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, or linkgit:git-help[1] command. For example, for the command 22"git clone <repo>", you can either use: 23 24------------------------------------------------ 25$ man git-clone 26------------------------------------------------ 27 28or: 29 30------------------------------------------------ 31$ git help clone 32------------------------------------------------ 33 34With the latter, you can use the manual viewer of your choice; see 35linkgit:git-help[1] for more information. 36 37See also <<git-quick-start>> for a brief overview of git commands, 38without any explanation. 39 40Finally, see <<todo>> for ways that you can help make this manual more 41complete. 42 43 44[[repositories-and-branches]] 45Repositories and Branches 46========================= 47 48[[how-to-get-a-git-repository]] 49How to get a git repository 50--------------------------- 51 52It will be useful to have a git repository to experiment with as you 53read this manual. 54 55The best way to get one is by using the linkgit:git-clone[1] command to 56download a copy of an existing repository. If you don't already have a 57project in mind, here are some interesting examples: 58 59------------------------------------------------ 60 # git itself (approx. 10MB download): 61$ git clone git://git.kernel.org/pub/scm/git/git.git 62 # the Linux kernel (approx. 150MB download): 63$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git 64------------------------------------------------ 65 66The initial clone may be time-consuming for a large project, but you 67will only need to clone once. 68 69The clone command creates a new directory named after the project ("git" 70or "linux-2.6" in the examples above). After you cd into this 71directory, you will see that it contains a copy of the project files, 72called the <<def_working_tree,working tree>>, together with a special 73top-level directory named ".git", which contains all the information 74about the history of the project. 75 76[[how-to-check-out]] 77How to check out a different version of a project 78------------------------------------------------- 79 80Git is best thought of as a tool for storing the history of a collection 81of files. It stores the history as a compressed collection of 82interrelated snapshots of the project's contents. In git each such 83version is called a <<def_commit,commit>>. 84 85Those snapshots aren't necessarily all arranged in a single line from 86oldest to newest; instead, work may simultaneously proceed along 87parallel lines of development, called <<def_branch,branches>>, which may 88merge and diverge. 89 90A single git repository can track development on multiple branches. It 91does this by keeping a list of <<def_head,heads>> which reference the 92latest commit on each branch; the linkgit:git-branch[1] command shows 93you the list of branch heads: 94 95------------------------------------------------ 96$ git branch 97* master 98------------------------------------------------ 99 100A freshly cloned repository contains a single branch head, by default 101named "master", with the working directory initialized to the state of 102the project referred to by that branch head. 103 104Most projects also use <<def_tag,tags>>. Tags, like heads, are 105references into the project's history, and can be listed using the 106linkgit:git-tag[1] command: 107 108------------------------------------------------ 109$ git tag -l 110v2.6.11 111v2.6.11-tree 112v2.6.12 113v2.6.12-rc2 114v2.6.12-rc3 115v2.6.12-rc4 116v2.6.12-rc5 117v2.6.12-rc6 118v2.6.13 119... 120------------------------------------------------ 121 122Tags are expected to always point at the same version of a project, 123while heads are expected to advance as development progresses. 124 125Create a new branch head pointing to one of these versions and check it 126out using linkgit:git-checkout[1]: 127 128------------------------------------------------ 129$ git checkout -b new v2.6.13 130------------------------------------------------ 131 132The working directory then reflects the contents that the project had 133when it was tagged v2.6.13, and linkgit:git-branch[1] shows two 134branches, with an asterisk marking the currently checked-out branch: 135 136------------------------------------------------ 137$ git branch 138 master 139* new 140------------------------------------------------ 141 142If you decide that you'd rather see version 2.6.17, you can modify 143the current branch to point at v2.6.17 instead, with 144 145------------------------------------------------ 146$ git reset --hard v2.6.17 147------------------------------------------------ 148 149Note that if the current branch head was your only reference to a 150particular point in history, then resetting that branch may leave you 151with no way to find the history it used to point to; so use this command 152carefully. 153 154[[understanding-commits]] 155Understanding History: Commits 156------------------------------ 157 158Every change in the history of a project is represented by a commit. 159The linkgit:git-show[1] command shows the most recent commit on the 160current branch: 161 162------------------------------------------------ 163$ git show 164commit 17cf781661e6d38f737f15f53ab552f1e95960d7 165Author: Linus Torvalds <torvalds@ppc970.osdl.org.(none)> 166Date: Tue Apr 19 14:11:06 2005 -0700 167 168 Remove duplicate getenv(DB_ENVIRONMENT) call 169 170 Noted by Tony Luck. 171 172diff --git a/init-db.c b/init-db.c 173index 65898fa..b002dc6 100644 174--- a/init-db.c 175+++ b/init-db.c 176@@ -7,7 +7,7 @@ 177 178 int main(int argc, char **argv) 179 { 180- char *sha1_dir = getenv(DB_ENVIRONMENT), *path; 181+ char *sha1_dir, *path; 182 int len, i; 183 184 if (mkdir(".git", 0755) < 0) { 185------------------------------------------------ 186 187As you can see, a commit shows who made the latest change, what they 188did, and why. 189 190Every commit has a 40-hexdigit id, sometimes called the "object name" or the 191"SHA-1 id", shown on the first line of the "git show" output. You can usually 192refer to a commit by a shorter name, such as a tag or a branch name, but this 193longer name can also be useful. Most importantly, it is a globally unique 194name for this commit: so if you tell somebody else the object name (for 195example in email), then you are guaranteed that name will refer to the same 196commit in their repository that it does in yours (assuming their repository 197has that commit at all). Since the object name is computed as a hash over the 198contents of the commit, you are guaranteed that the commit can never change 199without its name also changing. 200 201In fact, in <<git-concepts>> we shall see that everything stored in git 202history, including file data and directory contents, is stored in an object 203with a name that is a hash of its contents. 204 205[[understanding-reachability]] 206Understanding history: commits, parents, and reachability 207~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 208 209Every commit (except the very first commit in a project) also has a 210parent commit which shows what happened before this commit. 211Following the chain of parents will eventually take you back to the 212beginning of the project. 213 214However, the commits do not form a simple list; git allows lines of 215development to diverge and then reconverge, and the point where two 216lines of development reconverge is called a "merge". The commit 217representing a merge can therefore have more than one parent, with 218each parent representing the most recent commit on one of the lines 219of development leading to that point. 220 221The best way to see how this works is using the linkgit:gitk[1] 222command; running gitk now on a git repository and looking for merge 223commits will help understand how the git organizes history. 224 225In the following, we say that commit X is "reachable" from commit Y 226if commit X is an ancestor of commit Y. Equivalently, you could say 227that Y is a descendant of X, or that there is a chain of parents 228leading from commit Y to commit X. 229 230[[history-diagrams]] 231Understanding history: History diagrams 232~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 233 234We will sometimes represent git history using diagrams like the one 235below. Commits are shown as "o", and the links between them with 236lines drawn with - / and \. Time goes left to right: 237 238 239................................................ 240 o--o--o <-- Branch A 241 / 242 o--o--o <-- master 243 \ 244 o--o--o <-- Branch B 245................................................ 246 247If we need to talk about a particular commit, the character "o" may 248be replaced with another letter or number. 249 250[[what-is-a-branch]] 251Understanding history: What is a branch? 252~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 253 254When we need to be precise, we will use the word "branch" to mean a line 255of development, and "branch head" (or just "head") to mean a reference 256to the most recent commit on a branch. In the example above, the branch 257head named "A" is a pointer to one particular commit, but we refer to 258the line of three commits leading up to that point as all being part of 259"branch A". 260 261However, when no confusion will result, we often just use the term 262"branch" both for branches and for branch heads. 263 264[[manipulating-branches]] 265Manipulating branches 266--------------------- 267 268Creating, deleting, and modifying branches is quick and easy; here's 269a summary of the commands: 270 271git branch:: 272 list all branches 273git branch <branch>:: 274 create a new branch named <branch>, referencing the same 275 point in history as the current branch 276git branch <branch> <start-point>:: 277 create a new branch named <branch>, referencing 278 <start-point>, which may be specified any way you like, 279 including using a branch name or a tag name 280git branch -d <branch>:: 281 delete the branch <branch>; if the branch you are deleting 282 points to a commit which is not reachable from the current 283 branch, this command will fail with a warning. 284git branch -D <branch>:: 285 even if the branch points to a commit not reachable 286 from the current branch, you may know that that commit 287 is still reachable from some other branch or tag. In that 288 case it is safe to use this command to force git to delete 289 the branch. 290git checkout <branch>:: 291 make the current branch <branch>, updating the working 292 directory to reflect the version referenced by <branch> 293git checkout -b <new> <start-point>:: 294 create a new branch <new> referencing <start-point>, and 295 check it out. 296 297The special symbol "HEAD" can always be used to refer to the current 298branch. In fact, git uses a file named "HEAD" in the .git directory to 299remember which branch is current: 300 301------------------------------------------------ 302$ cat .git/HEAD 303ref: refs/heads/master 304------------------------------------------------ 305 306[[detached-head]] 307Examining an old version without creating a new branch 308------------------------------------------------------ 309 310The `git checkout` command normally expects a branch head, but will also 311accept an arbitrary commit; for example, you can check out the commit 312referenced by a tag: 313 314------------------------------------------------ 315$ git checkout v2.6.17 316Note: moving to "v2.6.17" which isn't a local branch 317If you want to create a new branch from this checkout, you may do so 318(now or later) by using -b with the checkout command again. Example: 319 git checkout -b <new_branch_name> 320HEAD is now at 427abfa... Linux v2.6.17 321------------------------------------------------ 322 323The HEAD then refers to the SHA-1 of the commit instead of to a branch, 324and git branch shows that you are no longer on a branch: 325 326------------------------------------------------ 327$ cat .git/HEAD 328427abfa28afedffadfca9dd8b067eb6d36bac53f 329$ git branch 330* (no branch) 331 master 332------------------------------------------------ 333 334In this case we say that the HEAD is "detached". 335 336This is an easy way to check out a particular version without having to 337make up a name for the new branch. You can still create a new branch 338(or tag) for this version later if you decide to. 339 340[[examining-remote-branches]] 341Examining branches from a remote repository 342------------------------------------------- 343 344The "master" branch that was created at the time you cloned is a copy 345of the HEAD in the repository that you cloned from. That repository 346may also have had other branches, though, and your local repository 347keeps branches which track each of those remote branches, which you 348can view using the "-r" option to linkgit:git-branch[1]: 349 350------------------------------------------------ 351$ git branch -r 352 origin/HEAD 353 origin/html 354 origin/maint 355 origin/man 356 origin/master 357 origin/next 358 origin/pu 359 origin/todo 360------------------------------------------------ 361 362You cannot check out these remote-tracking branches, but you can 363examine them on a branch of your own, just as you would a tag: 364 365------------------------------------------------ 366$ git checkout -b my-todo-copy origin/todo 367------------------------------------------------ 368 369Note that the name "origin" is just the name that git uses by default 370to refer to the repository that you cloned from. 371 372[[how-git-stores-references]] 373Naming branches, tags, and other references 374------------------------------------------- 375 376Branches, remote-tracking branches, and tags are all references to 377commits. All references are named with a slash-separated path name 378starting with "refs"; the names we've been using so far are actually 379shorthand: 380 381 - The branch "test" is short for "refs/heads/test". 382 - The tag "v2.6.18" is short for "refs/tags/v2.6.18". 383 - "origin/master" is short for "refs/remotes/origin/master". 384 385The full name is occasionally useful if, for example, there ever 386exists a tag and a branch with the same name. 387 388(Newly created refs are actually stored in the .git/refs directory, 389under the path given by their name. However, for efficiency reasons 390they may also be packed together in a single file; see 391linkgit:git-pack-refs[1]). 392 393As another useful shortcut, the "HEAD" of a repository can be referred 394to just using the name of that repository. So, for example, "origin" 395is usually a shortcut for the HEAD branch in the repository "origin". 396 397For the complete list of paths which git checks for references, and 398the order it uses to decide which to choose when there are multiple 399references with the same shorthand name, see the "SPECIFYING 400REVISIONS" section of linkgit:git-rev-parse[1]. 401 402[[Updating-a-repository-With-git-fetch]] 403Updating a repository with git fetch 404------------------------------------ 405 406Eventually the developer cloned from will do additional work in her 407repository, creating new commits and advancing the branches to point 408at the new commits. 409 410The command "git fetch", with no arguments, will update all of the 411remote-tracking branches to the latest version found in her 412repository. It will not touch any of your own branches--not even the 413"master" branch that was created for you on clone. 414 415[[fetching-branches]] 416Fetching branches from other repositories 417----------------------------------------- 418 419You can also track branches from repositories other than the one you 420cloned from, using linkgit:git-remote[1]: 421 422------------------------------------------------- 423$ git remote add linux-nfs git://linux-nfs.org/pub/nfs-2.6.git 424$ git fetch linux-nfs 425* refs/remotes/linux-nfs/master: storing branch 'master' ... 426 commit: bf81b46 427------------------------------------------------- 428 429New remote-tracking branches will be stored under the shorthand name 430that you gave "git remote add", in this case linux-nfs: 431 432------------------------------------------------- 433$ git branch -r 434linux-nfs/master 435origin/master 436------------------------------------------------- 437 438If you run "git fetch <remote>" later, the tracking branches for the 439named <remote> will be updated. 440 441If you examine the file .git/config, you will see that git has added 442a new stanza: 443 444------------------------------------------------- 445$ cat .git/config 446... 447[remote "linux-nfs"] 448 url = git://linux-nfs.org/pub/nfs-2.6.git 449 fetch = +refs/heads/*:refs/remotes/linux-nfs/* 450... 451------------------------------------------------- 452 453This is what causes git to track the remote's branches; you may modify 454or delete these configuration options by editing .git/config with a 455text editor. (See the "CONFIGURATION FILE" section of 456linkgit:git-config[1] for details.) 457 458[[exploring-git-history]] 459Exploring git history 460===================== 461 462Git is best thought of as a tool for storing the history of a 463collection of files. It does this by storing compressed snapshots of 464the contents of a file hierarchy, together with "commits" which show 465the relationships between these snapshots. 466 467Git provides extremely flexible and fast tools for exploring the 468history of a project. 469 470We start with one specialized tool that is useful for finding the 471commit that introduced a bug into a project. 472 473[[using-bisect]] 474How to use bisect to find a regression 475-------------------------------------- 476 477Suppose version 2.6.18 of your project worked, but the version at 478"master" crashes. Sometimes the best way to find the cause of such a 479regression is to perform a brute-force search through the project's 480history to find the particular commit that caused the problem. The 481linkgit:git-bisect[1] command can help you do this: 482 483------------------------------------------------- 484$ git bisect start 485$ git bisect good v2.6.18 486$ git bisect bad master 487Bisecting: 3537 revisions left to test after this 488[65934a9a028b88e83e2b0f8b36618fe503349f8e] BLOCK: Make USB storage depend on SCSI rather than selecting it [try #6] 489------------------------------------------------- 490 491If you run "git branch" at this point, you'll see that git has 492temporarily moved you in "(no branch)". HEAD is now detached from any 493branch and points directly to a commit (with commit id 65934...) that 494is reachable from "master" but not from v2.6.18. Compile and test it, 495and see whether it crashes. Assume it does crash. Then: 496 497------------------------------------------------- 498$ git bisect bad 499Bisecting: 1769 revisions left to test after this 500[7eff82c8b1511017ae605f0c99ac275a7e21b867] i2c-core: Drop useless bitmaskings 501------------------------------------------------- 502 503checks out an older version. Continue like this, telling git at each 504stage whether the version it gives you is good or bad, and notice 505that the number of revisions left to test is cut approximately in 506half each time. 507 508After about 13 tests (in this case), it will output the commit id of 509the guilty commit. You can then examine the commit with 510linkgit:git-show[1], find out who wrote it, and mail them your bug 511report with the commit id. Finally, run 512 513------------------------------------------------- 514$ git bisect reset 515------------------------------------------------- 516 517to return you to the branch you were on before. 518 519Note that the version which `git bisect` checks out for you at each 520point is just a suggestion, and you're free to try a different 521version if you think it would be a good idea. For example, 522occasionally you may land on a commit that broke something unrelated; 523run 524 525------------------------------------------------- 526$ git bisect visualize 527------------------------------------------------- 528 529which will run gitk and label the commit it chose with a marker that 530says "bisect". Choose a safe-looking commit nearby, note its commit 531id, and check it out with: 532 533------------------------------------------------- 534$ git reset --hard fb47ddb2db... 535------------------------------------------------- 536 537then test, run "bisect good" or "bisect bad" as appropriate, and 538continue. 539 540Instead of "git bisect visualize" and then "git reset --hard 541fb47ddb2db...", you might just want to tell git that you want to skip 542the current commit: 543 544------------------------------------------------- 545$ git bisect skip 546------------------------------------------------- 547 548In this case, though, git may not eventually be able to tell the first 549bad one between some first skipped commits and a later bad commit. 550 551There are also ways to automate the bisecting process if you have a 552test script that can tell a good from a bad commit. See 553linkgit:git-bisect[1] for more information about this and other "git 554bisect" features. 555 556[[naming-commits]] 557Naming commits 558-------------- 559 560We have seen several ways of naming commits already: 561 562 - 40-hexdigit object name 563 - branch name: refers to the commit at the head of the given 564 branch 565 - tag name: refers to the commit pointed to by the given tag 566 (we've seen branches and tags are special cases of 567 <<how-git-stores-references,references>>). 568 - HEAD: refers to the head of the current branch 569 570There are many more; see the "SPECIFYING REVISIONS" section of the 571linkgit:git-rev-parse[1] man page for the complete list of ways to 572name revisions. Some examples: 573 574------------------------------------------------- 575$ git show fb47ddb2 # the first few characters of the object name 576 # are usually enough to specify it uniquely 577$ git show HEAD^ # the parent of the HEAD commit 578$ git show HEAD^^ # the grandparent 579$ git show HEAD~4 # the great-great-grandparent 580------------------------------------------------- 581 582Recall that merge commits may have more than one parent; by default, 583^ and ~ follow the first parent listed in the commit, but you can 584also choose: 585 586------------------------------------------------- 587$ git show HEAD^1 # show the first parent of HEAD 588$ git show HEAD^2 # show the second parent of HEAD 589------------------------------------------------- 590 591In addition to HEAD, there are several other special names for 592commits: 593 594Merges (to be discussed later), as well as operations such as 595`git reset`, which change the currently checked-out commit, generally 596set ORIG_HEAD to the value HEAD had before the current operation. 597 598The `git fetch` operation always stores the head of the last fetched 599branch in FETCH_HEAD. For example, if you run `git fetch` without 600specifying a local branch as the target of the operation 601 602------------------------------------------------- 603$ git fetch git://example.com/proj.git theirbranch 604------------------------------------------------- 605 606the fetched commits will still be available from FETCH_HEAD. 607 608When we discuss merges we'll also see the special name MERGE_HEAD, 609which refers to the other branch that we're merging in to the current 610branch. 611 612The linkgit:git-rev-parse[1] command is a low-level command that is 613occasionally useful for translating some name for a commit to the object 614name for that commit: 615 616------------------------------------------------- 617$ git rev-parse origin 618e05db0fd4f31dde7005f075a84f96b360d05984b 619------------------------------------------------- 620 621[[creating-tags]] 622Creating tags 623------------- 624 625We can also create a tag to refer to a particular commit; after 626running 627 628------------------------------------------------- 629$ git tag stable-1 1b2e1d63ff 630------------------------------------------------- 631 632You can use stable-1 to refer to the commit 1b2e1d63ff. 633 634This creates a "lightweight" tag. If you would also like to include a 635comment with the tag, and possibly sign it cryptographically, then you 636should create a tag object instead; see the linkgit:git-tag[1] man page 637for details. 638 639[[browsing-revisions]] 640Browsing revisions 641------------------ 642 643The linkgit:git-log[1] command can show lists of commits. On its 644own, it shows all commits reachable from the parent commit; but you 645can also make more specific requests: 646 647------------------------------------------------- 648$ git log v2.5.. # commits since (not reachable from) v2.5 649$ git log test..master # commits reachable from master but not test 650$ git log master..test # ...reachable from test but not master 651$ git log master...test # ...reachable from either test or master, 652 # but not both 653$ git log --since="2 weeks ago" # commits from the last 2 weeks 654$ git log Makefile # commits which modify Makefile 655$ git log fs/ # ... which modify any file under fs/ 656$ git log -S'foo()' # commits which add or remove any file data 657 # matching the string 'foo()' 658------------------------------------------------- 659 660And of course you can combine all of these; the following finds 661commits since v2.5 which touch the Makefile or any file under fs: 662 663------------------------------------------------- 664$ git log v2.5.. Makefile fs/ 665------------------------------------------------- 666 667You can also ask git log to show patches: 668 669------------------------------------------------- 670$ git log -p 671------------------------------------------------- 672 673See the "--pretty" option in the linkgit:git-log[1] man page for more 674display options. 675 676Note that git log starts with the most recent commit and works 677backwards through the parents; however, since git history can contain 678multiple independent lines of development, the particular order that 679commits are listed in may be somewhat arbitrary. 680 681[[generating-diffs]] 682Generating diffs 683---------------- 684 685You can generate diffs between any two versions using 686linkgit:git-diff[1]: 687 688------------------------------------------------- 689$ git diff master..test 690------------------------------------------------- 691 692That will produce the diff between the tips of the two branches. If 693you'd prefer to find the diff from their common ancestor to test, you 694can use three dots instead of two: 695 696------------------------------------------------- 697$ git diff master...test 698------------------------------------------------- 699 700Sometimes what you want instead is a set of patches; for this you can 701use linkgit:git-format-patch[1]: 702 703------------------------------------------------- 704$ git format-patch master..test 705------------------------------------------------- 706 707will generate a file with a patch for each commit reachable from test 708but not from master. 709 710[[viewing-old-file-versions]] 711Viewing old file versions 712------------------------- 713 714You can always view an old version of a file by just checking out the 715correct revision first. But sometimes it is more convenient to be 716able to view an old version of a single file without checking 717anything out; this command does that: 718 719------------------------------------------------- 720$ git show v2.5:fs/locks.c 721------------------------------------------------- 722 723Before the colon may be anything that names a commit, and after it 724may be any path to a file tracked by git. 725 726[[history-examples]] 727Examples 728-------- 729 730[[counting-commits-on-a-branch]] 731Counting the number of commits on a branch 732~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 733 734Suppose you want to know how many commits you've made on "mybranch" 735since it diverged from "origin": 736 737------------------------------------------------- 738$ git log --pretty=oneline origin..mybranch | wc -l 739------------------------------------------------- 740 741Alternatively, you may often see this sort of thing done with the 742lower-level command linkgit:git-rev-list[1], which just lists the SHA-1's 743of all the given commits: 744 745------------------------------------------------- 746$ git rev-list origin..mybranch | wc -l 747------------------------------------------------- 748 749[[checking-for-equal-branches]] 750Check whether two branches point at the same history 751~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 752 753Suppose you want to check whether two branches point at the same point 754in history. 755 756------------------------------------------------- 757$ git diff origin..master 758------------------------------------------------- 759 760will tell you whether the contents of the project are the same at the 761two branches; in theory, however, it's possible that the same project 762contents could have been arrived at by two different historical 763routes. You could compare the object names: 764 765------------------------------------------------- 766$ git rev-list origin 767e05db0fd4f31dde7005f075a84f96b360d05984b 768$ git rev-list master 769e05db0fd4f31dde7005f075a84f96b360d05984b 770------------------------------------------------- 771 772Or you could recall that the ... operator selects all commits 773contained reachable from either one reference or the other but not 774both: so 775 776------------------------------------------------- 777$ git log origin...master 778------------------------------------------------- 779 780will return no commits when the two branches are equal. 781 782[[finding-tagged-descendants]] 783Find first tagged version including a given fix 784~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 785 786Suppose you know that the commit e05db0fd fixed a certain problem. 787You'd like to find the earliest tagged release that contains that 788fix. 789 790Of course, there may be more than one answer--if the history branched 791after commit e05db0fd, then there could be multiple "earliest" tagged 792releases. 793 794You could just visually inspect the commits since e05db0fd: 795 796------------------------------------------------- 797$ gitk e05db0fd.. 798------------------------------------------------- 799 800Or you can use linkgit:git-name-rev[1], which will give the commit a 801name based on any tag it finds pointing to one of the commit's 802descendants: 803 804------------------------------------------------- 805$ git name-rev --tags e05db0fd 806e05db0fd tags/v1.5.0-rc1^0~23 807------------------------------------------------- 808 809The linkgit:git-describe[1] command does the opposite, naming the 810revision using a tag on which the given commit is based: 811 812------------------------------------------------- 813$ git describe e05db0fd 814v1.5.0-rc0-260-ge05db0f 815------------------------------------------------- 816 817but that may sometimes help you guess which tags might come after the 818given commit. 819 820If you just want to verify whether a given tagged version contains a 821given commit, you could use linkgit:git-merge-base[1]: 822 823------------------------------------------------- 824$ git merge-base e05db0fd v1.5.0-rc1 825e05db0fd4f31dde7005f075a84f96b360d05984b 826------------------------------------------------- 827 828The merge-base command finds a common ancestor of the given commits, 829and always returns one or the other in the case where one is a 830descendant of the other; so the above output shows that e05db0fd 831actually is an ancestor of v1.5.0-rc1. 832 833Alternatively, note that 834 835------------------------------------------------- 836$ git log v1.5.0-rc1..e05db0fd 837------------------------------------------------- 838 839will produce empty output if and only if v1.5.0-rc1 includes e05db0fd, 840because it outputs only commits that are not reachable from v1.5.0-rc1. 841 842As yet another alternative, the linkgit:git-show-branch[1] command lists 843the commits reachable from its arguments with a display on the left-hand 844side that indicates which arguments that commit is reachable from. So, 845you can run something like 846 847------------------------------------------------- 848$ git show-branch e05db0fd v1.5.0-rc0 v1.5.0-rc1 v1.5.0-rc2 849! [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if 850available 851 ! [v1.5.0-rc0] GIT v1.5.0 preview 852 ! [v1.5.0-rc1] GIT v1.5.0-rc1 853 ! [v1.5.0-rc2] GIT v1.5.0-rc2 854... 855------------------------------------------------- 856 857then search for a line that looks like 858 859------------------------------------------------- 860+ ++ [e05db0fd] Fix warnings in sha1_file.c - use C99 printf format if 861available 862------------------------------------------------- 863 864Which shows that e05db0fd is reachable from itself, from v1.5.0-rc1, and 865from v1.5.0-rc2, but not from v1.5.0-rc0. 866 867[[showing-commits-unique-to-a-branch]] 868Showing commits unique to a given branch 869~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 870 871Suppose you would like to see all the commits reachable from the branch 872head named "master" but not from any other head in your repository. 873 874We can list all the heads in this repository with 875linkgit:git-show-ref[1]: 876 877------------------------------------------------- 878$ git show-ref --heads 879bf62196b5e363d73353a9dcf094c59595f3153b7 refs/heads/core-tutorial 880db768d5504c1bb46f63ee9d6e1772bd047e05bf9 refs/heads/maint 881a07157ac624b2524a059a3414e99f6f44bebc1e7 refs/heads/master 88224dbc180ea14dc1aebe09f14c8ecf32010690627 refs/heads/tutorial-2 8831e87486ae06626c2f31eaa63d26fc0fd646c8af2 refs/heads/tutorial-fixes 884------------------------------------------------- 885 886We can get just the branch-head names, and remove "master", with 887the help of the standard utilities cut and grep: 888 889------------------------------------------------- 890$ git show-ref --heads | cut -d' ' -f2 | grep -v '^refs/heads/master' 891refs/heads/core-tutorial 892refs/heads/maint 893refs/heads/tutorial-2 894refs/heads/tutorial-fixes 895------------------------------------------------- 896 897And then we can ask to see all the commits reachable from master 898but not from these other heads: 899 900------------------------------------------------- 901$ gitk master --not $( git show-ref --heads | cut -d' ' -f2 | 902 grep -v '^refs/heads/master' ) 903------------------------------------------------- 904 905Obviously, endless variations are possible; for example, to see all 906commits reachable from some head but not from any tag in the repository: 907 908------------------------------------------------- 909$ gitk $( git show-ref --heads ) --not $( git show-ref --tags ) 910------------------------------------------------- 911 912(See linkgit:git-rev-parse[1] for explanations of commit-selecting 913syntax such as `--not`.) 914 915[[making-a-release]] 916Creating a changelog and tarball for a software release 917~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 918 919The linkgit:git-archive[1] command can create a tar or zip archive from 920any version of a project; for example: 921 922------------------------------------------------- 923$ git archive --format=tar --prefix=project/ HEAD | gzip >latest.tar.gz 924------------------------------------------------- 925 926will use HEAD to produce a tar archive in which each filename is 927preceded by "project/". 928 929If you're releasing a new version of a software project, you may want 930to simultaneously make a changelog to include in the release 931announcement. 932 933Linus Torvalds, for example, makes new kernel releases by tagging them, 934then running: 935 936------------------------------------------------- 937$ release-script 2.6.12 2.6.13-rc6 2.6.13-rc7 938------------------------------------------------- 939 940where release-script is a shell script that looks like: 941 942------------------------------------------------- 943#!/bin/sh 944stable="$1" 945last="$2" 946new="$3" 947echo "# git tag v$new" 948echo "git archive --prefix=linux-$new/ v$new | gzip -9 > ../linux-$new.tar.gz" 949echo "git diff v$stable v$new | gzip -9 > ../patch-$new.gz" 950echo "git log --no-merges v$new ^v$last > ../ChangeLog-$new" 951echo "git shortlog --no-merges v$new ^v$last > ../ShortLog" 952echo "git diff --stat --summary -M v$last v$new > ../diffstat-$new" 953------------------------------------------------- 954 955and then he just cut-and-pastes the output commands after verifying that 956they look OK. 957 958[[Finding-comments-With-given-Content]] 959Finding commits referencing a file with given content 960~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 961 962Somebody hands you a copy of a file, and asks which commits modified a 963file such that it contained the given content either before or after the 964commit. You can find out with this: 965 966------------------------------------------------- 967$ git log --raw --abbrev=40 --pretty=oneline | 968 grep -B 1 `git hash-object filename` 969------------------------------------------------- 970 971Figuring out why this works is left as an exercise to the (advanced) 972student. The linkgit:git-log[1], linkgit:git-diff-tree[1], and 973linkgit:git-hash-object[1] man pages may prove helpful. 974 975[[Developing-With-git]] 976Developing with git 977=================== 978 979[[telling-git-your-name]] 980Telling git your name 981--------------------- 982 983Before creating any commits, you should introduce yourself to git. The 984easiest way to do so is to make sure the following lines appear in a 985file named .gitconfig in your home directory: 986 987------------------------------------------------ 988[user] 989 name = Your Name Comes Here 990 email = you@yourdomain.example.com 991------------------------------------------------ 992 993(See the "CONFIGURATION FILE" section of linkgit:git-config[1] for 994details on the configuration file.) 995 996 997[[creating-a-new-repository]] 998Creating a new repository 999-------------------------10001001Creating a new repository from scratch is very easy:10021003-------------------------------------------------1004$ mkdir project1005$ cd project1006$ git init1007-------------------------------------------------10081009If you have some initial content (say, a tarball):10101011-------------------------------------------------1012$ tar xzvf project.tar.gz1013$ cd project1014$ git init1015$ git add . # include everything below ./ in the first commit:1016$ git commit1017-------------------------------------------------10181019[[how-to-make-a-commit]]1020How to make a commit1021--------------------10221023Creating a new commit takes three steps:10241025 1. Making some changes to the working directory using your1026 favorite editor.1027 2. Telling git about your changes.1028 3. Creating the commit using the content you told git about1029 in step 2.10301031In practice, you can interleave and repeat steps 1 and 2 as many1032times as you want: in order to keep track of what you want committed1033at step 3, git maintains a snapshot of the tree's contents in a1034special staging area called "the index."10351036At the beginning, the content of the index will be identical to1037that of the HEAD. The command "git diff --cached", which shows1038the difference between the HEAD and the index, should therefore1039produce no output at that point.10401041Modifying the index is easy:10421043To update the index with the new contents of a modified file, use10441045-------------------------------------------------1046$ git add path/to/file1047-------------------------------------------------10481049To add the contents of a new file to the index, use10501051-------------------------------------------------1052$ git add path/to/file1053-------------------------------------------------10541055To remove a file from the index and from the working tree,10561057-------------------------------------------------1058$ git rm path/to/file1059-------------------------------------------------10601061After each step you can verify that10621063-------------------------------------------------1064$ git diff --cached1065-------------------------------------------------10661067always shows the difference between the HEAD and the index file--this1068is what you'd commit if you created the commit now--and that10691070-------------------------------------------------1071$ git diff1072-------------------------------------------------10731074shows the difference between the working tree and the index file.10751076Note that "git add" always adds just the current contents of a file1077to the index; further changes to the same file will be ignored unless1078you run `git add` on the file again.10791080When you're ready, just run10811082-------------------------------------------------1083$ git commit1084-------------------------------------------------10851086and git will prompt you for a commit message and then create the new1087commit. Check to make sure it looks like what you expected with10881089-------------------------------------------------1090$ git show1091-------------------------------------------------10921093As a special shortcut,10941095-------------------------------------------------1096$ git commit -a1097-------------------------------------------------10981099will update the index with any files that you've modified or removed1100and create a commit, all in one step.11011102A number of commands are useful for keeping track of what you're1103about to commit:11041105-------------------------------------------------1106$ git diff --cached # difference between HEAD and the index; what1107 # would be committed if you ran "commit" now.1108$ git diff # difference between the index file and your1109 # working directory; changes that would not1110 # be included if you ran "commit" now.1111$ git diff HEAD # difference between HEAD and working tree; what1112 # would be committed if you ran "commit -a" now.1113$ git status # a brief per-file summary of the above.1114-------------------------------------------------11151116You can also use linkgit:git-gui[1] to create commits, view changes in1117the index and the working tree files, and individually select diff hunks1118for inclusion in the index (by right-clicking on the diff hunk and1119choosing "Stage Hunk For Commit").11201121[[creating-good-commit-messages]]1122Creating good commit messages1123-----------------------------11241125Though not required, it's a good idea to begin the commit message1126with a single short (less than 50 character) line summarizing the1127change, followed by a blank line and then a more thorough1128description. Tools that turn commits into email, for example, use1129the first line on the Subject line and the rest of the commit in the1130body.11311132[[ignoring-files]]1133Ignoring files1134--------------11351136A project will often generate files that you do 'not' want to track with git.1137This typically includes files generated by a build process or temporary1138backup files made by your editor. Of course, 'not' tracking files with git1139is just a matter of 'not' calling `git add` on them. But it quickly becomes1140annoying to have these untracked files lying around; e.g. they make1141`git add .` practically useless, and they keep showing up in the output of1142`git status`.11431144You can tell git to ignore certain files by creating a file called .gitignore1145in the top level of your working directory, with contents such as:11461147-------------------------------------------------1148# Lines starting with '#' are considered comments.1149# Ignore any file named foo.txt.1150foo.txt1151# Ignore (generated) html files,1152*.html1153# except foo.html which is maintained by hand.1154!foo.html1155# Ignore objects and archives.1156*.[oa]1157-------------------------------------------------11581159See linkgit:gitignore[5] for a detailed explanation of the syntax. You can1160also place .gitignore files in other directories in your working tree, and they1161will apply to those directories and their subdirectories. The `.gitignore`1162files can be added to your repository like any other files (just run `git add1163.gitignore` and `git commit`, as usual), which is convenient when the exclude1164patterns (such as patterns matching build output files) would also make sense1165for other users who clone your repository.11661167If you wish the exclude patterns to affect only certain repositories1168(instead of every repository for a given project), you may instead put1169them in a file in your repository named .git/info/exclude, or in any file1170specified by the `core.excludesfile` configuration variable. Some git1171commands can also take exclude patterns directly on the command line.1172See linkgit:gitignore[5] for the details.11731174[[how-to-merge]]1175How to merge1176------------11771178You can rejoin two diverging branches of development using1179linkgit:git-merge[1]:11801181-------------------------------------------------1182$ git merge branchname1183-------------------------------------------------11841185merges the development in the branch "branchname" into the current1186branch. If there are conflicts--for example, if the same file is1187modified in two different ways in the remote branch and the local1188branch--then you are warned; the output may look something like this:11891190-------------------------------------------------1191$ git merge next1192 100% (4/4) done1193Auto-merged file.txt1194CONFLICT (content): Merge conflict in file.txt1195Automatic merge failed; fix conflicts and then commit the result.1196-------------------------------------------------11971198Conflict markers are left in the problematic files, and after1199you resolve the conflicts manually, you can update the index1200with the contents and run git commit, as you normally would when1201creating a new file.12021203If you examine the resulting commit using gitk, you will see that it1204has two parents, one pointing to the top of the current branch, and1205one to the top of the other branch.12061207[[resolving-a-merge]]1208Resolving a merge1209-----------------12101211When a merge isn't resolved automatically, git leaves the index and1212the working tree in a special state that gives you all the1213information you need to help resolve the merge.12141215Files with conflicts are marked specially in the index, so until you1216resolve the problem and update the index, linkgit:git-commit[1] will1217fail:12181219-------------------------------------------------1220$ git commit1221file.txt: needs merge1222-------------------------------------------------12231224Also, linkgit:git-status[1] will list those files as "unmerged", and the1225files with conflicts will have conflict markers added, like this:12261227-------------------------------------------------1228<<<<<<< HEAD:file.txt1229Hello world1230=======1231Goodbye1232>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt1233-------------------------------------------------12341235All you need to do is edit the files to resolve the conflicts, and then12361237-------------------------------------------------1238$ git add file.txt1239$ git commit1240-------------------------------------------------12411242Note that the commit message will already be filled in for you with1243some information about the merge. Normally you can just use this1244default message unchanged, but you may add additional commentary of1245your own if desired.12461247The above is all you need to know to resolve a simple merge. But git1248also provides more information to help resolve conflicts:12491250[[conflict-resolution]]1251Getting conflict-resolution help during a merge1252~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~12531254All of the changes that git was able to merge automatically are1255already added to the index file, so linkgit:git-diff[1] shows only1256the conflicts. It uses an unusual syntax:12571258-------------------------------------------------1259$ git diff1260diff --cc file.txt1261index 802992c,2b60207..00000001262--- a/file.txt1263+++ b/file.txt1264@@@ -1,1 -1,1 +1,5 @@@1265++<<<<<<< HEAD:file.txt1266 +Hello world1267++=======1268+ Goodbye1269++>>>>>>> 77976da35a11db4580b80ae27e8d65caf5208086:file.txt1270-------------------------------------------------12711272Recall that the commit which will be committed after we resolve this1273conflict will have two parents instead of the usual one: one parent1274will be HEAD, the tip of the current branch; the other will be the1275tip of the other branch, which is stored temporarily in MERGE_HEAD.12761277During the merge, the index holds three versions of each file. Each of1278these three "file stages" represents a different version of the file:12791280-------------------------------------------------1281$ git show :1:file.txt # the file in a common ancestor of both branches1282$ git show :2:file.txt # the version from HEAD.1283$ git show :3:file.txt # the version from MERGE_HEAD.1284-------------------------------------------------12851286When you ask linkgit:git-diff[1] to show the conflicts, it runs a1287three-way diff between the conflicted merge results in the work tree with1288stages 2 and 3 to show only hunks whose contents come from both sides,1289mixed (in other words, when a hunk's merge results come only from stage 2,1290that part is not conflicting and is not shown. Same for stage 3).12911292The diff above shows the differences between the working-tree version of1293file.txt and the stage 2 and stage 3 versions. So instead of preceding1294each line by a single "+" or "-", it now uses two columns: the first1295column is used for differences between the first parent and the working1296directory copy, and the second for differences between the second parent1297and the working directory copy. (See the "COMBINED DIFF FORMAT" section1298of linkgit:git-diff-files[1] for a details of the format.)12991300After resolving the conflict in the obvious way (but before updating the1301index), the diff will look like:13021303-------------------------------------------------1304$ git diff1305diff --cc file.txt1306index 802992c,2b60207..00000001307--- a/file.txt1308+++ b/file.txt1309@@@ -1,1 -1,1 +1,1 @@@1310- Hello world1311 -Goodbye1312++Goodbye world1313-------------------------------------------------13141315This shows that our resolved version deleted "Hello world" from the1316first parent, deleted "Goodbye" from the second parent, and added1317"Goodbye world", which was previously absent from both.13181319Some special diff options allow diffing the working directory against1320any of these stages:13211322-------------------------------------------------1323$ git diff -1 file.txt # diff against stage 11324$ git diff --base file.txt # same as the above1325$ git diff -2 file.txt # diff against stage 21326$ git diff --ours file.txt # same as the above1327$ git diff -3 file.txt # diff against stage 31328$ git diff --theirs file.txt # same as the above.1329-------------------------------------------------13301331The linkgit:git-log[1] and linkgit:gitk[1] commands also provide special help1332for merges:13331334-------------------------------------------------1335$ git log --merge1336$ gitk --merge1337-------------------------------------------------13381339These will display all commits which exist only on HEAD or on1340MERGE_HEAD, and which touch an unmerged file.13411342You may also use linkgit:git-mergetool[1], which lets you merge the1343unmerged files using external tools such as Emacs or kdiff3.13441345Each time you resolve the conflicts in a file and update the index:13461347-------------------------------------------------1348$ git add file.txt1349-------------------------------------------------13501351the different stages of that file will be "collapsed", after which1352`git diff` will (by default) no longer show diffs for that file.13531354[[undoing-a-merge]]1355Undoing a merge1356---------------13571358If you get stuck and decide to just give up and throw the whole mess1359away, you can always return to the pre-merge state with13601361-------------------------------------------------1362$ git reset --hard HEAD1363-------------------------------------------------13641365Or, if you've already committed the merge that you want to throw away,13661367-------------------------------------------------1368$ git reset --hard ORIG_HEAD1369-------------------------------------------------13701371However, this last command can be dangerous in some cases--never1372throw away a commit you have already committed if that commit may1373itself have been merged into another branch, as doing so may confuse1374further merges.13751376[[fast-forwards]]1377Fast-forward merges1378-------------------13791380There is one special case not mentioned above, which is treated1381differently. Normally, a merge results in a merge commit, with two1382parents, one pointing at each of the two lines of development that1383were merged.13841385However, if the current branch is a descendant of the other--so every1386commit present in the one is already contained in the other--then git1387just performs a "fast forward"; the head of the current branch is moved1388forward to point at the head of the merged-in branch, without any new1389commits being created.13901391[[fixing-mistakes]]1392Fixing mistakes1393---------------13941395If you've messed up the working tree, but haven't yet committed your1396mistake, you can return the entire working tree to the last committed1397state with13981399-------------------------------------------------1400$ git reset --hard HEAD1401-------------------------------------------------14021403If you make a commit that you later wish you hadn't, there are two1404fundamentally different ways to fix the problem:14051406 1. You can create a new commit that undoes whatever was done1407 by the old commit. This is the correct thing if your1408 mistake has already been made public.14091410 2. You can go back and modify the old commit. You should1411 never do this if you have already made the history public;1412 git does not normally expect the "history" of a project to1413 change, and cannot correctly perform repeated merges from1414 a branch that has had its history changed.14151416[[reverting-a-commit]]1417Fixing a mistake with a new commit1418~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~14191420Creating a new commit that reverts an earlier change is very easy;1421just pass the linkgit:git-revert[1] command a reference to the bad1422commit; for example, to revert the most recent commit:14231424-------------------------------------------------1425$ git revert HEAD1426-------------------------------------------------14271428This will create a new commit which undoes the change in HEAD. You1429will be given a chance to edit the commit message for the new commit.14301431You can also revert an earlier change, for example, the next-to-last:14321433-------------------------------------------------1434$ git revert HEAD^1435-------------------------------------------------14361437In this case git will attempt to undo the old change while leaving1438intact any changes made since then. If more recent changes overlap1439with the changes to be reverted, then you will be asked to fix1440conflicts manually, just as in the case of <<resolving-a-merge,1441resolving a merge>>.14421443[[fixing-a-mistake-by-rewriting-history]]1444Fixing a mistake by rewriting history1445~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~14461447If the problematic commit is the most recent commit, and you have not1448yet made that commit public, then you may just1449<<undoing-a-merge,destroy it using `git reset`>>.14501451Alternatively, you1452can edit the working directory and update the index to fix your1453mistake, just as if you were going to <<how-to-make-a-commit,create a1454new commit>>, then run14551456-------------------------------------------------1457$ git commit --amend1458-------------------------------------------------14591460which will replace the old commit by a new commit incorporating your1461changes, giving you a chance to edit the old commit message first.14621463Again, you should never do this to a commit that may already have1464been merged into another branch; use linkgit:git-revert[1] instead in1465that case.14661467It is also possible to replace commits further back in the history, but1468this is an advanced topic to be left for1469<<cleaning-up-history,another chapter>>.14701471[[checkout-of-path]]1472Checking out an old version of a file1473~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~14741475In the process of undoing a previous bad change, you may find it1476useful to check out an older version of a particular file using1477linkgit:git-checkout[1]. We've used `git checkout` before to switch1478branches, but it has quite different behavior if it is given a path1479name: the command14801481-------------------------------------------------1482$ git checkout HEAD^ path/to/file1483-------------------------------------------------14841485replaces path/to/file by the contents it had in the commit HEAD^, and1486also updates the index to match. It does not change branches.14871488If you just want to look at an old version of the file, without1489modifying the working directory, you can do that with1490linkgit:git-show[1]:14911492-------------------------------------------------1493$ git show HEAD^:path/to/file1494-------------------------------------------------14951496which will display the given version of the file.14971498[[interrupted-work]]1499Temporarily setting aside work in progress1500~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~15011502While you are in the middle of working on something complicated, you1503find an unrelated but obvious and trivial bug. You would like to fix it1504before continuing. You can use linkgit:git-stash[1] to save the current1505state of your work, and after fixing the bug (or, optionally after doing1506so on a different branch and then coming back), unstash the1507work-in-progress changes.15081509------------------------------------------------1510$ git stash save "work in progress for foo feature"1511------------------------------------------------15121513This command will save your changes away to the `stash`, and1514reset your working tree and the index to match the tip of your1515current branch. Then you can make your fix as usual.15161517------------------------------------------------1518... edit and test ...1519$ git commit -a -m "blorpl: typofix"1520------------------------------------------------15211522After that, you can go back to what you were working on with1523`git stash apply`:15241525------------------------------------------------1526$ git stash apply1527------------------------------------------------152815291530[[ensuring-good-performance]]1531Ensuring good performance1532-------------------------15331534On large repositories, git depends on compression to keep the history1535information from taking up too much space on disk or in memory.15361537This compression is not performed automatically. Therefore you1538should occasionally run linkgit:git-gc[1]:15391540-------------------------------------------------1541$ git gc1542-------------------------------------------------15431544to recompress the archive. This can be very time-consuming, so1545you may prefer to run `git gc` when you are not doing other work.154615471548[[ensuring-reliability]]1549Ensuring reliability1550--------------------15511552[[checking-for-corruption]]1553Checking the repository for corruption1554~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~15551556The linkgit:git-fsck[1] command runs a number of self-consistency checks1557on the repository, and reports on any problems. This may take some1558time. The most common warning by far is about "dangling" objects:15591560-------------------------------------------------1561$ git fsck1562dangling commit 7281251ddd2a61e38657c827739c57015671a6b31563dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a631564dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b51565dangling blob 218761f9d90712d37a9c5e36f406f92202db07eb1566dangling commit bf093535a34a4d35731aa2bd90fe6b176302f14f1567dangling commit 8e4bec7f2ddaa268bef999853c25755452100f8e1568dangling tree d50bb86186bf27b681d25af89d3b5b68382e40851569dangling tree b24c2473f1fd3d91352a624795be026d64c8841f1570...1571-------------------------------------------------15721573Dangling objects are not a problem. At worst they may take up a little1574extra disk space. They can sometimes provide a last-resort method for1575recovering lost work--see <<dangling-objects>> for details.15761577[[recovering-lost-changes]]1578Recovering lost changes1579~~~~~~~~~~~~~~~~~~~~~~~15801581[[reflogs]]1582Reflogs1583^^^^^^^15841585Say you modify a branch with `linkgit:git-reset[1] --hard`, and then1586realize that the branch was the only reference you had to that point in1587history.15881589Fortunately, git also keeps a log, called a "reflog", of all the1590previous values of each branch. So in this case you can still find the1591old history using, for example,15921593-------------------------------------------------1594$ git log master@{1}1595-------------------------------------------------15961597This lists the commits reachable from the previous version of the1598"master" branch head. This syntax can be used with any git command1599that accepts a commit, not just with git log. Some other examples:16001601-------------------------------------------------1602$ git show master@{2} # See where the branch pointed 2,1603$ git show master@{3} # 3, ... changes ago.1604$ gitk master@{yesterday} # See where it pointed yesterday,1605$ gitk master@{"1 week ago"} # ... or last week1606$ git log --walk-reflogs master # show reflog entries for master1607-------------------------------------------------16081609A separate reflog is kept for the HEAD, so16101611-------------------------------------------------1612$ git show HEAD@{"1 week ago"}1613-------------------------------------------------16141615will show what HEAD pointed to one week ago, not what the current branch1616pointed to one week ago. This allows you to see the history of what1617you've checked out.16181619The reflogs are kept by default for 30 days, after which they may be1620pruned. See linkgit:git-reflog[1] and linkgit:git-gc[1] to learn1621how to control this pruning, and see the "SPECIFYING REVISIONS"1622section of linkgit:git-rev-parse[1] for details.16231624Note that the reflog history is very different from normal git history.1625While normal history is shared by every repository that works on the1626same project, the reflog history is not shared: it tells you only about1627how the branches in your local repository have changed over time.16281629[[dangling-object-recovery]]1630Examining dangling objects1631^^^^^^^^^^^^^^^^^^^^^^^^^^16321633In some situations the reflog may not be able to save you. For example,1634suppose you delete a branch, then realize you need the history it1635contained. The reflog is also deleted; however, if you have not yet1636pruned the repository, then you may still be able to find the lost1637commits in the dangling objects that `git fsck` reports. See1638<<dangling-objects>> for the details.16391640-------------------------------------------------1641$ git fsck1642dangling commit 7281251ddd2a61e38657c827739c57015671a6b31643dangling commit 2706a059f258c6b245f298dc4ff2ccd30ec21a631644dangling commit 13472b7c4b80851a1bc551779171dcb03655e9b51645...1646-------------------------------------------------16471648You can examine1649one of those dangling commits with, for example,16501651------------------------------------------------1652$ gitk 7281251ddd --not --all1653------------------------------------------------16541655which does what it sounds like: it says that you want to see the commit1656history that is described by the dangling commit(s), but not the1657history that is described by all your existing branches and tags. Thus1658you get exactly the history reachable from that commit that is lost.1659(And notice that it might not be just one commit: we only report the1660"tip of the line" as being dangling, but there might be a whole deep1661and complex commit history that was dropped.)16621663If you decide you want the history back, you can always create a new1664reference pointing to it, for example, a new branch:16651666------------------------------------------------1667$ git branch recovered-branch 7281251ddd1668------------------------------------------------16691670Other types of dangling objects (blobs and trees) are also possible, and1671dangling objects can arise in other situations.167216731674[[sharing-development]]1675Sharing development with others1676===============================16771678[[getting-updates-With-git-pull]]1679Getting updates with git pull1680-----------------------------16811682After you clone a repository and make a few changes of your own, you1683may wish to check the original repository for updates and merge them1684into your own work.16851686We have already seen <<Updating-a-repository-With-git-fetch,how to1687keep remote tracking branches up to date>> with linkgit:git-fetch[1],1688and how to merge two branches. So you can merge in changes from the1689original repository's master branch with:16901691-------------------------------------------------1692$ git fetch1693$ git merge origin/master1694-------------------------------------------------16951696However, the linkgit:git-pull[1] command provides a way to do this in1697one step:16981699-------------------------------------------------1700$ git pull origin master1701-------------------------------------------------17021703In fact, if you have "master" checked out, then by default "git pull"1704merges from the HEAD branch of the origin repository. So often you can1705accomplish the above with just a simple17061707-------------------------------------------------1708$ git pull1709-------------------------------------------------17101711More generally, a branch that is created from a remote branch will pull1712by default from that branch. See the descriptions of the1713branch.<name>.remote and branch.<name>.merge options in1714linkgit:git-config[1], and the discussion of the `--track` option in1715linkgit:git-checkout[1], to learn how to control these defaults.17161717In addition to saving you keystrokes, "git pull" also helps you by1718producing a default commit message documenting the branch and1719repository that you pulled from.17201721(But note that no such commit will be created in the case of a1722<<fast-forwards,fast forward>>; instead, your branch will just be1723updated to point to the latest commit from the upstream branch.)17241725The `git pull` command can also be given "." as the "remote" repository,1726in which case it just merges in a branch from the current repository; so1727the commands17281729-------------------------------------------------1730$ git pull . branch1731$ git merge branch1732-------------------------------------------------17331734are roughly equivalent. The former is actually very commonly used.17351736[[submitting-patches]]1737Submitting patches to a project1738-------------------------------17391740If you just have a few changes, the simplest way to submit them may1741just be to send them as patches in email:17421743First, use linkgit:git-format-patch[1]; for example:17441745-------------------------------------------------1746$ git format-patch origin1747-------------------------------------------------17481749will produce a numbered series of files in the current directory, one1750for each patch in the current branch but not in origin/HEAD.17511752You can then import these into your mail client and send them by1753hand. However, if you have a lot to send at once, you may prefer to1754use the linkgit:git-send-email[1] script to automate the process.1755Consult the mailing list for your project first to determine how they1756prefer such patches be handled.17571758[[importing-patches]]1759Importing patches to a project1760------------------------------17611762Git also provides a tool called linkgit:git-am[1] (am stands for1763"apply mailbox"), for importing such an emailed series of patches.1764Just save all of the patch-containing messages, in order, into a1765single mailbox file, say "patches.mbox", then run17661767-------------------------------------------------1768$ git am -3 patches.mbox1769-------------------------------------------------17701771Git will apply each patch in order; if any conflicts are found, it1772will stop, and you can fix the conflicts as described in1773"<<resolving-a-merge,Resolving a merge>>". (The "-3" option tells1774git to perform a merge; if you would prefer it just to abort and1775leave your tree and index untouched, you may omit that option.)17761777Once the index is updated with the results of the conflict1778resolution, instead of creating a new commit, just run17791780-------------------------------------------------1781$ git am --resolved1782-------------------------------------------------17831784and git will create the commit for you and continue applying the1785remaining patches from the mailbox.17861787The final result will be a series of commits, one for each patch in1788the original mailbox, with authorship and commit log message each1789taken from the message containing each patch.17901791[[public-repositories]]1792Public git repositories1793-----------------------17941795Another way to submit changes to a project is to tell the maintainer1796of that project to pull the changes from your repository using1797linkgit:git-pull[1]. In the section "<<getting-updates-With-git-pull,1798Getting updates with `git pull`>>" we described this as a way to get1799updates from the "main" repository, but it works just as well in the1800other direction.18011802If you and the maintainer both have accounts on the same machine, then1803you can just pull changes from each other's repositories directly;1804commands that accept repository URLs as arguments will also accept a1805local directory name:18061807-------------------------------------------------1808$ git clone /path/to/repository1809$ git pull /path/to/other/repository1810-------------------------------------------------18111812or an ssh URL:18131814-------------------------------------------------1815$ git clone ssh://yourhost/~you/repository1816-------------------------------------------------18171818For projects with few developers, or for synchronizing a few private1819repositories, this may be all you need.18201821However, the more common way to do this is to maintain a separate public1822repository (usually on a different host) for others to pull changes1823from. This is usually more convenient, and allows you to cleanly1824separate private work in progress from publicly visible work.18251826You will continue to do your day-to-day work in your personal1827repository, but periodically "push" changes from your personal1828repository into your public repository, allowing other developers to1829pull from that repository. So the flow of changes, in a situation1830where there is one other developer with a public repository, looks1831like this:18321833 you push1834 your personal repo ------------------> your public repo1835 ^ |1836 | |1837 | you pull | they pull1838 | |1839 | |1840 | they push V1841 their public repo <------------------- their repo18421843We explain how to do this in the following sections.18441845[[setting-up-a-public-repository]]1846Setting up a public repository1847~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~18481849Assume your personal repository is in the directory ~/proj. We1850first create a new clone of the repository and tell `git daemon` that it1851is meant to be public:18521853-------------------------------------------------1854$ git clone --bare ~/proj proj.git1855$ touch proj.git/git-daemon-export-ok1856-------------------------------------------------18571858The resulting directory proj.git contains a "bare" git repository--it is1859just the contents of the ".git" directory, without any files checked out1860around it.18611862Next, copy proj.git to the server where you plan to host the1863public repository. You can use scp, rsync, or whatever is most1864convenient.18651866[[exporting-via-git]]1867Exporting a git repository via the git protocol1868~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~18691870This is the preferred method.18711872If someone else administers the server, they should tell you what1873directory to put the repository in, and what git:// URL it will appear1874at. You can then skip to the section1875"<<pushing-changes-to-a-public-repository,Pushing changes to a public1876repository>>", below.18771878Otherwise, all you need to do is start linkgit:git-daemon[1]; it will1879listen on port 9418. By default, it will allow access to any directory1880that looks like a git directory and contains the magic file1881git-daemon-export-ok. Passing some directory paths as `git daemon`1882arguments will further restrict the exports to those paths.18831884You can also run `git daemon` as an inetd service; see the1885linkgit:git-daemon[1] man page for details. (See especially the1886examples section.)18871888[[exporting-via-http]]1889Exporting a git repository via http1890~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~18911892The git protocol gives better performance and reliability, but on a1893host with a web server set up, http exports may be simpler to set up.18941895All you need to do is place the newly created bare git repository in1896a directory that is exported by the web server, and make some1897adjustments to give web clients some extra information they need:18981899-------------------------------------------------1900$ mv proj.git /home/you/public_html/proj.git1901$ cd proj.git1902$ git --bare update-server-info1903$ mv hooks/post-update.sample hooks/post-update1904-------------------------------------------------19051906(For an explanation of the last two lines, see1907linkgit:git-update-server-info[1] and linkgit:githooks[5].)19081909Advertise the URL of proj.git. Anybody else should then be able to1910clone or pull from that URL, for example with a command line like:19111912-------------------------------------------------1913$ git clone http://yourserver.com/~you/proj.git1914-------------------------------------------------19151916(See also1917link:howto/setup-git-server-over-http.txt[setup-git-server-over-http]1918for a slightly more sophisticated setup using WebDAV which also1919allows pushing over http.)19201921[[pushing-changes-to-a-public-repository]]1922Pushing changes to a public repository1923~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~19241925Note that the two techniques outlined above (exporting via1926<<exporting-via-http,http>> or <<exporting-via-git,git>>) allow other1927maintainers to fetch your latest changes, but they do not allow write1928access, which you will need to update the public repository with the1929latest changes created in your private repository.19301931The simplest way to do this is using linkgit:git-push[1] and ssh; to1932update the remote branch named "master" with the latest state of your1933branch named "master", run19341935-------------------------------------------------1936$ git push ssh://yourserver.com/~you/proj.git master:master1937-------------------------------------------------19381939or just19401941-------------------------------------------------1942$ git push ssh://yourserver.com/~you/proj.git master1943-------------------------------------------------19441945As with `git fetch`, `git push` will complain if this does not result in a1946<<fast-forwards,fast forward>>; see the following section for details on1947handling this case.19481949Note that the target of a "push" is normally a1950<<def_bare_repository,bare>> repository. You can also push to a1951repository that has a checked-out working tree, but the working tree1952will not be updated by the push. This may lead to unexpected results if1953the branch you push to is the currently checked-out branch!19541955As with `git fetch`, you may also set up configuration options to1956save typing; so, for example, after19571958-------------------------------------------------1959$ cat >>.git/config <<EOF1960[remote "public-repo"]1961 url = ssh://yourserver.com/~you/proj.git1962EOF1963-------------------------------------------------19641965you should be able to perform the above push with just19661967-------------------------------------------------1968$ git push public-repo master1969-------------------------------------------------19701971See the explanations of the remote.<name>.url, branch.<name>.remote,1972and remote.<name>.push options in linkgit:git-config[1] for1973details.19741975[[forcing-push]]1976What to do when a push fails1977~~~~~~~~~~~~~~~~~~~~~~~~~~~~19781979If a push would not result in a <<fast-forwards,fast forward>> of the1980remote branch, then it will fail with an error like:19811982-------------------------------------------------1983error: remote 'refs/heads/master' is not an ancestor of1984 local 'refs/heads/master'.1985 Maybe you are not up-to-date and need to pull first?1986error: failed to push to 'ssh://yourserver.com/~you/proj.git'1987-------------------------------------------------19881989This can happen, for example, if you:19901991 - use `git reset --hard` to remove already-published commits, or1992 - use `git commit --amend` to replace already-published commits1993 (as in <<fixing-a-mistake-by-rewriting-history>>), or1994 - use `git rebase` to rebase any already-published commits (as1995 in <<using-git-rebase>>).19961997You may force `git push` to perform the update anyway by preceding the1998branch name with a plus sign:19992000-------------------------------------------------2001$ git push ssh://yourserver.com/~you/proj.git +master2002-------------------------------------------------20032004Normally whenever a branch head in a public repository is modified, it2005is modified to point to a descendant of the commit that it pointed to2006before. By forcing a push in this situation, you break that convention.2007(See <<problems-With-rewriting-history>>.)20082009Nevertheless, this is a common practice for people that need a simple2010way to publish a work-in-progress patch series, and it is an acceptable2011compromise as long as you warn other developers that this is how you2012intend to manage the branch.20132014It's also possible for a push to fail in this way when other people have2015the right to push to the same repository. In that case, the correct2016solution is to retry the push after first updating your work: either by a2017pull, or by a fetch followed by a rebase; see the2018<<setting-up-a-shared-repository,next section>> and2019linkgit:gitcvs-migration[7] for more.20202021[[setting-up-a-shared-repository]]2022Setting up a shared repository2023~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~20242025Another way to collaborate is by using a model similar to that2026commonly used in CVS, where several developers with special rights2027all push to and pull from a single shared repository. See2028linkgit:gitcvs-migration[7] for instructions on how to2029set this up.20302031However, while there is nothing wrong with git's support for shared2032repositories, this mode of operation is not generally recommended,2033simply because the mode of collaboration that git supports--by2034exchanging patches and pulling from public repositories--has so many2035advantages over the central shared repository:20362037 - Git's ability to quickly import and merge patches allows a2038 single maintainer to process incoming changes even at very2039 high rates. And when that becomes too much, `git pull` provides2040 an easy way for that maintainer to delegate this job to other2041 maintainers while still allowing optional review of incoming2042 changes.2043 - Since every developer's repository has the same complete copy2044 of the project history, no repository is special, and it is2045 trivial for another developer to take over maintenance of a2046 project, either by mutual agreement, or because a maintainer2047 becomes unresponsive or difficult to work with.2048 - The lack of a central group of "committers" means there is2049 less need for formal decisions about who is "in" and who is2050 "out".20512052[[setting-up-gitweb]]2053Allowing web browsing of a repository2054~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~20552056The gitweb cgi script provides users an easy way to browse your2057project's files and history without having to install git; see the file2058gitweb/INSTALL in the git source tree for instructions on setting it up.20592060[[sharing-development-examples]]2061Examples2062--------20632064[[maintaining-topic-branches]]2065Maintaining topic branches for a Linux subsystem maintainer2066~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~20672068This describes how Tony Luck uses git in his role as maintainer of the2069IA64 architecture for the Linux kernel.20702071He uses two public branches:20722073 - A "test" tree into which patches are initially placed so that they2074 can get some exposure when integrated with other ongoing development.2075 This tree is available to Andrew for pulling into -mm whenever he2076 wants.20772078 - A "release" tree into which tested patches are moved for final sanity2079 checking, and as a vehicle to send them upstream to Linus (by sending2080 him a "please pull" request.)20812082He also uses a set of temporary branches ("topic branches"), each2083containing a logical grouping of patches.20842085To set this up, first create your work tree by cloning Linus's public2086tree:20872088-------------------------------------------------2089$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git work2090$ cd work2091-------------------------------------------------20922093Linus's tree will be stored in the remote branch named origin/master,2094and can be updated using linkgit:git-fetch[1]; you can track other2095public trees using linkgit:git-remote[1] to set up a "remote" and2096linkgit:git-fetch[1] to keep them up-to-date; see2097<<repositories-and-branches>>.20982099Now create the branches in which you are going to work; these start out2100at the current tip of origin/master branch, and should be set up (using2101the --track option to linkgit:git-branch[1]) to merge changes in from2102Linus by default.21032104-------------------------------------------------2105$ git branch --track test origin/master2106$ git branch --track release origin/master2107-------------------------------------------------21082109These can be easily kept up to date using linkgit:git-pull[1].21102111-------------------------------------------------2112$ git checkout test && git pull2113$ git checkout release && git pull2114-------------------------------------------------21152116Important note! If you have any local changes in these branches, then2117this merge will create a commit object in the history (with no local2118changes git will simply do a "Fast forward" merge). Many people dislike2119the "noise" that this creates in the Linux history, so you should avoid2120doing this capriciously in the "release" branch, as these noisy commits2121will become part of the permanent history when you ask Linus to pull2122from the release branch.21232124A few configuration variables (see linkgit:git-config[1]) can2125make it easy to push both branches to your public tree. (See2126<<setting-up-a-public-repository>>.)21272128-------------------------------------------------2129$ cat >> .git/config <<EOF2130[remote "mytree"]2131 url = master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6.git2132 push = release2133 push = test2134EOF2135-------------------------------------------------21362137Then you can push both the test and release trees using2138linkgit:git-push[1]:21392140-------------------------------------------------2141$ git push mytree2142-------------------------------------------------21432144or push just one of the test and release branches using:21452146-------------------------------------------------2147$ git push mytree test2148-------------------------------------------------21492150or21512152-------------------------------------------------2153$ git push mytree release2154-------------------------------------------------21552156Now to apply some patches from the community. Think of a short2157snappy name for a branch to hold this patch (or related group of2158patches), and create a new branch from the current tip of Linus's2159branch:21602161-------------------------------------------------2162$ git checkout -b speed-up-spinlocks origin2163-------------------------------------------------21642165Now you apply the patch(es), run some tests, and commit the change(s). If2166the patch is a multi-part series, then you should apply each as a separate2167commit to this branch.21682169-------------------------------------------------2170$ ... patch ... test ... commit [ ... patch ... test ... commit ]*2171-------------------------------------------------21722173When you are happy with the state of this change, you can pull it into the2174"test" branch in preparation to make it public:21752176-------------------------------------------------2177$ git checkout test && git pull . speed-up-spinlocks2178-------------------------------------------------21792180It is unlikely that you would have any conflicts here ... but you might if you2181spent a while on this step and had also pulled new versions from upstream.21822183Some time later when enough time has passed and testing done, you can pull the2184same branch into the "release" tree ready to go upstream. This is where you2185see the value of keeping each patch (or patch series) in its own branch. It2186means that the patches can be moved into the "release" tree in any order.21872188-------------------------------------------------2189$ git checkout release && git pull . speed-up-spinlocks2190-------------------------------------------------21912192After a while, you will have a number of branches, and despite the2193well chosen names you picked for each of them, you may forget what2194they are for, or what status they are in. To get a reminder of what2195changes are in a specific branch, use:21962197-------------------------------------------------2198$ git log linux..branchname | git shortlog2199-------------------------------------------------22002201To see whether it has already been merged into the test or release branches,2202use:22032204-------------------------------------------------2205$ git log test..branchname2206-------------------------------------------------22072208or22092210-------------------------------------------------2211$ git log release..branchname2212-------------------------------------------------22132214(If this branch has not yet been merged, you will see some log entries.2215If it has been merged, then there will be no output.)22162217Once a patch completes the great cycle (moving from test to release,2218then pulled by Linus, and finally coming back into your local2219"origin/master" branch), the branch for this change is no longer needed.2220You detect this when the output from:22212222-------------------------------------------------2223$ git log origin..branchname2224-------------------------------------------------22252226is empty. At this point the branch can be deleted:22272228-------------------------------------------------2229$ git branch -d branchname2230-------------------------------------------------22312232Some changes are so trivial that it is not necessary to create a separate2233branch and then merge into each of the test and release branches. For2234these changes, just apply directly to the "release" branch, and then2235merge that into the "test" branch.22362237To create diffstat and shortlog summaries of changes to include in a "please2238pull" request to Linus you can use:22392240-------------------------------------------------2241$ git diff --stat origin..release2242-------------------------------------------------22432244and22452246-------------------------------------------------2247$ git log -p origin..release | git shortlog2248-------------------------------------------------22492250Here are some of the scripts that simplify all this even further.22512252-------------------------------------------------2253==== update script ====2254# Update a branch in my GIT tree. If the branch to be updated2255# is origin, then pull from kernel.org. Otherwise merge2256# origin/master branch into test|release branch22572258case "$1" in2259test|release)2260 git checkout $1 && git pull . origin2261 ;;2262origin)2263 before=$(git rev-parse refs/remotes/origin/master)2264 git fetch origin2265 after=$(git rev-parse refs/remotes/origin/master)2266 if [ $before != $after ]2267 then2268 git log $before..$after | git shortlog2269 fi2270 ;;2271*)2272 echo "Usage: $0 origin|test|release" 1>&22273 exit 12274 ;;2275esac2276-------------------------------------------------22772278-------------------------------------------------2279==== merge script ====2280# Merge a branch into either the test or release branch22812282pname=$022832284usage()2285{2286 echo "Usage: $pname branch test|release" 1>&22287 exit 12288}22892290git show-ref -q --verify -- refs/heads/"$1" || {2291 echo "Can't see branch <$1>" 1>&22292 usage2293}22942295case "$2" in2296test|release)2297 if [ $(git log $2..$1 | wc -c) -eq 0 ]2298 then2299 echo $1 already merged into $2 1>&22300 exit 12301 fi2302 git checkout $2 && git pull . $12303 ;;2304*)2305 usage2306 ;;2307esac2308-------------------------------------------------23092310-------------------------------------------------2311==== status script ====2312# report on status of my ia64 GIT tree23132314gb=$(tput setab 2)2315rb=$(tput setab 1)2316restore=$(tput setab 9)23172318if [ `git rev-list test..release | wc -c` -gt 0 ]2319then2320 echo $rb Warning: commits in release that are not in test $restore2321 git log test..release2322fi23232324for branch in `git show-ref --heads | sed 's|^.*/||'`2325do2326 if [ $branch = test -o $branch = release ]2327 then2328 continue2329 fi23302331 echo -n $gb ======= $branch ====== $restore " "2332 status=2333 for ref in test release origin/master2334 do2335 if [ `git rev-list $ref..$branch | wc -c` -gt 0 ]2336 then2337 status=$status${ref:0:1}2338 fi2339 done2340 case $status in2341 trl)2342 echo $rb Need to pull into test $restore2343 ;;2344 rl)2345 echo "In test"2346 ;;2347 l)2348 echo "Waiting for linus"2349 ;;2350 "")2351 echo $rb All done $restore2352 ;;2353 *)2354 echo $rb "<$status>" $restore2355 ;;2356 esac2357 git log origin/master..$branch | git shortlog2358done2359-------------------------------------------------236023612362[[cleaning-up-history]]2363Rewriting history and maintaining patch series2364==============================================23652366Normally commits are only added to a project, never taken away or2367replaced. Git is designed with this assumption, and violating it will2368cause git's merge machinery (for example) to do the wrong thing.23692370However, there is a situation in which it can be useful to violate this2371assumption.23722373[[patch-series]]2374Creating the perfect patch series2375---------------------------------23762377Suppose you are a contributor to a large project, and you want to add a2378complicated feature, and to present it to the other developers in a way2379that makes it easy for them to read your changes, verify that they are2380correct, and understand why you made each change.23812382If you present all of your changes as a single patch (or commit), they2383may find that it is too much to digest all at once.23842385If you present them with the entire history of your work, complete with2386mistakes, corrections, and dead ends, they may be overwhelmed.23872388So the ideal is usually to produce a series of patches such that:23892390 1. Each patch can be applied in order.23912392 2. Each patch includes a single logical change, together with a2393 message explaining the change.23942395 3. No patch introduces a regression: after applying any initial2396 part of the series, the resulting project still compiles and2397 works, and has no bugs that it didn't have before.23982399 4. The complete series produces the same end result as your own2400 (probably much messier!) development process did.24012402We will introduce some tools that can help you do this, explain how to2403use them, and then explain some of the problems that can arise because2404you are rewriting history.24052406[[using-git-rebase]]2407Keeping a patch series up to date using git rebase2408--------------------------------------------------24092410Suppose that you create a branch "mywork" on a remote-tracking branch2411"origin", and create some commits on top of it:24122413-------------------------------------------------2414$ git checkout -b mywork origin2415$ vi file.txt2416$ git commit2417$ vi otherfile.txt2418$ git commit2419...2420-------------------------------------------------24212422You have performed no merges into mywork, so it is just a simple linear2423sequence of patches on top of "origin":24242425................................................2426 o--o--o <-- origin2427 \2428 o--o--o <-- mywork2429................................................24302431Some more interesting work has been done in the upstream project, and2432"origin" has advanced:24332434................................................2435 o--o--O--o--o--o <-- origin2436 \2437 a--b--c <-- mywork2438................................................24392440At this point, you could use "pull" to merge your changes back in;2441the result would create a new merge commit, like this:24422443................................................2444 o--o--O--o--o--o <-- origin2445 \ \2446 a--b--c--m <-- mywork2447................................................24482449However, if you prefer to keep the history in mywork a simple series of2450commits without any merges, you may instead choose to use2451linkgit:git-rebase[1]:24522453-------------------------------------------------2454$ git checkout mywork2455$ git rebase origin2456-------------------------------------------------24572458This will remove each of your commits from mywork, temporarily saving2459them as patches (in a directory named ".git/rebase-apply"), update mywork to2460point at the latest version of origin, then apply each of the saved2461patches to the new mywork. The result will look like:246224632464................................................2465 o--o--O--o--o--o <-- origin2466 \2467 a'--b'--c' <-- mywork2468................................................24692470In the process, it may discover conflicts. In that case it will stop2471and allow you to fix the conflicts; after fixing conflicts, use `git add`2472to update the index with those contents, and then, instead of2473running `git commit`, just run24742475-------------------------------------------------2476$ git rebase --continue2477-------------------------------------------------24782479and git will continue applying the rest of the patches.24802481At any point you may use the `--abort` option to abort this process and2482return mywork to the state it had before you started the rebase:24832484-------------------------------------------------2485$ git rebase --abort2486-------------------------------------------------24872488[[rewriting-one-commit]]2489Rewriting a single commit2490-------------------------24912492We saw in <<fixing-a-mistake-by-rewriting-history>> that you can replace the2493most recent commit using24942495-------------------------------------------------2496$ git commit --amend2497-------------------------------------------------24982499which will replace the old commit by a new commit incorporating your2500changes, giving you a chance to edit the old commit message first.25012502You can also use a combination of this and linkgit:git-rebase[1] to2503replace a commit further back in your history and recreate the2504intervening changes on top of it. First, tag the problematic commit2505with25062507-------------------------------------------------2508$ git tag bad mywork~52509-------------------------------------------------25102511(Either gitk or `git log` may be useful for finding the commit.)25122513Then check out that commit, edit it, and rebase the rest of the series2514on top of it (note that we could check out the commit on a temporary2515branch, but instead we're using a <<detached-head,detached head>>):25162517-------------------------------------------------2518$ git checkout bad2519$ # make changes here and update the index2520$ git commit --amend2521$ git rebase --onto HEAD bad mywork2522-------------------------------------------------25232524When you're done, you'll be left with mywork checked out, with the top2525patches on mywork reapplied on top of your modified commit. You can2526then clean up with25272528-------------------------------------------------2529$ git tag -d bad2530-------------------------------------------------25312532Note that the immutable nature of git history means that you haven't really2533"modified" existing commits; instead, you have replaced the old commits with2534new commits having new object names.25352536[[reordering-patch-series]]2537Reordering or selecting from a patch series2538-------------------------------------------25392540Given one existing commit, the linkgit:git-cherry-pick[1] command2541allows you to apply the change introduced by that commit and create a2542new commit that records it. So, for example, if "mywork" points to a2543series of patches on top of "origin", you might do something like:25442545-------------------------------------------------2546$ git checkout -b mywork-new origin2547$ gitk origin..mywork &2548-------------------------------------------------25492550and browse through the list of patches in the mywork branch using gitk,2551applying them (possibly in a different order) to mywork-new using2552cherry-pick, and possibly modifying them as you go using `git commit --amend`.2553The linkgit:git-gui[1] command may also help as it allows you to2554individually select diff hunks for inclusion in the index (by2555right-clicking on the diff hunk and choosing "Stage Hunk for Commit").25562557Another technique is to use `git format-patch` to create a series of2558patches, then reset the state to before the patches:25592560-------------------------------------------------2561$ git format-patch origin2562$ git reset --hard origin2563-------------------------------------------------25642565Then modify, reorder, or eliminate patches as preferred before applying2566them again with linkgit:git-am[1].25672568[[patch-series-tools]]2569Other tools2570-----------25712572There are numerous other tools, such as StGIT, which exist for the2573purpose of maintaining a patch series. These are outside of the scope of2574this manual.25752576[[problems-With-rewriting-history]]2577Problems with rewriting history2578-------------------------------25792580The primary problem with rewriting the history of a branch has to do2581with merging. Suppose somebody fetches your branch and merges it into2582their branch, with a result something like this:25832584................................................2585 o--o--O--o--o--o <-- origin2586 \ \2587 t--t--t--m <-- their branch:2588................................................25892590Then suppose you modify the last three commits:25912592................................................2593 o--o--o <-- new head of origin2594 /2595 o--o--O--o--o--o <-- old head of origin2596................................................25972598If we examined all this history together in one repository, it will2599look like:26002601................................................2602 o--o--o <-- new head of origin2603 /2604 o--o--O--o--o--o <-- old head of origin2605 \ \2606 t--t--t--m <-- their branch:2607................................................26082609Git has no way of knowing that the new head is an updated version of2610the old head; it treats this situation exactly the same as it would if2611two developers had independently done the work on the old and new heads2612in parallel. At this point, if someone attempts to merge the new head2613in to their branch, git will attempt to merge together the two (old and2614new) lines of development, instead of trying to replace the old by the2615new. The results are likely to be unexpected.26162617You may still choose to publish branches whose history is rewritten,2618and it may be useful for others to be able to fetch those branches in2619order to examine or test them, but they should not attempt to pull such2620branches into their own work.26212622For true distributed development that supports proper merging,2623published branches should never be rewritten.26242625[[bisect-merges]]2626Why bisecting merge commits can be harder than bisecting linear history2627-----------------------------------------------------------------------26282629The linkgit:git-bisect[1] command correctly handles history that2630includes merge commits. However, when the commit that it finds is a2631merge commit, the user may need to work harder than usual to figure out2632why that commit introduced a problem.26332634Imagine this history:26352636................................................2637 ---Z---o---X---...---o---A---C---D2638 \ /2639 o---o---Y---...---o---B2640................................................26412642Suppose that on the upper line of development, the meaning of one2643of the functions that exists at Z is changed at commit X. The2644commits from Z leading to A change both the function's2645implementation and all calling sites that exist at Z, as well2646as new calling sites they add, to be consistent. There is no2647bug at A.26482649Suppose that in the meantime on the lower line of development somebody2650adds a new calling site for that function at commit Y. The2651commits from Z leading to B all assume the old semantics of that2652function and the callers and the callee are consistent with each2653other. There is no bug at B, either.26542655Suppose further that the two development lines merge cleanly at C,2656so no conflict resolution is required.26572658Nevertheless, the code at C is broken, because the callers added2659on the lower line of development have not been converted to the new2660semantics introduced on the upper line of development. So if all2661you know is that D is bad, that Z is good, and that2662linkgit:git-bisect[1] identifies C as the culprit, how will you2663figure out that the problem is due to this change in semantics?26642665When the result of a `git bisect` is a non-merge commit, you should2666normally be able to discover the problem by examining just that commit.2667Developers can make this easy by breaking their changes into small2668self-contained commits. That won't help in the case above, however,2669because the problem isn't obvious from examination of any single2670commit; instead, a global view of the development is required. To2671make matters worse, the change in semantics in the problematic2672function may be just one small part of the changes in the upper2673line of development.26742675On the other hand, if instead of merging at C you had rebased the2676history between Z to B on top of A, you would have gotten this2677linear history:26782679................................................................2680 ---Z---o---X--...---o---A---o---o---Y*--...---o---B*--D*2681................................................................26822683Bisecting between Z and D* would hit a single culprit commit Y*,2684and understanding why Y* was broken would probably be easier.26852686Partly for this reason, many experienced git users, even when2687working on an otherwise merge-heavy project, keep the history2688linear by rebasing against the latest upstream version before2689publishing.26902691[[advanced-branch-management]]2692Advanced branch management2693==========================26942695[[fetching-individual-branches]]2696Fetching individual branches2697----------------------------26982699Instead of using linkgit:git-remote[1], you can also choose just2700to update one branch at a time, and to store it locally under an2701arbitrary name:27022703-------------------------------------------------2704$ git fetch origin todo:my-todo-work2705-------------------------------------------------27062707The first argument, "origin", just tells git to fetch from the2708repository you originally cloned from. The second argument tells git2709to fetch the branch named "todo" from the remote repository, and to2710store it locally under the name refs/heads/my-todo-work.27112712You can also fetch branches from other repositories; so27132714-------------------------------------------------2715$ git fetch git://example.com/proj.git master:example-master2716-------------------------------------------------27172718will create a new branch named "example-master" and store in it the2719branch named "master" from the repository at the given URL. If you2720already have a branch named example-master, it will attempt to2721<<fast-forwards,fast-forward>> to the commit given by example.com's2722master branch. In more detail:27232724[[fetch-fast-forwards]]2725git fetch and fast-forwards2726---------------------------27272728In the previous example, when updating an existing branch, "git fetch"2729checks to make sure that the most recent commit on the remote2730branch is a descendant of the most recent commit on your copy of the2731branch before updating your copy of the branch to point at the new2732commit. Git calls this process a <<fast-forwards,fast forward>>.27332734A fast forward looks something like this:27352736................................................2737 o--o--o--o <-- old head of the branch2738 \2739 o--o--o <-- new head of the branch2740................................................274127422743In some cases it is possible that the new head will *not* actually be2744a descendant of the old head. For example, the developer may have2745realized she made a serious mistake, and decided to backtrack,2746resulting in a situation like:27472748................................................2749 o--o--o--o--a--b <-- old head of the branch2750 \2751 o--o--o <-- new head of the branch2752................................................27532754In this case, "git fetch" will fail, and print out a warning.27552756In that case, you can still force git to update to the new head, as2757described in the following section. However, note that in the2758situation above this may mean losing the commits labeled "a" and "b",2759unless you've already created a reference of your own pointing to2760them.27612762[[forcing-fetch]]2763Forcing git fetch to do non-fast-forward updates2764------------------------------------------------27652766If git fetch fails because the new head of a branch is not a2767descendant of the old head, you may force the update with:27682769-------------------------------------------------2770$ git fetch git://example.com/proj.git +master:refs/remotes/example/master2771-------------------------------------------------27722773Note the addition of the "+" sign. Alternatively, you can use the "-f"2774flag to force updates of all the fetched branches, as in:27752776-------------------------------------------------2777$ git fetch -f origin2778-------------------------------------------------27792780Be aware that commits that the old version of example/master pointed at2781may be lost, as we saw in the previous section.27822783[[remote-branch-configuration]]2784Configuring remote branches2785---------------------------27862787We saw above that "origin" is just a shortcut to refer to the2788repository that you originally cloned from. This information is2789stored in git configuration variables, which you can see using2790linkgit:git-config[1]:27912792-------------------------------------------------2793$ git config -l2794core.repositoryformatversion=02795core.filemode=true2796core.logallrefupdates=true2797remote.origin.url=git://git.kernel.org/pub/scm/git/git.git2798remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*2799branch.master.remote=origin2800branch.master.merge=refs/heads/master2801-------------------------------------------------28022803If there are other repositories that you also use frequently, you can2804create similar configuration options to save typing; for example,2805after28062807-------------------------------------------------2808$ git config remote.example.url git://example.com/proj.git2809-------------------------------------------------28102811then the following two commands will do the same thing:28122813-------------------------------------------------2814$ git fetch git://example.com/proj.git master:refs/remotes/example/master2815$ git fetch example master:refs/remotes/example/master2816-------------------------------------------------28172818Even better, if you add one more option:28192820-------------------------------------------------2821$ git config remote.example.fetch master:refs/remotes/example/master2822-------------------------------------------------28232824then the following commands will all do the same thing:28252826-------------------------------------------------2827$ git fetch git://example.com/proj.git master:refs/remotes/example/master2828$ git fetch example master:refs/remotes/example/master2829$ git fetch example2830-------------------------------------------------28312832You can also add a "+" to force the update each time:28332834-------------------------------------------------2835$ git config remote.example.fetch +master:ref/remotes/example/master2836-------------------------------------------------28372838Don't do this unless you're sure you won't mind "git fetch" possibly2839throwing away commits on 'example/master'.28402841Also note that all of the above configuration can be performed by2842directly editing the file .git/config instead of using2843linkgit:git-config[1].28442845See linkgit:git-config[1] for more details on the configuration2846options mentioned above.284728482849[[git-concepts]]2850Git concepts2851============28522853Git is built on a small number of simple but powerful ideas. While it2854is possible to get things done without understanding them, you will find2855git much more intuitive if you do.28562857We start with the most important, the <<def_object_database,object2858database>> and the <<def_index,index>>.28592860[[the-object-database]]2861The Object Database2862-------------------286328642865We already saw in <<understanding-commits>> that all commits are stored2866under a 40-digit "object name". In fact, all the information needed to2867represent the history of a project is stored in objects with such names.2868In each case the name is calculated by taking the SHA-1 hash of the2869contents of the object. The SHA-1 hash is a cryptographic hash function.2870What that means to us is that it is impossible to find two different2871objects with the same name. This has a number of advantages; among2872others:28732874- Git can quickly determine whether two objects are identical or not,2875 just by comparing names.2876- Since object names are computed the same way in every repository, the2877 same content stored in two repositories will always be stored under2878 the same name.2879- Git can detect errors when it reads an object, by checking that the2880 object's name is still the SHA-1 hash of its contents.28812882(See <<object-details>> for the details of the object formatting and2883SHA-1 calculation.)28842885There are four different types of objects: "blob", "tree", "commit", and2886"tag".28872888- A <<def_blob_object,"blob" object>> is used to store file data.2889- A <<def_tree_object,"tree" object>> ties one or more2890 "blob" objects into a directory structure. In addition, a tree object2891 can refer to other tree objects, thus creating a directory hierarchy.2892- A <<def_commit_object,"commit" object>> ties such directory hierarchies2893 together into a <<def_DAG,directed acyclic graph>> of revisions--each2894 commit contains the object name of exactly one tree designating the2895 directory hierarchy at the time of the commit. In addition, a commit2896 refers to "parent" commit objects that describe the history of how we2897 arrived at that directory hierarchy.2898- A <<def_tag_object,"tag" object>> symbolically identifies and can be2899 used to sign other objects. It contains the object name and type of2900 another object, a symbolic name (of course!) and, optionally, a2901 signature.29022903The object types in some more detail:29042905[[commit-object]]2906Commit Object2907~~~~~~~~~~~~~29082909The "commit" object links a physical state of a tree with a description2910of how we got there and why. Use the --pretty=raw option to2911linkgit:git-show[1] or linkgit:git-log[1] to examine your favorite2912commit:29132914------------------------------------------------2915$ git show -s --pretty=raw 2be7fcb4762916commit 2be7fcb4764f2dbcee52635b91fedb1b3dcf7ab42917tree fb3a8bdd0ceddd019615af4d57a53f43d8cee2bf2918parent 257a84d9d02e90447b149af58b271c19405edb6a2919author Dave Watson <dwatson@mimvista.com> 1187576872 -04002920committer Junio C Hamano <gitster@pobox.com> 1187591163 -070029212922 Fix misspelling of 'suppress' in docs29232924 Signed-off-by: Junio C Hamano <gitster@pobox.com>2925------------------------------------------------29262927As you can see, a commit is defined by:29282929- a tree: The SHA-1 name of a tree object (as defined below), representing2930 the contents of a directory at a certain point in time.2931- parent(s): The SHA-1 name of some number of commits which represent the2932 immediately previous step(s) in the history of the project. The2933 example above has one parent; merge commits may have more than2934 one. A commit with no parents is called a "root" commit, and2935 represents the initial revision of a project. Each project must have2936 at least one root. A project can also have multiple roots, though2937 that isn't common (or necessarily a good idea).2938- an author: The name of the person responsible for this change, together2939 with its date.2940- a committer: The name of the person who actually created the commit,2941 with the date it was done. This may be different from the author, for2942 example, if the author was someone who wrote a patch and emailed it2943 to the person who used it to create the commit.2944- a comment describing this commit.29452946Note that a commit does not itself contain any information about what2947actually changed; all changes are calculated by comparing the contents2948of the tree referred to by this commit with the trees associated with2949its parents. In particular, git does not attempt to record file renames2950explicitly, though it can identify cases where the existence of the same2951file data at changing paths suggests a rename. (See, for example, the2952-M option to linkgit:git-diff[1]).29532954A commit is usually created by linkgit:git-commit[1], which creates a2955commit whose parent is normally the current HEAD, and whose tree is2956taken from the content currently stored in the index.29572958[[tree-object]]2959Tree Object2960~~~~~~~~~~~29612962The ever-versatile linkgit:git-show[1] command can also be used to2963examine tree objects, but linkgit:git-ls-tree[1] will give you more2964details:29652966------------------------------------------------2967$ git ls-tree fb3a8bdd0ce2968100644 blob 63c918c667fa005ff12ad89437f2fdc80926e21c .gitignore2969100644 blob 5529b198e8d14decbe4ad99db3f7fb632de0439d .mailmap2970100644 blob 6ff87c4664981e4397625791c8ea3bbb5f2279a3 COPYING2971040000 tree 2fb783e477100ce076f6bf57e4a6f026013dc745 Documentation2972100755 blob 3c0032cec592a765692234f1cba47dfdcc3a9200 GIT-VERSION-GEN2973100644 blob 289b046a443c0647624607d471289b2c7dcd470b INSTALL2974100644 blob 4eb463797adc693dc168b926b6932ff53f17d0b1 Makefile2975100644 blob 548142c327a6790ff8821d67c2ee1eff7a656b52 README2976...2977------------------------------------------------29782979As you can see, a tree object contains a list of entries, each with a2980mode, object type, SHA-1 name, and name, sorted by name. It represents2981the contents of a single directory tree.29822983The object type may be a blob, representing the contents of a file, or2984another tree, representing the contents of a subdirectory. Since trees2985and blobs, like all other objects, are named by the SHA-1 hash of their2986contents, two trees have the same SHA-1 name if and only if their2987contents (including, recursively, the contents of all subdirectories)2988are identical. This allows git to quickly determine the differences2989between two related tree objects, since it can ignore any entries with2990identical object names.29912992(Note: in the presence of submodules, trees may also have commits as2993entries. See <<submodules>> for documentation.)29942995Note that the files all have mode 644 or 755: git actually only pays2996attention to the executable bit.29972998[[blob-object]]2999Blob Object3000~~~~~~~~~~~30013002You can use linkgit:git-show[1] to examine the contents of a blob; take,3003for example, the blob in the entry for "COPYING" from the tree above:30043005------------------------------------------------3006$ git show 6ff87c466430073008 Note that the only valid version of the GPL as far as this project3009 is concerned is _this_ particular version of the license (ie v2, not3010 v2.2 or v3.x or whatever), unless explicitly otherwise stated.3011...3012------------------------------------------------30133014A "blob" object is nothing but a binary blob of data. It doesn't refer3015to anything else or have attributes of any kind.30163017Since the blob is entirely defined by its data, if two files in a3018directory tree (or in multiple different versions of the repository)3019have the same contents, they will share the same blob object. The object3020is totally independent of its location in the directory tree, and3021renaming a file does not change the object that file is associated with.30223023Note that any tree or blob object can be examined using3024linkgit:git-show[1] with the <revision>:<path> syntax. This can3025sometimes be useful for browsing the contents of a tree that is not3026currently checked out.30273028[[trust]]3029Trust3030~~~~~30313032If you receive the SHA-1 name of a blob from one source, and its contents3033from another (possibly untrusted) source, you can still trust that those3034contents are correct as long as the SHA-1 name agrees. This is because3035the SHA-1 is designed so that it is infeasible to find different contents3036that produce the same hash.30373038Similarly, you need only trust the SHA-1 name of a top-level tree object3039to trust the contents of the entire directory that it refers to, and if3040you receive the SHA-1 name of a commit from a trusted source, then you3041can easily verify the entire history of commits reachable through3042parents of that commit, and all of those contents of the trees referred3043to by those commits.30443045So to introduce some real trust in the system, the only thing you need3046to do is to digitally sign just 'one' special note, which includes the3047name of a top-level commit. Your digital signature shows others3048that you trust that commit, and the immutability of the history of3049commits tells others that they can trust the whole history.30503051In other words, you can easily validate a whole archive by just3052sending out a single email that tells the people the name (SHA-1 hash)3053of the top commit, and digitally sign that email using something3054like GPG/PGP.30553056To assist in this, git also provides the tag object...30573058[[tag-object]]3059Tag Object3060~~~~~~~~~~30613062A tag object contains an object, object type, tag name, the name of the3063person ("tagger") who created the tag, and a message, which may contain3064a signature, as can be seen using linkgit:git-cat-file[1]:30653066------------------------------------------------3067$ git cat-file tag v1.5.03068object 437b1b20df4b356c9342dac8d38849f24ef44f273069type commit3070tag v1.5.03071tagger Junio C Hamano <junkio@cox.net> 1171411200 +000030723073GIT 1.5.03074-----BEGIN PGP SIGNATURE-----3075Version: GnuPG v1.4.6 (GNU/Linux)30763077iD8DBQBF0lGqwMbZpPMRm5oRAuRiAJ9ohBLd7s2kqjkKlq1qqC57SbnmzQCdG4ui3078nLE/L9aUXdWeTFPron96DLA=3079=2E+03080-----END PGP SIGNATURE-----3081------------------------------------------------30823083See the linkgit:git-tag[1] command to learn how to create and verify tag3084objects. (Note that linkgit:git-tag[1] can also be used to create3085"lightweight tags", which are not tag objects at all, but just simple3086references whose names begin with "refs/tags/").30873088[[pack-files]]3089How git stores objects efficiently: pack files3090~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~30913092Newly created objects are initially created in a file named after the3093object's SHA-1 hash (stored in .git/objects).30943095Unfortunately this system becomes inefficient once a project has a3096lot of objects. Try this on an old project:30973098------------------------------------------------3099$ git count-objects31006930 objects, 47620 kilobytes3101------------------------------------------------31023103The first number is the number of objects which are kept in3104individual files. The second is the amount of space taken up by3105those "loose" objects.31063107You can save space and make git faster by moving these loose objects in3108to a "pack file", which stores a group of objects in an efficient3109compressed format; the details of how pack files are formatted can be3110found in link:technical/pack-format.txt[technical/pack-format.txt].31113112To put the loose objects into a pack, just run git repack:31133114------------------------------------------------3115$ git repack3116Generating pack...3117Done counting 6020 objects.3118Deltifying 6020 objects.3119 100% (6020/6020) done3120Writing 6020 objects.3121 100% (6020/6020) done3122Total 6020, written 6020 (delta 4070), reused 0 (delta 0)3123Pack pack-3e54ad29d5b2e05838c75df582c65257b8d08e1c created.3124------------------------------------------------31253126You can then run31273128------------------------------------------------3129$ git prune3130------------------------------------------------31313132to remove any of the "loose" objects that are now contained in the3133pack. This will also remove any unreferenced objects (which may be3134created when, for example, you use "git reset" to remove a commit).3135You can verify that the loose objects are gone by looking at the3136.git/objects directory or by running31373138------------------------------------------------3139$ git count-objects31400 objects, 0 kilobytes3141------------------------------------------------31423143Although the object files are gone, any commands that refer to those3144objects will work exactly as they did before.31453146The linkgit:git-gc[1] command performs packing, pruning, and more for3147you, so is normally the only high-level command you need.31483149[[dangling-objects]]3150Dangling objects3151~~~~~~~~~~~~~~~~31523153The linkgit:git-fsck[1] command will sometimes complain about dangling3154objects. They are not a problem.31553156The most common cause of dangling objects is that you've rebased a3157branch, or you have pulled from somebody else who rebased a branch--see3158<<cleaning-up-history>>. In that case, the old head of the original3159branch still exists, as does everything it pointed to. The branch3160pointer itself just doesn't, since you replaced it with another one.31613162There are also other situations that cause dangling objects. For3163example, a "dangling blob" may arise because you did a "git add" of a3164file, but then, before you actually committed it and made it part of the3165bigger picture, you changed something else in that file and committed3166that *updated* thing--the old state that you added originally ends up3167not being pointed to by any commit or tree, so it's now a dangling blob3168object.31693170Similarly, when the "recursive" merge strategy runs, and finds that3171there are criss-cross merges and thus more than one merge base (which is3172fairly unusual, but it does happen), it will generate one temporary3173midway tree (or possibly even more, if you had lots of criss-crossing3174merges and more than two merge bases) as a temporary internal merge3175base, and again, those are real objects, but the end result will not end3176up pointing to them, so they end up "dangling" in your repository.31773178Generally, dangling objects aren't anything to worry about. They can3179even be very useful: if you screw something up, the dangling objects can3180be how you recover your old tree (say, you did a rebase, and realized3181that you really didn't want to--you can look at what dangling objects3182you have, and decide to reset your head to some old dangling state).31833184For commits, you can just use:31853186------------------------------------------------3187$ gitk <dangling-commit-sha-goes-here> --not --all3188------------------------------------------------31893190This asks for all the history reachable from the given commit but not3191from any branch, tag, or other reference. If you decide it's something3192you want, you can always create a new reference to it, e.g.,31933194------------------------------------------------3195$ git branch recovered-branch <dangling-commit-sha-goes-here>3196------------------------------------------------31973198For blobs and trees, you can't do the same, but you can still examine3199them. You can just do32003201------------------------------------------------3202$ git show <dangling-blob/tree-sha-goes-here>3203------------------------------------------------32043205to show what the contents of the blob were (or, for a tree, basically3206what the "ls" for that directory was), and that may give you some idea3207of what the operation was that left that dangling object.32083209Usually, dangling blobs and trees aren't very interesting. They're3210almost always the result of either being a half-way mergebase (the blob3211will often even have the conflict markers from a merge in it, if you3212have had conflicting merges that you fixed up by hand), or simply3213because you interrupted a "git fetch" with ^C or something like that,3214leaving _some_ of the new objects in the object database, but just3215dangling and useless.32163217Anyway, once you are sure that you're not interested in any dangling3218state, you can just prune all unreachable objects:32193220------------------------------------------------3221$ git prune3222------------------------------------------------32233224and they'll be gone. But you should only run "git prune" on a quiescent3225repository--it's kind of like doing a filesystem fsck recovery: you3226don't want to do that while the filesystem is mounted.32273228(The same is true of "git fsck" itself, btw, but since3229`git fsck` never actually *changes* the repository, it just reports3230on what it found, `git fsck` itself is never 'dangerous' to run.3231Running it while somebody is actually changing the repository can cause3232confusing and scary messages, but it won't actually do anything bad. In3233contrast, running "git prune" while somebody is actively changing the3234repository is a *BAD* idea).32353236[[recovering-from-repository-corruption]]3237Recovering from repository corruption3238~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~32393240By design, git treats data trusted to it with caution. However, even in3241the absence of bugs in git itself, it is still possible that hardware or3242operating system errors could corrupt data.32433244The first defense against such problems is backups. You can back up a3245git directory using clone, or just using cp, tar, or any other backup3246mechanism.32473248As a last resort, you can search for the corrupted objects and attempt3249to replace them by hand. Back up your repository before attempting this3250in case you corrupt things even more in the process.32513252We'll assume that the problem is a single missing or corrupted blob,3253which is sometimes a solvable problem. (Recovering missing trees and3254especially commits is *much* harder).32553256Before starting, verify that there is corruption, and figure out where3257it is with linkgit:git-fsck[1]; this may be time-consuming.32583259Assume the output looks like this:32603261------------------------------------------------3262$ git fsck --full3263broken link from tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff83264 to blob 4b9458b3786228369c63936db65827de3cc062003265missing blob 4b9458b3786228369c63936db65827de3cc062003266------------------------------------------------32673268(Typically there will be some "dangling object" messages too, but they3269aren't interesting.)32703271Now you know that blob 4b9458b3 is missing, and that the tree 2d9263c63272points to it. If you could find just one copy of that missing blob3273object, possibly in some other repository, you could move it into3274.git/objects/4b/9458b3... and be done. Suppose you can't. You can3275still examine the tree that pointed to it with linkgit:git-ls-tree[1],3276which might output something like:32773278------------------------------------------------3279$ git ls-tree 2d9263c6d23595e7cb2a21e5ebbb53655278dff83280100644 blob 8d14531846b95bfa3564b58ccfb7913a034323b8 .gitignore3281100644 blob ebf9bf84da0aab5ed944264a5db2a65fe3a3e883 .mailmap3282100644 blob ca442d313d86dc67e0a2e5d584b465bd382cbf5c COPYING3283...3284100644 blob 4b9458b3786228369c63936db65827de3cc06200 myfile3285...3286------------------------------------------------32873288So now you know that the missing blob was the data for a file named3289"myfile". And chances are you can also identify the directory--let's3290say it's in "somedirectory". If you're lucky the missing copy might be3291the same as the copy you have checked out in your working tree at3292"somedirectory/myfile"; you can test whether that's right with3293linkgit:git-hash-object[1]:32943295------------------------------------------------3296$ git hash-object -w somedirectory/myfile3297------------------------------------------------32983299which will create and store a blob object with the contents of3300somedirectory/myfile, and output the SHA-1 of that object. if you're3301extremely lucky it might be 4b9458b3786228369c63936db65827de3cc06200, in3302which case you've guessed right, and the corruption is fixed!33033304Otherwise, you need more information. How do you tell which version of3305the file has been lost?33063307The easiest way to do this is with:33083309------------------------------------------------3310$ git log --raw --all --full-history -- somedirectory/myfile3311------------------------------------------------33123313Because you're asking for raw output, you'll now get something like33143315------------------------------------------------3316commit abc3317Author:3318Date:3319...3320:100644 100644 4b9458b... newsha... M somedirectory/myfile332133223323commit xyz3324Author:3325Date:33263327...3328:100644 100644 oldsha... 4b9458b... M somedirectory/myfile3329------------------------------------------------33303331This tells you that the immediately preceding version of the file was3332"newsha", and that the immediately following version was "oldsha".3333You also know the commit messages that went with the change from oldsha3334to 4b9458b and with the change from 4b9458b to newsha.33353336If you've been committing small enough changes, you may now have a good3337shot at reconstructing the contents of the in-between state 4b9458b.33383339If you can do that, you can now recreate the missing object with33403341------------------------------------------------3342$ git hash-object -w <recreated-file>3343------------------------------------------------33443345and your repository is good again!33463347(Btw, you could have ignored the fsck, and started with doing a33483349------------------------------------------------3350$ git log --raw --all3351------------------------------------------------33523353and just looked for the sha of the missing object (4b9458b..) in that3354whole thing. It's up to you - git does *have* a lot of information, it is3355just missing one particular blob version.33563357[[the-index]]3358The index3359-----------33603361The index is a binary file (generally kept in .git/index) containing a3362sorted list of path names, each with permissions and the SHA-1 of a blob3363object; linkgit:git-ls-files[1] can show you the contents of the index:33643365-------------------------------------------------3366$ git ls-files --stage3367100644 63c918c667fa005ff12ad89437f2fdc80926e21c 0 .gitignore3368100644 5529b198e8d14decbe4ad99db3f7fb632de0439d 0 .mailmap3369100644 6ff87c4664981e4397625791c8ea3bbb5f2279a3 0 COPYING3370100644 a37b2152bd26be2c2289e1f57a292534a51a93c7 0 Documentation/.gitignore3371100644 fbefe9a45b00a54b58d94d06eca48b03d40a50e0 0 Documentation/Makefile3372...3373100644 2511aef8d89ab52be5ec6a5e46236b4b6bcd07ea 0 xdiff/xtypes.h3374100644 2ade97b2574a9f77e7ae4002a4e07a6a38e46d07 0 xdiff/xutils.c3375100644 d5de8292e05e7c36c4b68857c1cf9855e3d2f70a 0 xdiff/xutils.h3376-------------------------------------------------33773378Note that in older documentation you may see the index called the3379"current directory cache" or just the "cache". It has three important3380properties:338133821. The index contains all the information necessary to generate a single3383(uniquely determined) tree object.3384+3385For example, running linkgit:git-commit[1] generates this tree object3386from the index, stores it in the object database, and uses it as the3387tree object associated with the new commit.338833892. The index enables fast comparisons between the tree object it defines3390and the working tree.3391+3392It does this by storing some additional data for each entry (such as3393the last modified time). This data is not displayed above, and is not3394stored in the created tree object, but it can be used to determine3395quickly which files in the working directory differ from what was3396stored in the index, and thus save git from having to read all of the3397data from such files to look for changes.339833993. It can efficiently represent information about merge conflicts3400between different tree objects, allowing each pathname to be3401associated with sufficient information about the trees involved that3402you can create a three-way merge between them.3403+3404We saw in <<conflict-resolution>> that during a merge the index can3405store multiple versions of a single file (called "stages"). The third3406column in the linkgit:git-ls-files[1] output above is the stage3407number, and will take on values other than 0 for files with merge3408conflicts.34093410The index is thus a sort of temporary staging area, which is filled with3411a tree which you are in the process of working on.34123413If you blow the index away entirely, you generally haven't lost any3414information as long as you have the name of the tree that it described.34153416[[submodules]]3417Submodules3418==========34193420Large projects are often composed of smaller, self-contained modules. For3421example, an embedded Linux distribution's source tree would include every3422piece of software in the distribution with some local modifications; a movie3423player might need to build against a specific, known-working version of a3424decompression library; several independent programs might all share the same3425build scripts.34263427With centralized revision control systems this is often accomplished by3428including every module in one single repository. Developers can check out3429all modules or only the modules they need to work with. They can even modify3430files across several modules in a single commit while moving things around3431or updating APIs and translations.34323433Git does not allow partial checkouts, so duplicating this approach in Git3434would force developers to keep a local copy of modules they are not3435interested in touching. Commits in an enormous checkout would be slower3436than you'd expect as Git would have to scan every directory for changes.3437If modules have a lot of local history, clones would take forever.34383439On the plus side, distributed revision control systems can much better3440integrate with external sources. In a centralized model, a single arbitrary3441snapshot of the external project is exported from its own revision control3442and then imported into the local revision control on a vendor branch. All3443the history is hidden. With distributed revision control you can clone the3444entire external history and much more easily follow development and re-merge3445local changes.34463447Git's submodule support allows a repository to contain, as a subdirectory, a3448checkout of an external project. Submodules maintain their own identity;3449the submodule support just stores the submodule repository location and3450commit ID, so other developers who clone the containing project3451("superproject") can easily clone all the submodules at the same revision.3452Partial checkouts of the superproject are possible: you can tell Git to3453clone none, some or all of the submodules.34543455The linkgit:git-submodule[1] command is available since Git 1.5.3. Users3456with Git 1.5.2 can look up the submodule commits in the repository and3457manually check them out; earlier versions won't recognize the submodules at3458all.34593460To see how submodule support works, create (for example) four example3461repositories that can be used later as a submodule:34623463-------------------------------------------------3464$ mkdir ~/git3465$ cd ~/git3466$ for i in a b c d3467do3468 mkdir $i3469 cd $i3470 git init3471 echo "module $i" > $i.txt3472 git add $i.txt3473 git commit -m "Initial commit, submodule $i"3474 cd ..3475done3476-------------------------------------------------34773478Now create the superproject and add all the submodules:34793480-------------------------------------------------3481$ mkdir super3482$ cd super3483$ git init3484$ for i in a b c d3485do3486 git submodule add ~/git/$i $i3487done3488-------------------------------------------------34893490NOTE: Do not use local URLs here if you plan to publish your superproject!34913492See what files `git submodule` created:34933494-------------------------------------------------3495$ ls -a3496. .. .git .gitmodules a b c d3497-------------------------------------------------34983499The `git submodule add <repo> <path>` command does a couple of things:35003501- It clones the submodule from <repo> to the given <path> under the3502 current directory and by default checks out the master branch.3503- It adds the submodule's clone path to the linkgit:gitmodules[5] file and3504 adds this file to the index, ready to be committed.3505- It adds the submodule's current commit ID to the index, ready to be3506 committed.35073508Commit the superproject:35093510-------------------------------------------------3511$ git commit -m "Add submodules a, b, c and d."3512-------------------------------------------------35133514Now clone the superproject:35153516-------------------------------------------------3517$ cd ..3518$ git clone super cloned3519$ cd cloned3520-------------------------------------------------35213522The submodule directories are there, but they're empty:35233524-------------------------------------------------3525$ ls -a a3526. ..3527$ git submodule status3528-d266b9873ad50488163457f025db7cdd9683d88b a3529-e81d457da15309b4fef4249aba9b50187999670d b3530-c1536a972b9affea0f16e0680ba87332dc059146 c3531-d96249ff5d57de5de093e6baff9e0aafa5276a74 d3532-------------------------------------------------35333534NOTE: The commit object names shown above would be different for you, but they3535should match the HEAD commit object names of your repositories. You can check3536it by running `git ls-remote ../a`.35373538Pulling down the submodules is a two-step process. First run `git submodule3539init` to add the submodule repository URLs to `.git/config`:35403541-------------------------------------------------3542$ git submodule init3543-------------------------------------------------35443545Now use `git submodule update` to clone the repositories and check out the3546commits specified in the superproject:35473548-------------------------------------------------3549$ git submodule update3550$ cd a3551$ ls -a3552. .. .git a.txt3553-------------------------------------------------35543555One major difference between `git submodule update` and `git submodule add` is3556that `git submodule update` checks out a specific commit, rather than the tip3557of a branch. It's like checking out a tag: the head is detached, so you're not3558working on a branch.35593560-------------------------------------------------3561$ git branch3562* (no branch)3563 master3564-------------------------------------------------35653566If you want to make a change within a submodule and you have a detached head,3567then you should create or checkout a branch, make your changes, publish the3568change within the submodule, and then update the superproject to reference the3569new commit:35703571-------------------------------------------------3572$ git checkout master3573-------------------------------------------------35743575or35763577-------------------------------------------------3578$ git checkout -b fix-up3579-------------------------------------------------35803581then35823583-------------------------------------------------3584$ echo "adding a line again" >> a.txt3585$ git commit -a -m "Updated the submodule from within the superproject."3586$ git push3587$ cd ..3588$ git diff3589diff --git a/a b/a3590index d266b98..261dfac 1600003591--- a/a3592+++ b/a3593@@ -1 +1 @@3594-Subproject commit d266b9873ad50488163457f025db7cdd9683d88b3595+Subproject commit 261dfac35cb99d380eb966e102c1197139f7fa243596$ git add a3597$ git commit -m "Updated submodule a."3598$ git push3599-------------------------------------------------36003601You have to run `git submodule update` after `git pull` if you want to update3602submodules, too.36033604Pitfalls with submodules3605------------------------36063607Always publish the submodule change before publishing the change to the3608superproject that references it. If you forget to publish the submodule change,3609others won't be able to clone the repository:36103611-------------------------------------------------3612$ cd ~/git/super/a3613$ echo i added another line to this file >> a.txt3614$ git commit -a -m "doing it wrong this time"3615$ cd ..3616$ git add a3617$ git commit -m "Updated submodule a again."3618$ git push3619$ cd ~/git/cloned3620$ git pull3621$ git submodule update3622error: pathspec '261dfac35cb99d380eb966e102c1197139f7fa24' did not match any file(s) known to git.3623Did you forget to 'git add'?3624Unable to checkout '261dfac35cb99d380eb966e102c1197139f7fa24' in submodule path 'a'3625-------------------------------------------------36263627You also should not rewind branches in a submodule beyond commits that were3628ever recorded in any superproject.36293630It's not safe to run `git submodule update` if you've made and committed3631changes within a submodule without checking out a branch first. They will be3632silently overwritten:36333634-------------------------------------------------3635$ cat a.txt3636module a3637$ echo line added from private2 >> a.txt3638$ git commit -a -m "line added inside private2"3639$ cd ..3640$ git submodule update3641Submodule path 'a': checked out 'd266b9873ad50488163457f025db7cdd9683d88b'3642$ cd a3643$ cat a.txt3644module a3645-------------------------------------------------36463647NOTE: The changes are still visible in the submodule's reflog.36483649This is not the case if you did not commit your changes.36503651[[low-level-operations]]3652Low-level git operations3653========================36543655Many of the higher-level commands were originally implemented as shell3656scripts using a smaller core of low-level git commands. These can still3657be useful when doing unusual things with git, or just as a way to3658understand its inner workings.36593660[[object-manipulation]]3661Object access and manipulation3662------------------------------36633664The linkgit:git-cat-file[1] command can show the contents of any object,3665though the higher-level linkgit:git-show[1] is usually more useful.36663667The linkgit:git-commit-tree[1] command allows constructing commits with3668arbitrary parents and trees.36693670A tree can be created with linkgit:git-write-tree[1] and its data can be3671accessed by linkgit:git-ls-tree[1]. Two trees can be compared with3672linkgit:git-diff-tree[1].36733674A tag is created with linkgit:git-mktag[1], and the signature can be3675verified by linkgit:git-verify-tag[1], though it is normally simpler to3676use linkgit:git-tag[1] for both.36773678[[the-workflow]]3679The Workflow3680------------36813682High-level operations such as linkgit:git-commit[1],3683linkgit:git-checkout[1] and linkgit:git-reset[1] work by moving data3684between the working tree, the index, and the object database. Git3685provides low-level operations which perform each of these steps3686individually.36873688Generally, all "git" operations work on the index file. Some operations3689work *purely* on the index file (showing the current state of the3690index), but most operations move data between the index file and either3691the database or the working directory. Thus there are four main3692combinations:36933694[[working-directory-to-index]]3695working directory -> index3696~~~~~~~~~~~~~~~~~~~~~~~~~~36973698The linkgit:git-update-index[1] command updates the index with3699information from the working directory. You generally update the3700index information by just specifying the filename you want to update,3701like so:37023703-------------------------------------------------3704$ git update-index filename3705-------------------------------------------------37063707but to avoid common mistakes with filename globbing etc, the command3708will not normally add totally new entries or remove old entries,3709i.e. it will normally just update existing cache entries.37103711To tell git that yes, you really do realize that certain files no3712longer exist, or that new files should be added, you3713should use the `--remove` and `--add` flags respectively.37143715NOTE! A `--remove` flag does 'not' mean that subsequent filenames will3716necessarily be removed: if the files still exist in your directory3717structure, the index will be updated with their new status, not3718removed. The only thing `--remove` means is that update-index will be3719considering a removed file to be a valid thing, and if the file really3720does not exist any more, it will update the index accordingly.37213722As a special case, you can also do `git update-index --refresh`, which3723will refresh the "stat" information of each index to match the current3724stat information. It will 'not' update the object status itself, and3725it will only update the fields that are used to quickly test whether3726an object still matches its old backing store object.37273728The previously introduced linkgit:git-add[1] is just a wrapper for3729linkgit:git-update-index[1].37303731[[index-to-object-database]]3732index -> object database3733~~~~~~~~~~~~~~~~~~~~~~~~37343735You write your current index file to a "tree" object with the program37363737-------------------------------------------------3738$ git write-tree3739-------------------------------------------------37403741that doesn't come with any options--it will just write out the3742current index into the set of tree objects that describe that state,3743and it will return the name of the resulting top-level tree. You can3744use that tree to re-generate the index at any time by going in the3745other direction:37463747[[object-database-to-index]]3748object database -> index3749~~~~~~~~~~~~~~~~~~~~~~~~37503751You read a "tree" file from the object database, and use that to3752populate (and overwrite--don't do this if your index contains any3753unsaved state that you might want to restore later!) your current3754index. Normal operation is just37553756-------------------------------------------------3757$ git read-tree <SHA-1 of tree>3758-------------------------------------------------37593760and your index file will now be equivalent to the tree that you saved3761earlier. However, that is only your 'index' file: your working3762directory contents have not been modified.37633764[[index-to-working-directory]]3765index -> working directory3766~~~~~~~~~~~~~~~~~~~~~~~~~~37673768You update your working directory from the index by "checking out"3769files. This is not a very common operation, since normally you'd just3770keep your files updated, and rather than write to your working3771directory, you'd tell the index files about the changes in your3772working directory (i.e. `git update-index`).37733774However, if you decide to jump to a new version, or check out somebody3775else's version, or just restore a previous tree, you'd populate your3776index file with read-tree, and then you need to check out the result3777with37783779-------------------------------------------------3780$ git checkout-index filename3781-------------------------------------------------37823783or, if you want to check out all of the index, use `-a`.37843785NOTE! `git checkout-index` normally refuses to overwrite old files, so3786if you have an old version of the tree already checked out, you will3787need to use the "-f" flag ('before' the "-a" flag or the filename) to3788'force' the checkout.378937903791Finally, there are a few odds and ends which are not purely moving3792from one representation to the other:37933794[[tying-it-all-together]]3795Tying it all together3796~~~~~~~~~~~~~~~~~~~~~37973798To commit a tree you have instantiated with "git write-tree", you'd3799create a "commit" object that refers to that tree and the history3800behind it--most notably the "parent" commits that preceded it in3801history.38023803Normally a "commit" has one parent: the previous state of the tree3804before a certain change was made. However, sometimes it can have two3805or more parent commits, in which case we call it a "merge", due to the3806fact that such a commit brings together ("merges") two or more3807previous states represented by other commits.38083809In other words, while a "tree" represents a particular directory state3810of a working directory, a "commit" represents that state in "time",3811and explains how we got there.38123813You create a commit object by giving it the tree that describes the3814state at the time of the commit, and a list of parents:38153816-------------------------------------------------3817$ git commit-tree <tree> -p <parent> [-p <parent2> ..]3818-------------------------------------------------38193820and then giving the reason for the commit on stdin (either through3821redirection from a pipe or file, or by just typing it at the tty).38223823`git commit-tree` will return the name of the object that represents3824that commit, and you should save it away for later use. Normally,3825you'd commit a new `HEAD` state, and while git doesn't care where you3826save the note about that state, in practice we tend to just write the3827result to the file pointed at by `.git/HEAD`, so that we can always see3828what the last committed state was.38293830Here is an ASCII art by Jon Loeliger that illustrates how3831various pieces fit together.38323833------------38343835 commit-tree3836 commit obj3837 +----+3838 | |3839 | |3840 V V3841 +-----------+3842 | Object DB |3843 | Backing |3844 | Store |3845 +-----------+3846 ^3847 write-tree | |3848 tree obj | |3849 | | read-tree3850 | | tree obj3851 V3852 +-----------+3853 | Index |3854 | "cache" |3855 +-----------+3856 update-index ^3857 blob obj | |3858 | |3859 checkout-index -u | | checkout-index3860 stat | | blob obj3861 V3862 +-----------+3863 | Working |3864 | Directory |3865 +-----------+38663867------------386838693870[[examining-the-data]]3871Examining the data3872------------------38733874You can examine the data represented in the object database and the3875index with various helper tools. For every object, you can use3876linkgit:git-cat-file[1] to examine details about the3877object:38783879-------------------------------------------------3880$ git cat-file -t <objectname>3881-------------------------------------------------38823883shows the type of the object, and once you have the type (which is3884usually implicit in where you find the object), you can use38853886-------------------------------------------------3887$ git cat-file blob|tree|commit|tag <objectname>3888-------------------------------------------------38893890to show its contents. NOTE! Trees have binary content, and as a result3891there is a special helper for showing that content, called3892`git ls-tree`, which turns the binary content into a more easily3893readable form.38943895It's especially instructive to look at "commit" objects, since those3896tend to be small and fairly self-explanatory. In particular, if you3897follow the convention of having the top commit name in `.git/HEAD`,3898you can do38993900-------------------------------------------------3901$ git cat-file commit HEAD3902-------------------------------------------------39033904to see what the top commit was.39053906[[merging-multiple-trees]]3907Merging multiple trees3908----------------------39093910Git helps you do a three-way merge, which you can expand to n-way by3911repeating the merge procedure arbitrary times until you finally3912"commit" the state. The normal situation is that you'd only do one3913three-way merge (two parents), and commit it, but if you like to, you3914can do multiple parents in one go.39153916To do a three-way merge, you need the two sets of "commit" objects3917that you want to merge, use those to find the closest common parent (a3918third "commit" object), and then use those commit objects to find the3919state of the directory ("tree" object) at these points.39203921To get the "base" for the merge, you first look up the common parent3922of two commits with39233924-------------------------------------------------3925$ git merge-base <commit1> <commit2>3926-------------------------------------------------39273928which will return you the commit they are both based on. You should3929now look up the "tree" objects of those commits, which you can easily3930do with (for example)39313932-------------------------------------------------3933$ git cat-file commit <commitname> | head -13934-------------------------------------------------39353936since the tree object information is always the first line in a commit3937object.39383939Once you know the three trees you are going to merge (the one "original"3940tree, aka the common tree, and the two "result" trees, aka the branches3941you want to merge), you do a "merge" read into the index. This will3942complain if it has to throw away your old index contents, so you should3943make sure that you've committed those--in fact you would normally3944always do a merge against your last commit (which should thus match what3945you have in your current index anyway).39463947To do the merge, do39483949-------------------------------------------------3950$ git read-tree -m -u <origtree> <yourtree> <targettree>3951-------------------------------------------------39523953which will do all trivial merge operations for you directly in the3954index file, and you can just write the result out with3955`git write-tree`.395639573958[[merging-multiple-trees-2]]3959Merging multiple trees, continued3960---------------------------------39613962Sadly, many merges aren't trivial. If there are files that have3963been added, moved or removed, or if both branches have modified the3964same file, you will be left with an index tree that contains "merge3965entries" in it. Such an index tree can 'NOT' be written out to a tree3966object, and you will have to resolve any such merge clashes using3967other tools before you can write out the result.39683969You can examine such index state with `git ls-files --unmerged`3970command. An example:39713972------------------------------------------------3973$ git read-tree -m $orig HEAD $target3974$ git ls-files --unmerged3975100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello.c3976100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello.c3977100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello.c3978------------------------------------------------39793980Each line of the `git ls-files --unmerged` output begins with3981the blob mode bits, blob SHA-1, 'stage number', and the3982filename. The 'stage number' is git's way to say which tree it3983came from: stage 1 corresponds to `$orig` tree, stage 2 `HEAD`3984tree, and stage3 `$target` tree.39853986Earlier we said that trivial merges are done inside3987`git read-tree -m`. For example, if the file did not change3988from `$orig` to `HEAD` nor `$target`, or if the file changed3989from `$orig` to `HEAD` and `$orig` to `$target` the same way,3990obviously the final outcome is what is in `HEAD`. What the3991above example shows is that file `hello.c` was changed from3992`$orig` to `HEAD` and `$orig` to `$target` in a different way.3993You could resolve this by running your favorite 3-way merge3994program, e.g. `diff3`, `merge`, or git's own merge-file, on3995the blob objects from these three stages yourself, like this:39963997------------------------------------------------3998$ git cat-file blob 263414f... >hello.c~13999$ git cat-file blob 06fa6a2... >hello.c~24000$ git cat-file blob cc44c73... >hello.c~34001$ git merge-file hello.c~2 hello.c~1 hello.c~34002------------------------------------------------40034004This would leave the merge result in `hello.c~2` file, along4005with conflict markers if there are conflicts. After verifying4006the merge result makes sense, you can tell git what the final4007merge result for this file is by:40084009-------------------------------------------------4010$ mv -f hello.c~2 hello.c4011$ git update-index hello.c4012-------------------------------------------------40134014When a path is in the "unmerged" state, running `git update-index` for4015that path tells git to mark the path resolved.40164017The above is the description of a git merge at the lowest level,4018to help you understand what conceptually happens under the hood.4019In practice, nobody, not even git itself, runs `git cat-file` three times4020for this. There is a `git merge-index` program that extracts the4021stages to temporary files and calls a "merge" script on it:40224023-------------------------------------------------4024$ git merge-index git-merge-one-file hello.c4025-------------------------------------------------40264027and that is what higher level `git merge -s resolve` is implemented with.40284029[[hacking-git]]4030Hacking git4031===========40324033This chapter covers internal details of the git implementation which4034probably only git developers need to understand.40354036[[object-details]]4037Object storage format4038---------------------40394040All objects have a statically determined "type" which identifies the4041format of the object (i.e. how it is used, and how it can refer to other4042objects). There are currently four different object types: "blob",4043"tree", "commit", and "tag".40444045Regardless of object type, all objects share the following4046characteristics: they are all deflated with zlib, and have a header4047that not only specifies their type, but also provides size information4048about the data in the object. It's worth noting that the SHA-1 hash4049that is used to name the object is the hash of the original data4050plus this header, so `sha1sum` 'file' does not match the object name4051for 'file'.4052(Historical note: in the dawn of the age of git the hash4053was the SHA-1 of the 'compressed' object.)40544055As a result, the general consistency of an object can always be tested4056independently of the contents or the type of the object: all objects can4057be validated by verifying that (a) their hashes match the content of the4058file and (b) the object successfully inflates to a stream of bytes that4059forms a sequence of <ascii type without space> {plus} <space> {plus} <ascii decimal4060size> {plus} <byte\0> {plus} <binary object data>.40614062The structured objects can further have their structure and4063connectivity to other objects verified. This is generally done with4064the `git fsck` program, which generates a full dependency graph4065of all objects, and verifies their internal consistency (in addition4066to just verifying their superficial consistency through the hash).40674068[[birdview-on-the-source-code]]4069A birds-eye view of Git's source code4070-------------------------------------40714072It is not always easy for new developers to find their way through Git's4073source code. This section gives you a little guidance to show where to4074start.40754076A good place to start is with the contents of the initial commit, with:40774078----------------------------------------------------4079$ git checkout e83c51634080----------------------------------------------------40814082The initial revision lays the foundation for almost everything git has4083today, but is small enough to read in one sitting.40844085Note that terminology has changed since that revision. For example, the4086README in that revision uses the word "changeset" to describe what we4087now call a <<def_commit_object,commit>>.40884089Also, we do not call it "cache" any more, but rather "index"; however, the4090file is still called `cache.h`. Remark: Not much reason to change it now,4091especially since there is no good single name for it anyway, because it is4092basically _the_ header file which is included by _all_ of Git's C sources.40934094If you grasp the ideas in that initial commit, you should check out a4095more recent version and skim `cache.h`, `object.h` and `commit.h`.40964097In the early days, Git (in the tradition of UNIX) was a bunch of programs4098which were extremely simple, and which you used in scripts, piping the4099output of one into another. This turned out to be good for initial4100development, since it was easier to test new things. However, recently4101many of these parts have become builtins, and some of the core has been4102"libified", i.e. put into libgit.a for performance, portability reasons,4103and to avoid code duplication.41044105By now, you know what the index is (and find the corresponding data4106structures in `cache.h`), and that there are just a couple of object types4107(blobs, trees, commits and tags) which inherit their common structure from4108`struct object`, which is their first member (and thus, you can cast e.g.4109`(struct object *)commit` to achieve the _same_ as `&commit->object`, i.e.4110get at the object name and flags).41114112Now is a good point to take a break to let this information sink in.41134114Next step: get familiar with the object naming. Read <<naming-commits>>.4115There are quite a few ways to name an object (and not only revisions!).4116All of these are handled in `sha1_name.c`. Just have a quick look at4117the function `get_sha1()`. A lot of the special handling is done by4118functions like `get_sha1_basic()` or the likes.41194120This is just to get you into the groove for the most libified part of Git:4121the revision walker.41224123Basically, the initial version of `git log` was a shell script:41244125----------------------------------------------------------------4126$ git-rev-list --pretty $(git-rev-parse --default HEAD "$@") | \4127 LESS=-S ${PAGER:-less}4128----------------------------------------------------------------41294130What does this mean?41314132`git rev-list` is the original version of the revision walker, which4133_always_ printed a list of revisions to stdout. It is still functional,4134and needs to, since most new Git programs start out as scripts using4135`git rev-list`.41364137`git rev-parse` is not as important any more; it was only used to filter out4138options that were relevant for the different plumbing commands that were4139called by the script.41404141Most of what `git rev-list` did is contained in `revision.c` and4142`revision.h`. It wraps the options in a struct named `rev_info`, which4143controls how and what revisions are walked, and more.41444145The original job of `git rev-parse` is now taken by the function4146`setup_revisions()`, which parses the revisions and the common command line4147options for the revision walker. This information is stored in the struct4148`rev_info` for later consumption. You can do your own command line option4149parsing after calling `setup_revisions()`. After that, you have to call4150`prepare_revision_walk()` for initialization, and then you can get the4151commits one by one with the function `get_revision()`.41524153If you are interested in more details of the revision walking process,4154just have a look at the first implementation of `cmd_log()`; call4155`git show v1.3.0{tilde}155^2{tilde}4` and scroll down to that function (note that you4156no longer need to call `setup_pager()` directly).41574158Nowadays, `git log` is a builtin, which means that it is _contained_ in the4159command `git`. The source side of a builtin is41604161- a function called `cmd_<bla>`, typically defined in `builtin-<bla>.c`,4162 and declared in `builtin.h`,41634164- an entry in the `commands[]` array in `git.c`, and41654166- an entry in `BUILTIN_OBJECTS` in the `Makefile`.41674168Sometimes, more than one builtin is contained in one source file. For4169example, `cmd_whatchanged()` and `cmd_log()` both reside in `builtin-log.c`,4170since they share quite a bit of code. In that case, the commands which are4171_not_ named like the `.c` file in which they live have to be listed in4172`BUILT_INS` in the `Makefile`.41734174`git log` looks more complicated in C than it does in the original script,4175but that allows for a much greater flexibility and performance.41764177Here again it is a good point to take a pause.41784179Lesson three is: study the code. Really, it is the best way to learn about4180the organization of Git (after you know the basic concepts).41814182So, think about something which you are interested in, say, "how can I4183access a blob just knowing the object name of it?". The first step is to4184find a Git command with which you can do it. In this example, it is either4185`git show` or `git cat-file`.41864187For the sake of clarity, let's stay with `git cat-file`, because it41884189- is plumbing, and41904191- was around even in the initial commit (it literally went only through4192 some 20 revisions as `cat-file.c`, was renamed to `builtin-cat-file.c`4193 when made a builtin, and then saw less than 10 versions).41944195So, look into `builtin-cat-file.c`, search for `cmd_cat_file()` and look what4196it does.41974198------------------------------------------------------------------4199 git_config(git_default_config);4200 if (argc != 3)4201 usage("git cat-file [-t|-s|-e|-p|<type>] <sha1>");4202 if (get_sha1(argv[2], sha1))4203 die("Not a valid object name %s", argv[2]);4204------------------------------------------------------------------42054206Let's skip over the obvious details; the only really interesting part4207here is the call to `get_sha1()`. It tries to interpret `argv[2]` as an4208object name, and if it refers to an object which is present in the current4209repository, it writes the resulting SHA-1 into the variable `sha1`.42104211Two things are interesting here:42124213- `get_sha1()` returns 0 on _success_. This might surprise some new4214 Git hackers, but there is a long tradition in UNIX to return different4215 negative numbers in case of different errors--and 0 on success.42164217- the variable `sha1` in the function signature of `get_sha1()` is `unsigned4218 char \*`, but is actually expected to be a pointer to `unsigned4219 char[20]`. This variable will contain the 160-bit SHA-1 of the given4220 commit. Note that whenever a SHA-1 is passed as `unsigned char \*`, it4221 is the binary representation, as opposed to the ASCII representation in4222 hex characters, which is passed as `char *`.42234224You will see both of these things throughout the code.42254226Now, for the meat:42274228-----------------------------------------------------------------------------4229 case 0:4230 buf = read_object_with_reference(sha1, argv[1], &size, NULL);4231-----------------------------------------------------------------------------42324233This is how you read a blob (actually, not only a blob, but any type of4234object). To know how the function `read_object_with_reference()` actually4235works, find the source code for it (something like `git grep4236read_object_with | grep ":[a-z]"` in the git repository), and read4237the source.42384239To find out how the result can be used, just read on in `cmd_cat_file()`:42404241-----------------------------------4242 write_or_die(1, buf, size);4243-----------------------------------42444245Sometimes, you do not know where to look for a feature. In many such cases,4246it helps to search through the output of `git log`, and then `git show` the4247corresponding commit.42484249Example: If you know that there was some test case for `git bundle`, but4250do not remember where it was (yes, you _could_ `git grep bundle t/`, but that4251does not illustrate the point!):42524253------------------------4254$ git log --no-merges t/4255------------------------42564257In the pager (`less`), just search for "bundle", go a few lines back,4258and see that it is in commit 18449ab0... Now just copy this object name,4259and paste it into the command line42604261-------------------4262$ git show 18449ab04263-------------------42644265Voila.42664267Another example: Find out what to do in order to make some script a4268builtin:42694270-------------------------------------------------4271$ git log --no-merges --diff-filter=A builtin-*.c4272-------------------------------------------------42734274You see, Git is actually the best tool to find out about the source of Git4275itself!42764277[[glossary]]4278GIT Glossary4279============42804281include::glossary-content.txt[]42824283[[git-quick-start]]4284Appendix A: Git Quick Reference4285===============================42864287This is a quick summary of the major commands; the previous chapters4288explain how these work in more detail.42894290[[quick-creating-a-new-repository]]4291Creating a new repository4292-------------------------42934294From a tarball:42954296-----------------------------------------------4297$ tar xzf project.tar.gz4298$ cd project4299$ git init4300Initialized empty Git repository in .git/4301$ git add .4302$ git commit4303-----------------------------------------------43044305From a remote repository:43064307-----------------------------------------------4308$ git clone git://example.com/pub/project.git4309$ cd project4310-----------------------------------------------43114312[[managing-branches]]4313Managing branches4314-----------------43154316-----------------------------------------------4317$ git branch # list all local branches in this repo4318$ git checkout test # switch working directory to branch "test"4319$ git branch new # create branch "new" starting at current HEAD4320$ git branch -d new # delete branch "new"4321-----------------------------------------------43224323Instead of basing a new branch on current HEAD (the default), use:43244325-----------------------------------------------4326$ git branch new test # branch named "test"4327$ git branch new v2.6.15 # tag named v2.6.154328$ git branch new HEAD^ # commit before the most recent4329$ git branch new HEAD^^ # commit before that4330$ git branch new test~10 # ten commits before tip of branch "test"4331-----------------------------------------------43324333Create and switch to a new branch at the same time:43344335-----------------------------------------------4336$ git checkout -b new v2.6.154337-----------------------------------------------43384339Update and examine branches from the repository you cloned from:43404341-----------------------------------------------4342$ git fetch # update4343$ git branch -r # list4344 origin/master4345 origin/next4346 ...4347$ git checkout -b masterwork origin/master4348-----------------------------------------------43494350Fetch a branch from a different repository, and give it a new4351name in your repository:43524353-----------------------------------------------4354$ git fetch git://example.com/project.git theirbranch:mybranch4355$ git fetch git://example.com/project.git v2.6.15:mybranch4356-----------------------------------------------43574358Keep a list of repositories you work with regularly:43594360-----------------------------------------------4361$ git remote add example git://example.com/project.git4362$ git remote # list remote repositories4363example4364origin4365$ git remote show example # get details4366* remote example4367 URL: git://example.com/project.git4368 Tracked remote branches4369 master4370 next4371 ...4372$ git fetch example # update branches from example4373$ git branch -r # list all remote branches4374-----------------------------------------------437543764377[[exploring-history]]4378Exploring history4379-----------------43804381-----------------------------------------------4382$ gitk # visualize and browse history4383$ git log # list all commits4384$ git log src/ # ...modifying src/4385$ git log v2.6.15..v2.6.16 # ...in v2.6.16, not in v2.6.154386$ git log master..test # ...in branch test, not in branch master4387$ git log test..master # ...in branch master, but not in test4388$ git log test...master # ...in one branch, not in both4389$ git log -S'foo()' # ...where difference contain "foo()"4390$ git log --since="2 weeks ago"4391$ git log -p # show patches as well4392$ git show # most recent commit4393$ git diff v2.6.15..v2.6.16 # diff between two tagged versions4394$ git diff v2.6.15..HEAD # diff with current head4395$ git grep "foo()" # search working directory for "foo()"4396$ git grep v2.6.15 "foo()" # search old tree for "foo()"4397$ git show v2.6.15:a.txt # look at old version of a.txt4398-----------------------------------------------43994400Search for regressions:44014402-----------------------------------------------4403$ git bisect start4404$ git bisect bad # current version is bad4405$ git bisect good v2.6.13-rc2 # last known good revision4406Bisecting: 675 revisions left to test after this4407 # test here, then:4408$ git bisect good # if this revision is good, or4409$ git bisect bad # if this revision is bad.4410 # repeat until done.4411-----------------------------------------------44124413[[making-changes]]4414Making changes4415--------------44164417Make sure git knows who to blame:44184419------------------------------------------------4420$ cat >>~/.gitconfig <<\EOF4421[user]4422 name = Your Name Comes Here4423 email = you@yourdomain.example.com4424EOF4425------------------------------------------------44264427Select file contents to include in the next commit, then make the4428commit:44294430-----------------------------------------------4431$ git add a.txt # updated file4432$ git add b.txt # new file4433$ git rm c.txt # old file4434$ git commit4435-----------------------------------------------44364437Or, prepare and create the commit in one step:44384439-----------------------------------------------4440$ git commit d.txt # use latest content only of d.txt4441$ git commit -a # use latest content of all tracked files4442-----------------------------------------------44434444[[merging]]4445Merging4446-------44474448-----------------------------------------------4449$ git merge test # merge branch "test" into the current branch4450$ git pull git://example.com/project.git master4451 # fetch and merge in remote branch4452$ git pull . test # equivalent to git merge test4453-----------------------------------------------44544455[[sharing-your-changes]]4456Sharing your changes4457--------------------44584459Importing or exporting patches:44604461-----------------------------------------------4462$ git format-patch origin..HEAD # format a patch for each commit4463 # in HEAD but not in origin4464$ git am mbox # import patches from the mailbox "mbox"4465-----------------------------------------------44664467Fetch a branch in a different git repository, then merge into the4468current branch:44694470-----------------------------------------------4471$ git pull git://example.com/project.git theirbranch4472-----------------------------------------------44734474Store the fetched branch into a local branch before merging into the4475current branch:44764477-----------------------------------------------4478$ git pull git://example.com/project.git theirbranch:mybranch4479-----------------------------------------------44804481After creating commits on a local branch, update the remote4482branch with your commits:44834484-----------------------------------------------4485$ git push ssh://example.com/project.git mybranch:theirbranch4486-----------------------------------------------44874488When remote and local branch are both named "test":44894490-----------------------------------------------4491$ git push ssh://example.com/project.git test4492-----------------------------------------------44934494Shortcut version for a frequently used remote repository:44954496-----------------------------------------------4497$ git remote add example ssh://example.com/project.git4498$ git push example test4499-----------------------------------------------45004501[[repository-maintenance]]4502Repository maintenance4503----------------------45044505Check for corruption:45064507-----------------------------------------------4508$ git fsck4509-----------------------------------------------45104511Recompress, remove unused cruft:45124513-----------------------------------------------4514$ git gc4515-----------------------------------------------451645174518[[todo]]4519Appendix B: Notes and todo list for this manual4520===============================================45214522This is a work in progress.45234524The basic requirements:45254526- It must be readable in order, from beginning to end, by someone4527 intelligent with a basic grasp of the UNIX command line, but without4528 any special knowledge of git. If necessary, any other prerequisites4529 should be specifically mentioned as they arise.4530- Whenever possible, section headings should clearly describe the task4531 they explain how to do, in language that requires no more knowledge4532 than necessary: for example, "importing patches into a project" rather4533 than "the `git am` command"45344535Think about how to create a clear chapter dependency graph that will4536allow people to get to important topics without necessarily reading4537everything in between.45384539Scan Documentation/ for other stuff left out; in particular:45404541- howto's4542- some of technical/?4543- hooks4544- list of commands in linkgit:git[1]45454546Scan email archives for other stuff left out45474548Scan man pages to see if any assume more background than this manual4549provides.45504551Simplify beginning by suggesting disconnected head instead of4552temporary branch creation?45534554Add more good examples. Entire sections of just cookbook examples4555might be a good idea; maybe make an "advanced examples" section a4556standard end-of-chapter section?45574558Include cross-references to the glossary, where appropriate.45594560Document shallow clones? See draft 1.5.0 release notes for some4561documentation.45624563Add a section on working with other version control systems, including4564CVS, Subversion, and just imports of series of release tarballs.45654566More details on gitweb?45674568Write a chapter on using plumbing and writing scripts.45694570Alternates, clone -reference, etc.45714572More on recovery from repository corruption. See:4573 http://marc.theaimsgroup.com/?l=git&m=117263864820799&w=24574 http://marc.theaimsgroup.com/?l=git&m=117147855503798&w=2