f24ef44f27
3059type commit
3060tag v1.5.0
3061tagger Junio C Hamano <junkio@cox.net> 1171411200 +0000
3062
3063GIT 1.5.0
3064-----BEGIN PGP SIGNATURE-----
3065Version: GnuPG v1.4.6 (GNU/Linux)
3066
3067iD8DBQBF0lGqwMbZpPMRm5oRAuRiAJ9ohBLd7s2kqjkKlq1qqC57SbnmzQCdG4ui
3068nLE/L9aUXdWeTFPron96DLA=
3069=2E+0
3070-----END PGP SIGNATURE-----
3071------------------------------------------------
3072
3073See the linkgit:git-tag[1] command to learn how to create and verify tag
3074objects.  (Note that linkgit:git-tag[1] can also be used to create
3075"lightweight tags", which are not tag objects at all, but just simple
3076references whose names begin with "refs/tags/").
3077
3078[[pack-files]]
3079How git stores objects efficiently: pack files
3080~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3081
3082Newly created objects are initially created in a file named after the
3083object's SHA1 hash (stored in .git/objects).
3084
3085Unfortunately this system becomes inefficient once a project has a
3086lot of objects.  Try this on an old project:
3087
3088------------------------------------------------
3089$ git count-objects
30906930 objects, 47620 kilobytes
3091------------------------------------------------
3092
3093The first number is the number of objects which are kept in
3094individual files.  The second is the amount of space taken up by
3095those "loose" objects.
3096
3097You can save space and make git faster by moving these loose objects in
3098to a "pack file", which stores a group of objects in an efficient
3099compressed format; the details of how pack files are formatted can be
3100found in link:technical/pack-format.txt[technical/pack-format.txt].
3101
3102To put the loose objects into a pack, just run git repack:
3103
3104------------------------------------------------
3105$ git repack
3106Generating pack...
3107Done counting 6020 objects.
3108Deltifying 6020 objects.
3109 100% (6020/6020) done
3110Writing 6020 objects.
3111 100% (6020/6020) done
3112Total 6020, written 6020 (delta 4070), reused 0 (delta 0)
3113Pack pack-3e54ad29d5b2e05838c75df582c65257b8d08e1c created.
3114------------------------------------------------
3115
3116You can then run
3117
3118------------------------------------------------
3119$ git prune
3120------------------------------------------------
3121
3122to remove any of the "loose" objects that are now contained in the
3123pack.  This will also remove any unreferenced objects (which may be
3124created when, for example, you use "git-reset" to remove a commit).
3125You can verify that the loose objects are gone by looking at the
3126.git/objects directory or by running
3127
3128------------------------------------------------
3129$ git count-objects
31300 objects, 0 kilobytes
3131------------------------------------------------
3132
3133Although the object files are gone, any commands that refer to those
3134objects will work exactly as they did before.
3135
3136The linkgit:git-gc[1] command performs packing, pruning, and more for
3137you, so is normally the only high-level command you need.
3138
3139[[dangling-objects]]
3140Dangling objects
3141~~~~~~~~~~~~~~~~
3142
3143The linkgit:git-fsck[1] command will sometimes complain about dangling
3144objects.  They are not a problem.
3145
3146The most common cause of dangling objects is that you've rebased a
3147branch, or you have pulled from somebody else who rebased a branch--see
3148<<cleaning-up-history>>.  In that case, the old head of the original
3149branch still exists, as does everything it pointed to. The branch
3150pointer itself just doesn't, since you replaced it with another one.
3151
3152There are also other situations that cause dangling objects. For
3153example, a "dangling blob" may arise because you did a "git-add" of a
3154file, but then, before you actually committed it and made it part of the
3155bigger picture, you changed something else in that file and committed
3156that *updated* thing--the old state that you added originally ends up
3157not being pointed to by any commit or tree, so it's now a dangling blob
3158object.
3159
3160Similarly, when the "recursive" merge strategy runs, and finds that
3161there are criss-cross merges and thus more than one merge base (which is
3162fairly unusual, but it does happen), it will generate one temporary
3163midway tree (or possibly even more, if you had lots of criss-crossing
3164merges and more than two merge bases) as a temporary internal merge
3165base, and again, those are real objects, but the end result will not end
3166up pointing to them, so they end up "dangling" in your repository.
3167
3168Generally, dangling objects aren't anything to worry about. They can
3169even be very useful: if you screw something up, the dangling objects can
3170be how you recover your old tree (say, you did a rebase, and realized
3171that you really didn't want to--you can look at what dangling objects
3172you have, and decide to reset your head to some old dangling state).
3173
3174For commits, you can just use:
3175
3176------------------------------------------------
3177$ gitk <dangling-commit-sha-goes-here> --not --all
3178------------------------------------------------
3179
3180This asks for all the history reachable from the given commit but not
3181from any branch, tag, or other reference.  If you decide it's something
3182you want, you can always create a new reference to it, e.g.,
3183
3184------------------------------------------------
3185$ git branch recovered-branch <dangling-commit-sha-goes-here>
3186------------------------------------------------
3187
3188For blobs and trees, you can't do the same, but you can still examine
3189them.  You can just do
3190
3191------------------------------------------------
3192$ git show <dangling-blob/tree-sha-goes-here>
3193------------------------------------------------
3194
3195to show what the contents of the blob were (or, for a tree, basically
3196what the "ls" for that directory was), and that may give you some idea
3197of what the operation was that left that dangling object.
3198
3199Usually, dangling blobs and trees aren't very interesting. They're
3200almost always the result of either being a half-way mergebase (the blob
3201will often even have the conflict markers from a merge in it, if you
3202have had conflicting merges that you fixed up by hand), or simply
3203because you interrupted a "git-fetch" with ^C or something like that,
3204leaving _some_ of the new objects in the object database, but just
3205dangling and useless.
3206
3207Anyway, once you are sure that you're not interested in any dangling
3208state, you can just prune all unreachable objects:
3209
3210------------------------------------------------
3211$ git prune
3212------------------------------------------------
3213
3214and they'll be gone. But you should only run "git prune" on a quiescent
3215repository--it's kind of like doing a filesystem fsck recovery: you
3216don't want to do that while the filesystem is mounted.
3217
3218(The same is true of "git-fsck" itself, btw, but since
3219git-fsck never actually *changes* the repository, it just reports
3220on what it found, git-fsck itself is never "dangerous" to run.
3221Running it while somebody is actually changing the repository can cause
3222confusing and scary messages, but it won't actually do anything bad. In
3223contrast, running "git prune" while somebody is actively changing the
3224repository is a *BAD* idea).
3225
3226[[recovering-from-repository-corruption]]
3227Recovering from repository corruption
3228~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3229
3230By design, git treats data trusted to it with caution.  However, even in
3231the absence of bugs in git itself, it is still possible that hardware or
3232operating system errors could corrupt data.
3233
3234The first defense against such problems is backups.  You can back up a
3235git directory using clone, or just using cp, tar, or any other backup
3236mechanism.
3237
3238As a last resort, you can search for the corrupted objects and attempt
3239to replace them by hand.  Back up your repository before attempting this
3240in case you corrupt things even more in the process.
3241
3242We'll assume that the problem is a single missing or corrupted blob,
3243which is sometimes a solvable problem.  (Recovering missing trees and
3244especially commits is *much* harder).
3245
3246Before starting, verify that there is corruption, and figure out where
3247it is with linkgit:git-fsck[1]; this may be time-consuming.
3248
3249Assume the output looks like this:
3250
3251------------------------------------------------
3252$ git fsck --full
3253broken link from    tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff8
3254              to    blob 4b9458b3786228369c63936db65827de3cc06200
3255missing blob 4b9458b3786228369c63936db65827de3cc06200
3256------------------------------------------------
3257
3258(Typically there will be some "dangling object" messages too, but they
3259aren't interesting.)
3260
3261Now you know that blob 4b9458b3 is missing, and that the tree 2d9263c6
3262points to it.  If you could find just one copy of that missing blob
3263object, possibly in some other repository, you could move it into
3264.git/objects/4b/9458b3... and be done.  Suppose you can't.  You can
3265still examine the tree that pointed to it with linkgit:git-ls-tree[1],
3266which might output something like:
3267
3268------------------------------------------------
3269$ git ls-tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff8
3270100644 blob 8d14531846b95bfa3564b58ccfb7913a034323b8    .gitignore
3271100644 blob ebf9bf84da0aab5ed944264a5db2a65fe3a3e883    .mailmap
3272100644 blob ca442d313d86dc67e0a2e5d584b465bd382cbf5c    COPYING
3273...
3274100644 blob 4b9458b3786228369c63936db65827de3cc06200    myfile
3275...
3276------------------------------------------------
3277
3278So now you know that the missing blob was the data for a file named
3279"myfile".  And chances are you can also identify the directory--let's
3280say it's in "somedirectory".  If you're lucky the missing copy might be
3281the same as the copy you have checked out in your working tree at
3282"somedirectory/myfile"; you can test whether that's right with
3283linkgit:git-hash-object[1]:
3284
3285------------------------------------------------
3286$ git hash-object -w somedirectory/myfile
3287------------------------------------------------
3288
3289which will create and store a blob object with the contents of
3290somedirectory/myfile, and output the sha1 of that object.  if you're
3291extremely lucky it might be 4b9458b3786228369c63936db65827de3cc06200, in
3292which case you've guessed right, and the corruption is fixed!
3293
3294Otherwise, you need more information.  How do you tell which version of
3295the file has been lost?
3296
3297The easiest way to do this is with:
3298
3299------------------------------------------------
3300$ git log --raw --all --full-history -- somedirectory/myfile
3301------------------------------------------------
3302
3303Because you're asking for raw output, you'll now get something like
3304
3305------------------------------------------------
3306commit abc
3307Author:
3308Date:
3309...
3310:100644 100644 4b9458b... newsha... M somedirectory/myfile
3311
3312
3313commit xyz
3314Author:
3315Date:
3316
3317...
3318:100644 100644 oldsha... 4b9458b... M somedirectory/myfile
3319------------------------------------------------
3320
3321This tells you that the immediately preceding version of the file was
3322"newsha", and that the immediately following version was "oldsha".
3323You also know the commit messages that went with the change from oldsha
3324to 4b9458b and with the change from 4b9458b to newsha.
3325
3326If you've been committing small enough changes, you may now have a good
3327shot at reconstructing the contents of the in-between state 4b9458b.
3328
3329If you can do that, you can now recreate the missing object with
3330
3331------------------------------------------------
3332$ git hash-object -w <recreated-file>
3333------------------------------------------------
3334
3335and your repository is good again!
3336
3337(Btw, you could have ignored the fsck, and started with doing a
3338
3339------------------------------------------------
3340$ git log --raw --all
3341------------------------------------------------
3342
3343and just looked for the sha of the missing object (4b9458b..) in that
3344whole thing. It's up to you - git does *have* a lot of information, it is
3345just missing one particular blob version.
3346
3347[[the-index]]
3348The index
3349-----------
3350
3351The index is a binary file (generally kept in .git/index) containing a
3352sorted list of path names, each with permissions and the SHA1 of a blob
3353object; linkgit:git-ls-files[1] can show you the contents of the index:
3354
3355-------------------------------------------------
3356$ git ls-files --stage
3357100644 63c918c667fa005ff12ad89437f2fdc80926e21c 0       .gitignore
3358100644 5529b198e8d14decbe4ad99db3f7fb632de0439d 0       .mailmap
3359100644 6ff87c4664981e4397625791c8ea3bbb5f2279a3 0       COPYING
3360100644 a37b2152bd26be2c2289e1f57a292534a51a93c7 0       Documentation/.gitignore
3361100644 fbefe9a45b00a54b58d94d06eca48b03d40a50e0 0       Documentation/Makefile
3362...
3363100644 2511aef8d89ab52be5ec6a5e46236b4b6bcd07ea 0       xdiff/xtypes.h
3364100644 2ade97b2574a9f77e7ae4002a4e07a6a38e46d07 0       xdiff/xutils.c
3365100644 d5de8292e05e7c36c4b68857c1cf9855e3d2f70a 0       xdiff/xutils.h
3366-------------------------------------------------
3367
3368Note that in older documentation you may see the index called the
3369"current directory cache" or just the "cache".  It has three important
3370properties:
3371
33721. The index contains all the information necessary to generate a single
3373(uniquely determined) tree object.
3374+
3375For example, running linkgit:git-commit[1] generates this tree object
3376from the index, stores it in the object database, and uses it as the
3377tree object associated with the new commit.
3378
33792. The index enables fast comparisons between the tree object it defines
3380and the working tree.
3381+
3382It does this by storing some additional data for each entry (such as
3383the last modified time).  This data is not displayed above, and is not
3384stored in the created tree object, but it can be used to determine
3385quickly which files in the working directory differ from what was
3386stored in the index, and thus save git from having to read all of the
3387data from such files to look for changes.
3388
33893. It can efficiently represent information about merge conflicts
3390between different tree objects, allowing each pathname to be
3391associated with sufficient information about the trees involved that
3392you can create a three-way merge between them.
3393+
3394We saw in <<conflict-resolution>> that during a merge the index can
3395store multiple versions of a single file (called "stages").  The third
3396column in the linkgit:git-ls-files[1] output above is the stage
3397number, and will take on values other than 0 for files with merge
3398conflicts.
3399
3400The index is thus a sort of temporary staging area, which is filled with
3401a tree which you are in the process of working on.
3402
3403If you blow the index away entirely, you generally haven't lost any
3404information as long as you have the name of the tree that it described.
3405
3406[[submodules]]
3407Submodules
3408==========
3409
3410Large projects are often composed of smaller, self-contained modules.  For
3411example, an embedded Linux distribution's source tree would include every
3412piece of software in the distribution with some local modifications; a movie
3413player might need to build against a specific, known-working version of a
3414decompression library; several independent programs might all share the same
3415build scripts.
3416
3417With centralized revision control systems this is often accomplished by
3418including every module in one single repository.  Developers can check out
3419all modules or only the modules they need to work with.  They can even modify
3420files across several modules in a single commit while moving things around
3421or updating APIs and translations.
3422
3423Git does not allow partial checkouts, so duplicating this approach in Git
3424would force developers to keep a local copy of modules they are not
3425interested in touching.  Commits in an enormous checkout would be slower
3426than you'd expect as Git would have to scan every directory for changes.
3427If modules have a lot of local history, clones would take forever.
3428
3429On the plus side, distributed revision control systems can much better
3430integrate with external sources.  In a centralized model, a single arbitrary
3431snapshot of the external project is exported from its own revision control
3432and then imported into the local revision control on a vendor branch.  All
3433the history is hidden.  With distributed revision control you can clone the
3434entire external history and much more easily follow development and re-merge
3435local changes.
3436
3437Git's submodule support allows a repository to contain, as a subdirectory, a
3438checkout of an external project.  Submodules maintain their own identity;
3439the submodule support just stores the submodule repository location and
3440commit ID, so other developers who clone the containing project
3441("superproject") can easily clone all the submodules at the same revision.
3442Partial checkouts of the superproject are possible: you can tell Git to
3443clone none, some or all of the submodules.
3444
3445The linkgit:git-submodule[1] command is available since Git 1.5.3.  Users
3446with Git 1.5.2 can look up the submodule commits in the repository and
3447manually check them out; earlier versions won't recognize the submodules at
3448all.
3449
3450To see how submodule support works, create (for example) four example
3451repositories that can be used later as a submodule:
3452
3453-------------------------------------------------
3454$ mkdir ~/git
3455$ cd ~/git
3456$ for i in a b c d
3457do
3458        mkdir $i
3459        cd $i
3460        git init
3461        echo "module $i" > $i.txt
3462        git add $i.txt
3463        git commit -m "Initial commit, submodule $i"
3464        cd ..
3465done
3466-------------------------------------------------
3467
3468Now create the superproject and add all the submodules:
3469
3470-------------------------------------------------
3471$ mkdir super
3472$ cd super
3473$ git init
3474$ for i in a b c d
3475do
3476        git submodule add ~/git/$i $i
3477done
3478-------------------------------------------------
3479
3480NOTE: Do not use local URLs here if you plan to publish your superproject!
3481
3482See what files `git-submodule` created:
3483
3484-------------------------------------------------
3485$ ls -a
3486.  ..  .git  .gitmodules  a  b  c  d
3487-------------------------------------------------
3488
3489The `git-submodule add <repo> <path>` command does a couple of things:
3490
3491- It clones the submodule from <repo> to the given <path> under the
3492  current directory and by default checks out the master branch.
3493- It adds the submodule's clone path to the linkgit:gitmodules[5] file and
3494  adds this file to the index, ready to be committed.
3495- It adds the submodule's current commit ID to the index, ready to be
3496  committed.
3497
3498Commit the superproject:
3499
3500-------------------------------------------------
3501$ git commit -m "Add submodules a, b, c and d."
3502-------------------------------------------------
3503
3504Now clone the superproject:
3505
3506-------------------------------------------------
3507$ cd ..
3508$ git clone super cloned
3509$ cd cloned
3510-------------------------------------------------
3511
3512The submodule directories are there, but they're empty:
3513
3514-------------------------------------------------
3515$ ls -a a
3516.  ..
3517$ git submodule status
3518-d266b9873ad50488163457f025db7cdd9683d88b a
3519-e81d457da15309b4fef4249aba9b50187999670d b
3520-c1536a972b9affea0f16e0680ba87332dc059146 c
3521-d96249ff5d57de5de093e6baff9e0aafa5276a74 d
3522-------------------------------------------------
3523
3524NOTE: The commit object names shown above would be different for you, but they
3525should match the HEAD commit object names of your repositories.  You can check
3526it by running `git ls-remote ../a`.
3527
3528Pulling down the submodules is a two-step process. First run `git submodule
3529init` to add the submodule repository URLs to `.git/config`:
3530
3531-------------------------------------------------
3532$ git submodule init
3533-------------------------------------------------
3534
3535Now use `git-submodule update` to clone the repositories and check out the
3536commits specified in the superproject:
3537
3538-------------------------------------------------
3539$ git submodule update
3540$ cd a
3541$ ls -a
3542.  ..  .git  a.txt
3543-------------------------------------------------
3544
3545One major difference between `git-submodule update` and `git-submodule add` is
3546that `git-submodule update` checks out a specific commit, rather than the tip
3547of a branch. It's like checking out a tag: the head is detached, so you're not
3548working on a branch.
3549
3550-------------------------------------------------
3551$ git branch
3552* (no branch)
3553  master
3554-------------------------------------------------
3555
3556If you want to make a change within a submodule and you have a detached head,
3557then you should create or checkout a branch, make your changes, publish the
3558change within the submodule, and then update the superproject to reference the
3559new commit:
3560
3561-------------------------------------------------
3562$ git checkout master
3563-------------------------------------------------
3564
3565or
3566
3567-------------------------------------------------
3568$ git checkout -b fix-up
3569-------------------------------------------------
3570
3571then
3572
3573-------------------------------------------------
3574$ echo "adding a line again" >> a.txt
3575$ git commit -a -m "Updated the submodule from within the superproject."
3576$ git push
3577$ cd ..
3578$ git diff
3579diff --git a/a b/a
3580index d266b98..261dfac 160000
3581--- a/a
3582+++ b/a
3583@@ -1 +1 @@
3584-Subproject commit d266b9873ad50488163457f025db7cdd9683d88b
3585+Subproject commit 261dfac35cb99d380eb966e102c1197139f7fa24
3586$ git add a
3587$ git commit -m "Updated submodule a."
3588$ git push
3589-------------------------------------------------
3590
3591You have to run `git submodule update` after `git pull` if you want to update
3592submodules, too.
3593
3594Pitfalls with submodules
3595------------------------
3596
3597Always publish the submodule change before publishing the change to the
3598superproject that references it. If you forget to publish the submodule change,
3599others won't be able to clone the repository:
3600
3601-------------------------------------------------
3602$ cd ~/git/super/a
3603$ echo i added another line to this file >> a.txt
3604$ git commit -a -m "doing it wrong this time"
3605$ cd ..
3606$ git add a
3607$ git commit -m "Updated submodule a again."
3608$ git push
3609$ cd ~/git/cloned
3610$ git pull
3611$ git submodule update
3612error: pathspec '261dfac35cb99d380eb966e102c1197139f7fa24' did not match any file(s) known to git.
3613Did you forget to 'git add'?
3614Unable to checkout '261dfac35cb99d380eb966e102c1197139f7fa24' in submodule path 'a'
3615-------------------------------------------------
3616
3617You also should not rewind branches in a submodule beyond commits that were
3618ever recorded in any superproject.
3619
3620It's not safe to run `git submodule update` if you've made and committed
3621changes within a submodule without checking out a branch first. They will be
3622silently overwritten:
3623
3624-------------------------------------------------
3625$ cat a.txt
3626module a
3627$ echo line added from private2 >> a.txt
3628$ git commit -a -m "line added inside private2"
3629$ cd ..
3630$ git submodule update
3631Submodule path 'a': checked out 'd266b9873ad50488163457f025db7cdd9683d88b'
3632$ cd a
3633$ cat a.txt
3634module a
3635-------------------------------------------------
3636
3637NOTE: The changes are still visible in the submodule's reflog.
3638
3639This is not the case if you did not commit your changes.
3640
3641[[low-level-operations]]
3642Low-level git operations
3643========================
3644
3645Many of the higher-level commands were originally implemented as shell
3646scripts using a smaller core of low-level git commands.  These can still
3647be useful when doing unusual things with git, or just as a way to
3648understand its inner workings.
3649
3650[[object-manipulation]]
3651Object access and manipulation
3652------------------------------
3653
3654The linkgit:git-cat-file[1] command can show the contents of any object,
3655though the higher-level linkgit:git-show[1] is usually more useful.
3656
3657The linkgit:git-commit-tree[1] command allows constructing commits with
3658arbitrary parents and trees.
3659
3660A tree can be created with linkgit:git-write-tree[1] and its data can be
3661accessed by linkgit:git-ls-tree[1].  Two trees can be compared with
3662linkgit:git-diff-tree[1].
3663
3664A tag is created with linkgit:git-mktag[1], and the signature can be
3665verified by linkgit:git-verify-tag[1], though it is normally simpler to
3666use linkgit:git-tag[1] for both.
3667
3668[[the-workflow]]
3669The Workflow
3670------------
3671
3672High-level operations such as linkgit:git-commit[1],
3673linkgit:git-checkout[1] and linkgit:git-reset[1] work by moving data
3674between the working tree, the index, and the object database.  Git
3675provides low-level operations which perform each of these steps
3676individually.
3677
3678Generally, all "git" operations work on the index file. Some operations
3679work *purely* on the index file (showing the current state of the
3680index), but most operations move data between the index file and either
3681the database or the working directory. Thus there are four main
3682combinations:
3683
3684[[working-directory-to-index]]
3685working directory -> index
3686~~~~~~~~~~~~~~~~~~~~~~~~~~
3687
3688The linkgit:git-update-index[1] command updates the index with
3689information from the working directory.  You generally update the
3690index information by just specifying the filename you want to update,
3691like so:
3692
3693-------------------------------------------------
3694$ git update-index filename
3695-------------------------------------------------
3696
3697but to avoid common mistakes with filename globbing etc, the command
3698will not normally add totally new entries or remove old entries,
3699i.e. it will normally just update existing cache entries.
3700
3701To tell git that yes, you really do realize that certain files no
3702longer exist, or that new files should be added, you
3703should use the `--remove` and `--add` flags respectively.
3704
3705NOTE! A `--remove` flag does 'not' mean that subsequent filenames will
3706necessarily be removed: if the files still exist in your directory
3707structure, the index will be updated with their new status, not
3708removed. The only thing `--remove` means is that update-index will be
3709considering a removed file to be a valid thing, and if the file really
3710does not exist any more, it will update the index accordingly.
3711
3712As a special case, you can also do `git update-index --refresh`, which
3713will refresh the "stat" information of each index to match the current
3714stat information. It will 'not' update the object status itself, and
3715it will only update the fields that are used to quickly test whether
3716an object still matches its old backing store object.
3717
3718The previously introduced linkgit:git-add[1] is just a wrapper for
3719linkgit:git-update-index[1].
3720
3721[[index-to-object-database]]
3722index -> object database
3723~~~~~~~~~~~~~~~~~~~~~~~~
3724
3725You write your current index file to a "tree" object with the program
3726
3727-------------------------------------------------
3728$ git write-tree
3729-------------------------------------------------
3730
3731that doesn't come with any options--it will just write out the
3732current index into the set of tree objects that describe that state,
3733and it will return the name of the resulting top-level tree. You can
3734use that tree to re-generate the index at any time by going in the
3735other direction:
3736
3737[[object-database-to-index]]
3738object database -> index
3739~~~~~~~~~~~~~~~~~~~~~~~~
3740
3741You read a "tree" file from the object database, and use that to
3742populate (and overwrite--don't do this if your index contains any
3743unsaved state that you might want to restore later!) your current
3744index.  Normal operation is just
3745
3746-------------------------------------------------
3747$ git read-tree <sha1 of tree>
3748-------------------------------------------------
3749
3750and your index file will now be equivalent to the tree that you saved
3751earlier. However, that is only your 'index' file: your working
3752directory contents have not been modified.
3753
3754[[index-to-working-directory]]
3755index -> working directory
3756~~~~~~~~~~~~~~~~~~~~~~~~~~
3757
3758You update your working directory from the index by "checking out"
3759files. This is not a very common operation, since normally you'd just
3760keep your files updated, and rather than write to your working
3761directory, you'd tell the index files about the changes in your
3762working directory (i.e. `git-update-index`).
3763
3764However, if you decide to jump to a new version, or check out somebody
3765else's version, or just restore a previous tree, you'd populate your
3766index file with read-tree, and then you need to check out the result
3767with
3768
3769-------------------------------------------------
3770$ git checkout-index filename
3771-------------------------------------------------
3772
3773or, if you want to check out all of the index, use `-a`.
3774
3775NOTE! git-checkout-index normally refuses to overwrite old files, so
3776if you have an old version of the tree already checked out, you will
3777need to use the "-f" flag ('before' the "-a" flag or the filename) to
3778'force' the checkout.
3779
3780
3781Finally, there are a few odds and ends which are not purely moving
3782from one representation to the other:
3783
3784[[tying-it-all-together]]
3785Tying it all together
3786~~~~~~~~~~~~~~~~~~~~~
3787
3788To commit a tree you have instantiated with "git write-tree", you'd
3789create a "commit" object that refers to that tree and the history
3790behind it--most notably the "parent" commits that preceded it in
3791history.
3792
3793Normally a "commit" has one parent: the previous state of the tree
3794before a certain change was made. However, sometimes it can have two
3795or more parent commits, in which case we call it a "merge", due to the
3796fact that such a commit brings together ("merges") two or more
3797previous states represented by other commits.
3798
3799In other words, while a "tree" represents a particular directory state
3800of a working directory, a "commit" represents that state in "time",
3801and explains how we got there.
3802
3803You create a commit object by giving it the tree that describes the
3804state at the time of the commit, and a list of parents:
3805
3806-------------------------------------------------
3807$ git commit-tree <tree> -p <parent> [-p <parent2> ..]
3808-------------------------------------------------
3809
3810and then giving the reason for the commit on stdin (either through
3811redirection from a pipe or file, or by just typing it at the tty).
3812
3813git-commit-tree will return the name of the object that represents
3814that commit, and you should save it away for later use. Normally,
3815you'd commit a new `HEAD` state, and while git doesn't care where you
3816save the note about that state, in practice we tend to just write the
3817result to the file pointed at by `.git/HEAD`, so that we can always see
3818what the last committed state was.
3819
3820Here is an ASCII art by Jon Loeliger that illustrates how
3821various pieces fit together.
3822
3823------------
3824
3825                     commit-tree
3826                      commit obj
3827                       +----+
3828                       |    |
3829                       |    |
3830                       V    V
3831                    +-----------+
3832                    | Object DB |
3833                    |  Backing  |
3834                    |   Store   |
3835                    +-----------+
3836                       ^
3837           write-tree  |     |
3838             tree obj  |     |
3839                       |     |  read-tree
3840                       |     |  tree obj
3841                             V
3842                    +-----------+
3843                    |   Index   |
3844                    |  "cache"  |
3845                    +-----------+
3846         update-index  ^
3847             blob obj  |     |
3848                       |     |
3849    checkout-index -u  |     |  checkout-index
3850             stat      |     |  blob obj
3851                             V
3852                    +-----------+
3853                    |  Working  |
3854                    | Directory |
3855                    +-----------+
3856
3857------------
3858
3859
3860[[examining-the-data]]
3861Examining the data
3862------------------
3863
3864You can examine the data represented in the object database and the
3865index with various helper tools. For every object, you can use
3866linkgit:git-cat-file[1] to examine details about the
3867object:
3868
3869-------------------------------------------------
3870$ git cat-file -t <objectname>
3871-------------------------------------------------
3872
3873shows the type of the object, and once you have the type (which is
3874usually implicit in where you find the object), you can use
3875
3876-------------------------------------------------
3877$ git cat-file blob|tree|commit|tag <objectname>
3878-------------------------------------------------
3879
3880to show its contents. NOTE! Trees have binary content, and as a result
3881there is a special helper for showing that content, called
3882`git-ls-tree`, which turns the binary content into a more easily
3883readable form.
3884
3885It's especially instructive to look at "commit" objects, since those
3886tend to be small and fairly self-explanatory. In particular, if you
3887follow the convention of having the top commit name in `.git/HEAD`,
3888you can do
3889
3890-------------------------------------------------
3891$ git cat-file commit HEAD
3892-------------------------------------------------
3893
3894to see what the top commit was.
3895
3896[[merging-multiple-trees]]
3897Merging multiple trees
3898----------------------
3899
3900Git helps you do a three-way merge, which you can expand to n-way by
3901repeating the merge procedure arbitrary times until you finally
3902"commit" the state.  The normal situation is that you'd only do one
3903three-way merge (two parents), and commit it, but if you like to, you
3904can do multiple parents in one go.
3905
3906To do a three-way merge, you need the two sets of "commit" objects
3907that you want to merge, use those to find the closest common parent (a
3908third "commit" object), and then use those commit objects to find the
3909state of the directory ("tree" object) at these points.
3910
3911To get the "base" for the merge, you first look up the common parent
3912of two commits with
3913
3914-------------------------------------------------
3915$ git merge-base <commit1> <commit2>
3916-------------------------------------------------
3917
3918which will return you the commit they are both based on.  You should
3919now look up the "tree" objects of those commits, which you can easily
3920do with (for example)
3921
3922-------------------------------------------------
3923$ git cat-file commit <commitname> | head -1
3924-------------------------------------------------
3925
3926since the tree object information is always the first line in a commit
3927object.
3928
3929Once you know the three trees you are going to merge (the one "original"
3930tree, aka the common tree, and the two "result" trees, aka the branches
3931you want to merge), you do a "merge" read into the index. This will
3932complain if it has to throw away your old index contents, so you should
3933make sure that you've committed those--in fact you would normally
3934always do a merge against your last commit (which should thus match what
3935you have in your current index anyway).
3936
3937To do the merge, do
3938
3939-------------------------------------------------
3940$ git read-tree -m -u <origtree> <yourtree> <targettree>
3941-------------------------------------------------
3942
3943which will do all trivial merge operations for you directly in the
3944index file, and you can just write the result out with
3945`git write-tree`.
3946
3947
3948[[merging-multiple-trees-2]]
3949Merging multiple trees, continued
3950---------------------------------
3951
3952Sadly, many merges aren't trivial. If there are files that have
3953been added, moved or removed, or if both branches have modified the
3954same file, you will be left with an index tree that contains "merge
3955entries" in it. Such an index tree can 'NOT' be written out to a tree
3956object, and you will have to resolve any such merge clashes using
3957other tools before you can write out the result.
3958
3959You can examine such index state with `git ls-files --unmerged`
3960command.  An example:
3961
3962------------------------------------------------
3963$ git read-tree -m $orig HEAD $target
3964$ git ls-files --unmerged
3965100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1       hello.c
3966100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2       hello.c
3967100644 cc44c73eb783565da5831b4d820c962954019b69 3       hello.c
3968------------------------------------------------
3969
3970Each line of the `git ls-files --unmerged` output begins with
3971the blob mode bits, blob SHA1, 'stage number', and the
3972filename.  The 'stage number' is git's way to say which tree it
3973came from: stage 1 corresponds to `$orig` tree, stage 2 `HEAD`
3974tree, and stage3 `$target` tree.
3975
3976Earlier we said that trivial merges are done inside
3977`git-read-tree -m`.  For example, if the file did not change
3978from `$orig` to `HEAD` nor `$target`, or if the file changed
3979from `$orig` to `HEAD` and `$orig` to `$target` the same way,
3980obviously the final outcome is what is in `HEAD`.  What the
3981above example shows is that file `hello.c` was changed from
3982`$orig` to `HEAD` and `$orig` to `$target` in a different way.
3983You could resolve this by running your favorite 3-way merge
3984program, e.g.  `diff3`, `merge`, or git's own merge-file, on
3985the blob objects from these three stages yourself, like this:
3986
3987------------------------------------------------
3988$ git cat-file blob 263414f... >hello.c~1
3989$ git cat-file blob 06fa6a2... >hello.c~2
3990$ git cat-file blob cc44c73... >hello.c~3
3991$ git merge-file hello.c~2 hello.c~1 hello.c~3
3992------------------------------------------------
3993
3994This would leave the merge result in `hello.c~2` file, along
3995with conflict markers if there are conflicts.  After verifying
3996the merge result makes sense, you can tell git what the final
3997merge result for this file is by:
3998
3999-------------------------------------------------
4000$ mv -f hello.c~2 hello.c
4001$ git update-index hello.c
4002-------------------------------------------------
4003
4004When a path is in the "unmerged" state, running `git-update-index` for
4005that path tells git to mark the path resolved.
4006
4007The above is the description of a git merge at the lowest level,
4008to help you understand what conceptually happens under the hood.
4009In practice, nobody, not even git itself, runs `git-cat-file` three times
4010for this.  There is a `git-merge-index` program that extracts the
4011stages to temporary files and calls a "merge" script on it:
4012
4013-------------------------------------------------
4014$ git merge-index git-merge-one-file hello.c
4015-------------------------------------------------
4016
4017and that is what higher level `git-merge -s resolve` is implemented with.
4018
4019[[hacking-git]]
4020Hacking git
4021===========
4022
4023This chapter covers internal details of the git implementation which
4024probably only git developers need to understand.
4025
4026[[object-details]]
4027Object storage format
4028---------------------
4029
4030All objects have a statically determined "type" which identifies the
4031format of the object (i.e. how it is used, and how it can refer to other
4032objects).  There are currently four different object types: "blob",
4033"tree", "commit", and "tag".
4034
4035Regardless of object type, all objects share the following
4036characteristics: they are all deflated with zlib, and have a header
4037that not only specifies their type, but also provides size information
4038about the data in the object.  It's worth noting that the SHA1 hash
4039that is used to name the object is the hash of the original data
4040plus this header, so `sha1sum` 'file' does not match the object name
4041for 'file'.
4042(Historical note: in the dawn of the age of git the hash
4043was the sha1 of the 'compressed' object.)
4044
4045As a result, the general consistency of an object can always be tested
4046independently of the contents or the type of the object: all objects can
4047be validated by verifying that (a) their hashes match the content of the
4048file and (b) the object successfully inflates to a stream of bytes that
4049forms a sequence of <ascii type without space> {plus} <space> {plus} <ascii decimal
4050size> {plus} <byte\0> {plus} <binary object data>.
4051
4052The structured objects can further have their structure and
4053connectivity to other objects verified. This is generally done with
4054the `git-fsck` program, which generates a full dependency graph
4055of all objects, and verifies their internal consistency (in addition
4056to just verifying their superficial consistency through the hash).
4057
4058[[birdview-on-the-source-code]]
4059A birds-eye view of Git's source code
4060-------------------------------------
4061
4062It is not always easy for new developers to find their way through Git's
4063source code.  This section gives you a little guidance to show where to
4064start.
4065
4066A good place to start is with the contents of the initial commit, with:
4067
4068----------------------------------------------------
4069$ git checkout e83c5163
4070----------------------------------------------------
4071
4072The initial revision lays the foundation for almost everything git has
4073today, but is small enough to read in one sitting.
4074
4075Note that terminology has changed since that revision.  For example, the
4076README in that revision uses the word "changeset" to describe what we
4077now call a <<def_commit_object,commit>>.
4078
4079Also, we do not call it "cache" any more, but rather "index"; however, the
4080file is still called `cache.h`.  Remark: Not much reason to change it now,
4081especially since there is no good single name for it anyway, because it is
4082basically _the_ header file which is included by _all_ of Git's C sources.
4083
4084If you grasp the ideas in that initial commit, you should check out a
4085more recent version and skim `cache.h`, `object.h` and `commit.h`.
4086
4087In the early days, Git (in the tradition of UNIX) was a bunch of programs
4088which were extremely simple, and which you used in scripts, piping the
4089output of one into another. This turned out to be good for initial
4090development, since it was easier to test new things.  However, recently
4091many of these parts have become builtins, and some of the core has been
4092"libified", i.e. put into libgit.a for performance, portability reasons,
4093and to avoid code duplication.
4094
4095By now, you know what the index is (and find the corresponding data
4096structures in `cache.h`), and that there are just a couple of object types
4097(blobs, trees, commits and tags) which inherit their common structure from
4098`struct object`, which is their first member (and thus, you can cast e.g.
4099`(struct object *)commit` to achieve the _same_ as `&commit->object`, i.e.
4100get at the object name and flags).
4101
4102Now is a good point to take a break to let this information sink in.
4103
4104Next step: get familiar with the object naming.  Read <<naming-commits>>.
4105There are quite a few ways to name an object (and not only revisions!).
4106All of these are handled in `sha1_name.c`. Just have a quick look at
4107the function `get_sha1()`. A lot of the special handling is done by
4108functions like `get_sha1_basic()` or the likes.
4109
4110This is just to get you into the groove for the most libified part of Git:
4111the revision walker.
4112
4113Basically, the initial version of `git-log` was a shell script:
4114
4115----------------------------------------------------------------
4116$ git-rev-list --pretty $(git-rev-parse --default HEAD "$@") | \
4117        LESS=-S ${PAGER:-less}
4118----------------------------------------------------------------
4119
4120What does this mean?
4121
4122`git-rev-list` is the original version of the revision walker, which
4123_always_ printed a list of revisions to stdout.  It is still functional,
4124and needs to, since most new Git programs start out as scripts using
4125`git-rev-list`.
4126
4127`git-rev-parse` is not as important any more; it was only used to filter out
4128options that were relevant for the different plumbing commands that were
4129called by the script.
4130
4131Most of what `git-rev-list` did is contained in `revision.c` and
4132`revision.h`.  It wraps the options in a struct named `rev_info`, which
4133controls how and what revisions are walked, and more.
4134
4135The original job of `git-rev-parse` is now taken by the function
4136`setup_revisions()`, which parses the revisions and the common command line
4137options for the revision walker. This information is stored in the struct
4138`rev_info` for later consumption. You can do your own command line option
4139parsing after calling `setup_revisions()`. After that, you have to call
4140`prepare_revision_walk()` for initialization, and then you can get the
4141commits one by one with the function `get_revision()`.
4142
4143If you are interested in more details of the revision walking process,
4144just have a look at the first implementation of `cmd_log()`; call
4145`git show v1.3.0{tilde}155^2{tilde}4` and scroll down to that function (note that you
4146no longer need to call `setup_pager()` directly).
4147
4148Nowadays, `git-log` is a builtin, which means that it is _contained_ in the
4149command `git`.  The source side of a builtin is
4150
4151- a function called `cmd_<bla>`, typically defined in `builtin-<bla>.c`,
4152  and declared in `builtin.h`,
4153
4154- an entry in the `commands[]` array in `git.c`, and
4155
4156- an entry in `BUILTIN_OBJECTS` in the `Makefile`.
4157
4158Sometimes, more than one builtin is contained in one source file.  For
4159example, `cmd_whatchanged()` and `cmd_log()` both reside in `builtin-log.c`,
4160since they share quite a bit of code.  In that case, the commands which are
4161_not_ named like the `.c` file in which they live have to be listed in
4162`BUILT_INS` in the `Makefile`.
4163
4164`git-log` looks more complicated in C than it does in the original script,
4165but that allows for a much greater flexibility and performance.
4166
4167Here again it is a good point to take a pause.
4168
4169Lesson three is: study the code.  Really, it is the best way to learn about
4170the organization of Git (after you know the basic concepts).
4171
4172So, think about something which you are interested in, say, "how can I
4173access a blob just knowing the object name of it?".  The first step is to
4174find a Git command with which you can do it.  In this example, it is either
4175`git-show` or `git-cat-file`.
4176
4177For the sake of clarity, let's stay with `git-cat-file`, because it
4178
4179- is plumbing, and
4180
4181- was around even in the initial commit (it literally went only through
4182  some 20 revisions as `cat-file.c`, was renamed to `builtin-cat-file.c`
4183  when made a builtin, and then saw less than 10 versions).
4184
4185So, look into `builtin-cat-file.c`, search for `cmd_cat_file()` and look what
4186it does.
4187
4188------------------------------------------------------------------
4189        git_config(git_default_config);
4190        if (argc != 3)
4191                usage("git-cat-file [-t|-s|-e|-p|<type>] <sha1>");
4192        if (get_sha1(argv[2], sha1))
4193                die("Not a valid object name %s", argv[2]);
4194------------------------------------------------------------------
4195
4196Let's skip over the obvious details; the only really interesting part
4197here is the call to `get_sha1()`.  It tries to interpret `argv[2]` as an
4198object name, and if it refers to an object which is present in the current
4199repository, it writes the resulting SHA-1 into the variable `sha1`.
4200
4201Two things are interesting here:
4202
4203- `get_sha1()` returns 0 on _success_.  This might surprise some new
4204  Git hackers, but there is a long tradition in UNIX to return different
4205  negative numbers in case of different errors--and 0 on success.
4206
4207- the variable `sha1` in the function signature of `get_sha1()` is `unsigned
4208  char \*`, but is actually expected to be a pointer to `unsigned
4209  char[20]`.  This variable will contain the 160-bit SHA-1 of the given
4210  commit.  Note that whenever a SHA-1 is passed as `unsigned char \*`, it
4211  is the binary representation, as opposed to the ASCII representation in
4212  hex characters, which is passed as `char *`.
4213
4214You will see both of these things throughout the code.
4215
4216Now, for the meat:
4217
4218-----------------------------------------------------------------------------
4219        case 0:
4220                buf = read_object_with_reference(sha1, argv[1], &size, NULL);
4221-----------------------------------------------------------------------------
4222
4223This is how you read a blob (actually, not only a blob, but any type of
4224object).  To know how the function `read_object_with_reference()` actually
4225works, find the source code for it (something like `git grep
4226read_object_with | grep ":[a-z]"` in the git repository), and read
4227the source.
4228
4229To find out how the result can be used, just read on in `cmd_cat_file()`:
4230
4231-----------------------------------
4232        write_or_die(1, buf, size);
4233-----------------------------------
4234
4235Sometimes, you do not know where to look for a feature.  In many such cases,
4236it helps to search through the output of `git log`, and then `git-show` the
4237corresponding commit.
4238
4239Example: If you know that there was some test case for `git-bundle`, but
4240do not remember where it was (yes, you _could_ `git grep bundle t/`, but that
4241does not illustrate the point!):
4242
4243------------------------
4244$ git log --no-merges t/
4245------------------------
4246
4247In the pager (`less`), just search for "bundle", go a few lines back,
4248and see that it is in commit 18449ab0...  Now just copy this object name,
4249and paste it into the command line
4250
4251-------------------
4252$ git show 18449ab0
4253-------------------
4254
4255Voila.
4256
4257Another example: Find out what to do in order to make some script a
4258builtin:
4259
4260-------------------------------------------------
4261$ git log --no-merges --diff-filter=A builtin-*.c
4262-------------------------------------------------
4263
4264You see, Git is actually the best tool to find out about the source of Git
4265itself!
4266
4267[[glossary]]
4268GIT Glossary
4269============
4270
4271include::glossary-content.txt[]
4272
4273[[git-quick-start]]
4274Appendix A: Git Quick Reference
4275===============================
4276
4277This is a quick summary of the major commands; the previous chapters
4278explain how these work in more detail.
4279
4280[[quick-creating-a-new-repository]]
4281Creating a new repository
4282-------------------------
4283
4284From a tarball:
4285
4286-----------------------------------------------
4287$ tar xzf project.tar.gz
4288$ cd project
4289$ git init
4290Initialized empty Git repository in .git/
4291$ git add .
4292$ git commit
4293-----------------------------------------------
4294
4295From a remote repository:
4296
4297-----------------------------------------------
4298$ git clone git://example.com/pub/project.git
4299$ cd project
4300-----------------------------------------------
4301
4302[[managing-branches]]
4303Managing branches
4304-----------------
4305
4306-----------------------------------------------
4307$ git branch         # list all local branches in this repo
4308$ git checkout test  # switch working directory to branch "test"
4309$ git branch new     # create branch "new" starting at current HEAD
4310$ git branch -d new  # delete branch "new"
4311-----------------------------------------------
4312
4313Instead of basing a new branch on current HEAD (the default), use:
4314
4315-----------------------------------------------
4316$ git branch new test    # branch named "test"
4317$ git branch new v2.6.15 # tag named v2.6.15
4318$ git branch new HEAD^   # commit before the most recent
4319$ git branch new HEAD^^  # commit before that
4320$ git branch new test~10 # ten commits before tip of branch "test"
4321-----------------------------------------------
4322
4323Create and switch to a new branch at the same time:
4324
4325-----------------------------------------------
4326$ git checkout -b new v2.6.15
4327-----------------------------------------------
4328
4329Update and examine branches from the repository you cloned from:
4330
4331-----------------------------------------------
4332$ git fetch             # update
4333$ git branch -r         # list
4334  origin/master
4335  origin/next
4336  ...
4337$ git checkout -b masterwork origin/master
4338-----------------------------------------------
4339
4340Fetch a branch from a different repository, and give it a new
4341name in your repository:
4342
4343-----------------------------------------------
4344$ git fetch git://example.com/project.git theirbranch:mybranch
4345$ git fetch git://example.com/project.git v2.6.15:mybranch
4346-----------------------------------------------
4347
4348Keep a list of repositories you work with regularly:
4349
4350-----------------------------------------------
4351$ git remote add example git://example.com/project.git
4352$ git remote                    # list remote repositories
4353example
4354origin
4355$ git remote show example       # get details
4356* remote example
4357  URL: git://example.com/project.git
4358  Tracked remote branches
4359    master next ...
4360$ git fetch example             # update branches from example
4361$ git branch -r                 # list all remote branches
4362-----------------------------------------------
4363
4364
4365[[exploring-history]]
4366Exploring history
4367-----------------
4368
4369-----------------------------------------------
4370$ gitk                      # visualize and browse history
4371$ git log                   # list all commits
4372$ git log src/              # ...modifying src/
4373$ git log v2.6.15..v2.6.16  # ...in v2.6.16, not in v2.6.15
4374$ git log master..test      # ...in branch test, not in branch master
4375$ git log test..master      # ...in branch master, but not in test
4376$ git log test...master     # ...in one branch, not in both
4377$ git log -S'foo()'         # ...where difference contain "foo()"
4378$ git log --since="2 weeks ago"
4379$ git log -p                # show patches as well
4380$ git show                  # most recent commit
4381$ git diff v2.6.15..v2.6.16 # diff between two tagged versions
4382$ git diff v2.6.15..HEAD    # diff with current head
4383$ git grep "foo()"          # search working directory for "foo()"
4384$ git grep v2.6.15 "foo()"  # search old tree for "foo()"
4385$ git show v2.6.15:a.txt    # look at old version of a.txt
4386-----------------------------------------------
4387
4388Search for regressions:
4389
4390-----------------------------------------------
4391$ git bisect start
4392$ git bisect bad                # current version is bad
4393$ git bisect good v2.6.13-rc2   # last known good revision
4394Bisecting: 675 revisions left to test after this
4395                                # test here, then:
4396$ git bisect good               # if this revision is good, or
4397$ git bisect bad                # if this revision is bad.
4398                                # repeat until done.
4399-----------------------------------------------
4400
4401[[making-changes]]
4402Making changes
4403--------------
4404
4405Make sure git knows who to blame:
4406
4407------------------------------------------------
4408$ cat >>~/.gitconfig <<\EOF
4409[user]
4410        name = Your Name Comes Here
4411        email = you@yourdomain.example.com
4412EOF
4413------------------------------------------------
4414
4415Select file contents to include in the next commit, then make the
4416commit:
4417
4418-----------------------------------------------
4419$ git add a.txt    # updated file
4420$ git add b.txt    # new file
4421$ git rm c.txt     # old file
4422$ git commit
4423-----------------------------------------------
4424
4425Or, prepare and create the commit in one step:
4426
4427-----------------------------------------------
4428$ git commit d.txt # use latest content only of d.txt
4429$ git commit -a    # use latest content of all tracked files
4430-----------------------------------------------
4431
4432[[merging]]
4433Merging
4434-------
4435
4436-----------------------------------------------
4437$ git merge test   # merge branch "test" into the current branch
4438$ git pull git://example.com/project.git master
4439                   # fetch and merge in remote branch
4440$ git pull . test  # equivalent to git merge test
4441-----------------------------------------------
4442
4443[[sharing-your-changes]]
4444Sharing your changes
4445--------------------
4446
4447Importing or exporting patches:
4448
4449-----------------------------------------------
4450$ git format-patch origin..HEAD # format a patch for each commit
4451                                # in HEAD but not in origin
4452$ git am mbox # import patches from the mailbox "mbox"
4453-----------------------------------------------
4454
4455Fetch a branch in a different git repository, then merge into the
4456current branch:
4457
4458-----------------------------------------------
4459$ git pull git://example.com/project.git theirbranch
4460-----------------------------------------------
4461
4462Store the fetched branch into a local branch before merging into the
4463current branch:
4464
4465-----------------------------------------------
4466$ git pull git://example.com/project.git theirbranch:mybranch
4467-----------------------------------------------
4468
4469After creating commits on a local branch, update the remote
4470branch with your commits:
4471
4472-----------------------------------------------
4473$ git push ssh://example.com/project.git mybranch:theirbranch
4474-----------------------------------------------
4475
4476When remote and local branch are both named "test":
4477
4478-----------------------------------------------
4479$ git push ssh://example.com/project.git test
4480-----------------------------------------------
4481
4482Shortcut version for a frequently used remote repository:
4483
4484-----------------------------------------------
4485$ git remote add example ssh://example.com/project.git
4486$ git push example test
4487-----------------------------------------------
4488
4489[[repository-maintenance]]
4490Repository maintenance
4491----------------------
4492
4493Check for corruption:
4494
4495-----------------------------------------------
4496$ git fsck
4497-----------------------------------------------
4498
4499Recompress, remove unused cruft:
4500
4501-----------------------------------------------
4502$ git gc
4503-----------------------------------------------
4504
4505
4506[[todo]]
4507Appendix B: Notes and todo list for this manual
4508===============================================
4509
4510This is a work in progress.
4511
4512The basic requirements:
4513
4514- It must be readable in order, from beginning to end, by someone
4515  intelligent with a basic grasp of the UNIX command line, but without
4516  any special knowledge of git.  If necessary, any other prerequisites
4517  should be specifically mentioned as they arise.
4518- Whenever possible, section headings should clearly describe the task
4519  they explain how to do, in language that requires no more knowledge
4520  than necessary: for example, "importing patches into a project" rather
4521  than "the git-am command"
4522
4523Think about how to create a clear chapter dependency graph that will
4524allow people to get to important topics without necessarily reading
4525everything in between.
4526
4527Scan Documentation/ for other stuff left out; in particular:
4528
4529- howto's
4530- some of technical/?
4531- hooks
4532- list of commands in linkgit:git[1]
4533
4534Scan email archives for other stuff left out
4535
4536Scan man pages to see if any assume more background than this manual
4537provides.
4538
4539Simplify beginning by suggesting disconnected head instead of
4540temporary branch creation?
4541
4542Add more good examples.  Entire sections of just cookbook examples
4543might be a good idea; maybe make an "advanced examples" section a
4544standard end-of-chapter section?
4545
4546Include cross-references to the glossary, where appropriate.
4547
4548Document shallow clones?  See draft 1.5.0 release notes for some
4549documentation.
4550
4551Add a section on working with other version control systems, including
4552CVS, Subversion, and just imports of series of release tarballs.
4553
4554More details on gitweb?
4555
4556Write a chapter on using plumbing and writing scripts.
4557
4558Alternates, clone -reference, etc.
4559
4560More on recovery from repository corruption.  See:
4561        http://marc.theaimsgroup.com/?l=git&m=117263864820799&w=2
4562        http://marc.theaimsgroup.com/?l=git&m=117147855503798&w=2