Merge branch 'aw/mirror-push' into jk/send-pack
[git/git.git] / Documentation / core-tutorial.txt
1 A git core tutorial for developers
2 ==================================
3
4 Introduction
5 ------------
6
7 This tutorial explains how to use the "core" git programs to set up and
8 work with a git repository.
9
10 If you just need to use git as a revision control system you may prefer
11 to start with link:tutorial.html[a tutorial introduction to git] or
12 link:user-manual.html[the git user manual].
13
14 However, an understanding of these low-level tools can be helpful if
15 you want to understand git's internals.
16
17 The core git is often called "plumbing", with the prettier user
18 interfaces on top of it called "porcelain". You may not want to use the
19 plumbing directly very often, but it can be good to know what the
20 plumbing does for when the porcelain isn't flushing.
21
22 [NOTE]
23 Deeper technical details are often marked as Notes, which you can
24 skip on your first reading.
25
26
27 Creating a git repository
28 -------------------------
29
30 Creating a new git repository couldn't be easier: all git repositories start
31 out empty, and the only thing you need to do is find yourself a
32 subdirectory that you want to use as a working tree - either an empty
33 one for a totally new project, or an existing working tree that you want
34 to import into git.
35
36 For our first example, we're going to start a totally new repository from
37 scratch, with no pre-existing files, and we'll call it `git-tutorial`.
38 To start up, create a subdirectory for it, change into that
39 subdirectory, and initialize the git infrastructure with `git-init`:
40
41 ------------------------------------------------
42 $ mkdir git-tutorial
43 $ cd git-tutorial
44 $ git-init
45 ------------------------------------------------
46
47 to which git will reply
48
49 ----------------
50 Initialized empty Git repository in .git/
51 ----------------
52
53 which is just git's way of saying that you haven't been doing anything
54 strange, and that it will have created a local `.git` directory setup for
55 your new project. You will now have a `.git` directory, and you can
56 inspect that with `ls`. For your new empty project, it should show you
57 three entries, among other things:
58
59 - a file called `HEAD`, that has `ref: refs/heads/master` in it.
60 This is similar to a symbolic link and points at
61 `refs/heads/master` relative to the `HEAD` file.
62 +
63 Don't worry about the fact that the file that the `HEAD` link points to
64 doesn't even exist yet -- you haven't created the commit that will
65 start your `HEAD` development branch yet.
66
67 - a subdirectory called `objects`, which will contain all the
68 objects of your project. You should never have any real reason to
69 look at the objects directly, but you might want to know that these
70 objects are what contains all the real 'data' in your repository.
71
72 - a subdirectory called `refs`, which contains references to objects.
73
74 In particular, the `refs` subdirectory will contain two other
75 subdirectories, named `heads` and `tags` respectively. They do
76 exactly what their names imply: they contain references to any number
77 of different 'heads' of development (aka 'branches'), and to any
78 'tags' that you have created to name specific versions in your
79 repository.
80
81 One note: the special `master` head is the default branch, which is
82 why the `.git/HEAD` file was created points to it even if it
83 doesn't yet exist. Basically, the `HEAD` link is supposed to always
84 point to the branch you are working on right now, and you always
85 start out expecting to work on the `master` branch.
86
87 However, this is only a convention, and you can name your branches
88 anything you want, and don't have to ever even 'have' a `master`
89 branch. A number of the git tools will assume that `.git/HEAD` is
90 valid, though.
91
92 [NOTE]
93 An 'object' is identified by its 160-bit SHA1 hash, aka 'object name',
94 and a reference to an object is always the 40-byte hex
95 representation of that SHA1 name. The files in the `refs`
96 subdirectory are expected to contain these hex references
97 (usually with a final `\'\n\'` at the end), and you should thus
98 expect to see a number of 41-byte files containing these
99 references in these `refs` subdirectories when you actually start
100 populating your tree.
101
102 [NOTE]
103 An advanced user may want to take a look at the
104 link:repository-layout.html[repository layout] document
105 after finishing this tutorial.
106
107 You have now created your first git repository. Of course, since it's
108 empty, that's not very useful, so let's start populating it with data.
109
110
111 Populating a git repository
112 ---------------------------
113
114 We'll keep this simple and stupid, so we'll start off with populating a
115 few trivial files just to get a feel for it.
116
117 Start off with just creating any random files that you want to maintain
118 in your git repository. We'll start off with a few bad examples, just to
119 get a feel for how this works:
120
121 ------------------------------------------------
122 $ echo "Hello World" >hello
123 $ echo "Silly example" >example
124 ------------------------------------------------
125
126 you have now created two files in your working tree (aka 'working directory'),
127 but to actually check in your hard work, you will have to go through two steps:
128
129 - fill in the 'index' file (aka 'cache') with the information about your
130 working tree state.
131
132 - commit that index file as an object.
133
134 The first step is trivial: when you want to tell git about any changes
135 to your working tree, you use the `git-update-index` program. That
136 program normally just takes a list of filenames you want to update, but
137 to avoid trivial mistakes, it refuses to add new entries to the index
138 (or remove existing ones) unless you explicitly tell it that you're
139 adding a new entry with the `\--add` flag (or removing an entry with the
140 `\--remove`) flag.
141
142 So to populate the index with the two files you just created, you can do
143
144 ------------------------------------------------
145 $ git-update-index --add hello example
146 ------------------------------------------------
147
148 and you have now told git to track those two files.
149
150 In fact, as you did that, if you now look into your object directory,
151 you'll notice that git will have added two new objects to the object
152 database. If you did exactly the steps above, you should now be able to do
153
154
155 ----------------
156 $ ls .git/objects/??/*
157 ----------------
158
159 and see two files:
160
161 ----------------
162 .git/objects/55/7db03de997c86a4a028e1ebd3a1ceb225be238
163 .git/objects/f2/4c74a2e500f5ee1332c86b94199f52b1d1d962
164 ----------------
165
166 which correspond with the objects with names of `557db...` and
167 `f24c7...` respectively.
168
169 If you want to, you can use `git-cat-file` to look at those objects, but
170 you'll have to use the object name, not the filename of the object:
171
172 ----------------
173 $ git-cat-file -t 557db03de997c86a4a028e1ebd3a1ceb225be238
174 ----------------
175
176 where the `-t` tells `git-cat-file` to tell you what the "type" of the
177 object is. git will tell you that you have a "blob" object (i.e., just a
178 regular file), and you can see the contents with
179
180 ----------------
181 $ git-cat-file "blob" 557db03
182 ----------------
183
184 which will print out "Hello World". The object `557db03` is nothing
185 more than the contents of your file `hello`.
186
187 [NOTE]
188 Don't confuse that object with the file `hello` itself. The
189 object is literally just those specific *contents* of the file, and
190 however much you later change the contents in file `hello`, the object
191 we just looked at will never change. Objects are immutable.
192
193 [NOTE]
194 The second example demonstrates that you can
195 abbreviate the object name to only the first several
196 hexadecimal digits in most places.
197
198 Anyway, as we mentioned previously, you normally never actually take a
199 look at the objects themselves, and typing long 40-character hex
200 names is not something you'd normally want to do. The above digression
201 was just to show that `git-update-index` did something magical, and
202 actually saved away the contents of your files into the git object
203 database.
204
205 Updating the index did something else too: it created a `.git/index`
206 file. This is the index that describes your current working tree, and
207 something you should be very aware of. Again, you normally never worry
208 about the index file itself, but you should be aware of the fact that
209 you have not actually really "checked in" your files into git so far,
210 you've only *told* git about them.
211
212 However, since git knows about them, you can now start using some of the
213 most basic git commands to manipulate the files or look at their status.
214
215 In particular, let's not even check in the two files into git yet, we'll
216 start off by adding another line to `hello` first:
217
218 ------------------------------------------------
219 $ echo "It's a new day for git" >>hello
220 ------------------------------------------------
221
222 and you can now, since you told git about the previous state of `hello`, ask
223 git what has changed in the tree compared to your old index, using the
224 `git-diff-files` command:
225
226 ------------
227 $ git-diff-files
228 ------------
229
230 Oops. That wasn't very readable. It just spit out its own internal
231 version of a `diff`, but that internal version really just tells you
232 that it has noticed that "hello" has been modified, and that the old object
233 contents it had have been replaced with something else.
234
235 To make it readable, we can tell git-diff-files to output the
236 differences as a patch, using the `-p` flag:
237
238 ------------
239 $ git-diff-files -p
240 diff --git a/hello b/hello
241 index 557db03..263414f 100644
242 --- a/hello
243 +++ b/hello
244 @@ -1 +1,2 @@
245 Hello World
246 +It's a new day for git
247 ----
248
249 i.e. the diff of the change we caused by adding another line to `hello`.
250
251 In other words, `git-diff-files` always shows us the difference between
252 what is recorded in the index, and what is currently in the working
253 tree. That's very useful.
254
255 A common shorthand for `git-diff-files -p` is to just write `git
256 diff`, which will do the same thing.
257
258 ------------
259 $ git diff
260 diff --git a/hello b/hello
261 index 557db03..263414f 100644
262 --- a/hello
263 +++ b/hello
264 @@ -1 +1,2 @@
265 Hello World
266 +It's a new day for git
267 ------------
268
269
270 Committing git state
271 --------------------
272
273 Now, we want to go to the next stage in git, which is to take the files
274 that git knows about in the index, and commit them as a real tree. We do
275 that in two phases: creating a 'tree' object, and committing that 'tree'
276 object as a 'commit' object together with an explanation of what the
277 tree was all about, along with information of how we came to that state.
278
279 Creating a tree object is trivial, and is done with `git-write-tree`.
280 There are no options or other input: git-write-tree will take the
281 current index state, and write an object that describes that whole
282 index. In other words, we're now tying together all the different
283 filenames with their contents (and their permissions), and we're
284 creating the equivalent of a git "directory" object:
285
286 ------------------------------------------------
287 $ git-write-tree
288 ------------------------------------------------
289
290 and this will just output the name of the resulting tree, in this case
291 (if you have done exactly as I've described) it should be
292
293 ----------------
294 8988da15d077d4829fc51d8544c097def6644dbb
295 ----------------
296
297 which is another incomprehensible object name. Again, if you want to,
298 you can use `git-cat-file -t 8988d\...` to see that this time the object
299 is not a "blob" object, but a "tree" object (you can also use
300 `git-cat-file` to actually output the raw object contents, but you'll see
301 mainly a binary mess, so that's less interesting).
302
303 However -- normally you'd never use `git-write-tree` on its own, because
304 normally you always commit a tree into a commit object using the
305 `git-commit-tree` command. In fact, it's easier to not actually use
306 `git-write-tree` on its own at all, but to just pass its result in as an
307 argument to `git-commit-tree`.
308
309 `git-commit-tree` normally takes several arguments -- it wants to know
310 what the 'parent' of a commit was, but since this is the first commit
311 ever in this new repository, and it has no parents, we only need to pass in
312 the object name of the tree. However, `git-commit-tree` also wants to get a
313 commit message on its standard input, and it will write out the resulting
314 object name for the commit to its standard output.
315
316 And this is where we create the `.git/refs/heads/master` file
317 which is pointed at by `HEAD`. This file is supposed to contain
318 the reference to the top-of-tree of the master branch, and since
319 that's exactly what `git-commit-tree` spits out, we can do this
320 all with a sequence of simple shell commands:
321
322 ------------------------------------------------
323 $ tree=$(git-write-tree)
324 $ commit=$(echo 'Initial commit' | git-commit-tree $tree)
325 $ git-update-ref HEAD $commit
326 ------------------------------------------------
327
328 In this case this creates a totally new commit that is not related to
329 anything else. Normally you do this only *once* for a project ever, and
330 all later commits will be parented on top of an earlier commit.
331
332 Again, normally you'd never actually do this by hand. There is a
333 helpful script called `git commit` that will do all of this for you. So
334 you could have just written `git commit`
335 instead, and it would have done the above magic scripting for you.
336
337
338 Making a change
339 ---------------
340
341 Remember how we did the `git-update-index` on file `hello` and then we
342 changed `hello` afterward, and could compare the new state of `hello` with the
343 state we saved in the index file?
344
345 Further, remember how I said that `git-write-tree` writes the contents
346 of the *index* file to the tree, and thus what we just committed was in
347 fact the *original* contents of the file `hello`, not the new ones. We did
348 that on purpose, to show the difference between the index state, and the
349 state in the working tree, and how they don't have to match, even
350 when we commit things.
351
352 As before, if we do `git-diff-files -p` in our git-tutorial project,
353 we'll still see the same difference we saw last time: the index file
354 hasn't changed by the act of committing anything. However, now that we
355 have committed something, we can also learn to use a new command:
356 `git-diff-index`.
357
358 Unlike `git-diff-files`, which showed the difference between the index
359 file and the working tree, `git-diff-index` shows the differences
360 between a committed *tree* and either the index file or the working
361 tree. In other words, `git-diff-index` wants a tree to be diffed
362 against, and before we did the commit, we couldn't do that, because we
363 didn't have anything to diff against.
364
365 But now we can do
366
367 ----------------
368 $ git-diff-index -p HEAD
369 ----------------
370
371 (where `-p` has the same meaning as it did in `git-diff-files`), and it
372 will show us the same difference, but for a totally different reason.
373 Now we're comparing the working tree not against the index file,
374 but against the tree we just wrote. It just so happens that those two
375 are obviously the same, so we get the same result.
376
377 Again, because this is a common operation, you can also just shorthand
378 it with
379
380 ----------------
381 $ git diff HEAD
382 ----------------
383
384 which ends up doing the above for you.
385
386 In other words, `git-diff-index` normally compares a tree against the
387 working tree, but when given the `\--cached` flag, it is told to
388 instead compare against just the index cache contents, and ignore the
389 current working tree state entirely. Since we just wrote the index
390 file to HEAD, doing `git-diff-index \--cached -p HEAD` should thus return
391 an empty set of differences, and that's exactly what it does.
392
393 [NOTE]
394 ================
395 `git-diff-index` really always uses the index for its
396 comparisons, and saying that it compares a tree against the working
397 tree is thus not strictly accurate. In particular, the list of
398 files to compare (the "meta-data") *always* comes from the index file,
399 regardless of whether the `\--cached` flag is used or not. The `\--cached`
400 flag really only determines whether the file *contents* to be compared
401 come from the working tree or not.
402
403 This is not hard to understand, as soon as you realize that git simply
404 never knows (or cares) about files that it is not told about
405 explicitly. git will never go *looking* for files to compare, it
406 expects you to tell it what the files are, and that's what the index
407 is there for.
408 ================
409
410 However, our next step is to commit the *change* we did, and again, to
411 understand what's going on, keep in mind the difference between "working
412 tree contents", "index file" and "committed tree". We have changes
413 in the working tree that we want to commit, and we always have to
414 work through the index file, so the first thing we need to do is to
415 update the index cache:
416
417 ------------------------------------------------
418 $ git-update-index hello
419 ------------------------------------------------
420
421 (note how we didn't need the `\--add` flag this time, since git knew
422 about the file already).
423
424 Note what happens to the different `git-diff-\*` versions here. After
425 we've updated `hello` in the index, `git-diff-files -p` now shows no
426 differences, but `git-diff-index -p HEAD` still *does* show that the
427 current state is different from the state we committed. In fact, now
428 `git-diff-index` shows the same difference whether we use the `--cached`
429 flag or not, since now the index is coherent with the working tree.
430
431 Now, since we've updated `hello` in the index, we can commit the new
432 version. We could do it by writing the tree by hand again, and
433 committing the tree (this time we'd have to use the `-p HEAD` flag to
434 tell commit that the HEAD was the *parent* of the new commit, and that
435 this wasn't an initial commit any more), but you've done that once
436 already, so let's just use the helpful script this time:
437
438 ------------------------------------------------
439 $ git commit
440 ------------------------------------------------
441
442 which starts an editor for you to write the commit message and tells you
443 a bit about what you have done.
444
445 Write whatever message you want, and all the lines that start with '#'
446 will be pruned out, and the rest will be used as the commit message for
447 the change. If you decide you don't want to commit anything after all at
448 this point (you can continue to edit things and update the index), you
449 can just leave an empty message. Otherwise `git commit` will commit
450 the change for you.
451
452 You've now made your first real git commit. And if you're interested in
453 looking at what `git commit` really does, feel free to investigate:
454 it's a few very simple shell scripts to generate the helpful (?) commit
455 message headers, and a few one-liners that actually do the
456 commit itself (`git-commit`).
457
458
459 Inspecting Changes
460 ------------------
461
462 While creating changes is useful, it's even more useful if you can tell
463 later what changed. The most useful command for this is another of the
464 `diff` family, namely `git-diff-tree`.
465
466 `git-diff-tree` can be given two arbitrary trees, and it will tell you the
467 differences between them. Perhaps even more commonly, though, you can
468 give it just a single commit object, and it will figure out the parent
469 of that commit itself, and show the difference directly. Thus, to get
470 the same diff that we've already seen several times, we can now do
471
472 ----------------
473 $ git-diff-tree -p HEAD
474 ----------------
475
476 (again, `-p` means to show the difference as a human-readable patch),
477 and it will show what the last commit (in `HEAD`) actually changed.
478
479 [NOTE]
480 ============
481 Here is an ASCII art by Jon Loeliger that illustrates how
482 various diff-\* commands compare things.
483
484 diff-tree
485 +----+
486 | |
487 | |
488 V V
489 +-----------+
490 | Object DB |
491 | Backing |
492 | Store |
493 +-----------+
494 ^ ^
495 | |
496 | | diff-index --cached
497 | |
498 diff-index | V
499 | +-----------+
500 | | Index |
501 | | "cache" |
502 | +-----------+
503 | ^
504 | |
505 | | diff-files
506 | |
507 V V
508 +-----------+
509 | Working |
510 | Directory |
511 +-----------+
512 ============
513
514 More interestingly, you can also give `git-diff-tree` the `--pretty` flag,
515 which tells it to also show the commit message and author and date of the
516 commit, and you can tell it to show a whole series of diffs.
517 Alternatively, you can tell it to be "silent", and not show the diffs at
518 all, but just show the actual commit message.
519
520 In fact, together with the `git-rev-list` program (which generates a
521 list of revisions), `git-diff-tree` ends up being a veritable fount of
522 changes. A trivial (but very useful) script called `git-whatchanged` is
523 included with git which does exactly this, and shows a log of recent
524 activities.
525
526 To see the whole history of our pitiful little git-tutorial project, you
527 can do
528
529 ----------------
530 $ git log
531 ----------------
532
533 which shows just the log messages, or if we want to see the log together
534 with the associated patches use the more complex (and much more
535 powerful)
536
537 ----------------
538 $ git-whatchanged -p --root
539 ----------------
540
541 and you will see exactly what has changed in the repository over its
542 short history.
543
544 [NOTE]
545 The `\--root` flag is a flag to `git-diff-tree` to tell it to
546 show the initial aka 'root' commit too. Normally you'd probably not
547 want to see the initial import diff, but since the tutorial project
548 was started from scratch and is so small, we use it to make the result
549 a bit more interesting.
550
551 With that, you should now be having some inkling of what git does, and
552 can explore on your own.
553
554 [NOTE]
555 Most likely, you are not directly using the core
556 git Plumbing commands, but using Porcelain such as `git-add`, `git-rm'
557 and `git-commit'.
558
559
560 Tagging a version
561 -----------------
562
563 In git, there are two kinds of tags, a "light" one, and an "annotated tag".
564
565 A "light" tag is technically nothing more than a branch, except we put
566 it in the `.git/refs/tags/` subdirectory instead of calling it a `head`.
567 So the simplest form of tag involves nothing more than
568
569 ------------------------------------------------
570 $ git tag my-first-tag
571 ------------------------------------------------
572
573 which just writes the current `HEAD` into the `.git/refs/tags/my-first-tag`
574 file, after which point you can then use this symbolic name for that
575 particular state. You can, for example, do
576
577 ----------------
578 $ git diff my-first-tag
579 ----------------
580
581 to diff your current state against that tag (which at this point will
582 obviously be an empty diff, but if you continue to develop and commit
583 stuff, you can use your tag as an "anchor-point" to see what has changed
584 since you tagged it.
585
586 An "annotated tag" is actually a real git object, and contains not only a
587 pointer to the state you want to tag, but also a small tag name and
588 message, along with optionally a PGP signature that says that yes,
589 you really did
590 that tag. You create these annotated tags with either the `-a` or
591 `-s` flag to `git tag`:
592
593 ----------------
594 $ git tag -s <tagname>
595 ----------------
596
597 which will sign the current `HEAD` (but you can also give it another
598 argument that specifies the thing to tag, i.e., you could have tagged the
599 current `mybranch` point by using `git tag <tagname> mybranch`).
600
601 You normally only do signed tags for major releases or things
602 like that, while the light-weight tags are useful for any marking you
603 want to do -- any time you decide that you want to remember a certain
604 point, just create a private tag for it, and you have a nice symbolic
605 name for the state at that point.
606
607
608 Copying repositories
609 --------------------
610
611 git repositories are normally totally self-sufficient and relocatable.
612 Unlike CVS, for example, there is no separate notion of
613 "repository" and "working tree". A git repository normally *is* the
614 working tree, with the local git information hidden in the `.git`
615 subdirectory. There is nothing else. What you see is what you got.
616
617 [NOTE]
618 You can tell git to split the git internal information from
619 the directory that it tracks, but we'll ignore that for now: it's not
620 how normal projects work, and it's really only meant for special uses.
621 So the mental model of "the git information is always tied directly to
622 the working tree that it describes" may not be technically 100%
623 accurate, but it's a good model for all normal use.
624
625 This has two implications:
626
627 - if you grow bored with the tutorial repository you created (or you've
628 made a mistake and want to start all over), you can just do simple
629 +
630 ----------------
631 $ rm -rf git-tutorial
632 ----------------
633 +
634 and it will be gone. There's no external repository, and there's no
635 history outside the project you created.
636
637 - if you want to move or duplicate a git repository, you can do so. There
638 is `git clone` command, but if all you want to do is just to
639 create a copy of your repository (with all the full history that
640 went along with it), you can do so with a regular
641 `cp -a git-tutorial new-git-tutorial`.
642 +
643 Note that when you've moved or copied a git repository, your git index
644 file (which caches various information, notably some of the "stat"
645 information for the files involved) will likely need to be refreshed.
646 So after you do a `cp -a` to create a new copy, you'll want to do
647 +
648 ----------------
649 $ git-update-index --refresh
650 ----------------
651 +
652 in the new repository to make sure that the index file is up-to-date.
653
654 Note that the second point is true even across machines. You can
655 duplicate a remote git repository with *any* regular copy mechanism, be it
656 `scp`, `rsync` or `wget`.
657
658 When copying a remote repository, you'll want to at a minimum update the
659 index cache when you do this, and especially with other peoples'
660 repositories you often want to make sure that the index cache is in some
661 known state (you don't know *what* they've done and not yet checked in),
662 so usually you'll precede the `git-update-index` with a
663
664 ----------------
665 $ git-read-tree --reset HEAD
666 $ git-update-index --refresh
667 ----------------
668
669 which will force a total index re-build from the tree pointed to by `HEAD`.
670 It resets the index contents to `HEAD`, and then the `git-update-index`
671 makes sure to match up all index entries with the checked-out files.
672 If the original repository had uncommitted changes in its
673 working tree, `git-update-index --refresh` notices them and
674 tells you they need to be updated.
675
676 The above can also be written as simply
677
678 ----------------
679 $ git reset
680 ----------------
681
682 and in fact a lot of the common git command combinations can be scripted
683 with the `git xyz` interfaces. You can learn things by just looking
684 at what the various git scripts do. For example, `git reset` used to be
685 the above two lines implemented in `git-reset`, but some things like
686 `git status` and `git commit` are slightly more complex scripts around
687 the basic git commands.
688
689 Many (most?) public remote repositories will not contain any of
690 the checked out files or even an index file, and will *only* contain the
691 actual core git files. Such a repository usually doesn't even have the
692 `.git` subdirectory, but has all the git files directly in the
693 repository.
694
695 To create your own local live copy of such a "raw" git repository, you'd
696 first create your own subdirectory for the project, and then copy the
697 raw repository contents into the `.git` directory. For example, to
698 create your own copy of the git repository, you'd do the following
699
700 ----------------
701 $ mkdir my-git
702 $ cd my-git
703 $ rsync -rL rsync://rsync.kernel.org/pub/scm/git/git.git/ .git
704 ----------------
705
706 followed by
707
708 ----------------
709 $ git-read-tree HEAD
710 ----------------
711
712 to populate the index. However, now you have populated the index, and
713 you have all the git internal files, but you will notice that you don't
714 actually have any of the working tree files to work on. To get
715 those, you'd check them out with
716
717 ----------------
718 $ git-checkout-index -u -a
719 ----------------
720
721 where the `-u` flag means that you want the checkout to keep the index
722 up-to-date (so that you don't have to refresh it afterward), and the
723 `-a` flag means "check out all files" (if you have a stale copy or an
724 older version of a checked out tree you may also need to add the `-f`
725 flag first, to tell git-checkout-index to *force* overwriting of any old
726 files).
727
728 Again, this can all be simplified with
729
730 ----------------
731 $ git clone rsync://rsync.kernel.org/pub/scm/git/git.git/ my-git
732 $ cd my-git
733 $ git checkout
734 ----------------
735
736 which will end up doing all of the above for you.
737
738 You have now successfully copied somebody else's (mine) remote
739 repository, and checked it out.
740
741
742 Creating a new branch
743 ---------------------
744
745 Branches in git are really nothing more than pointers into the git
746 object database from within the `.git/refs/` subdirectory, and as we
747 already discussed, the `HEAD` branch is nothing but a symlink to one of
748 these object pointers.
749
750 You can at any time create a new branch by just picking an arbitrary
751 point in the project history, and just writing the SHA1 name of that
752 object into a file under `.git/refs/heads/`. You can use any filename you
753 want (and indeed, subdirectories), but the convention is that the
754 "normal" branch is called `master`. That's just a convention, though,
755 and nothing enforces it.
756
757 To show that as an example, let's go back to the git-tutorial repository we
758 used earlier, and create a branch in it. You do that by simply just
759 saying that you want to check out a new branch:
760
761 ------------
762 $ git checkout -b mybranch
763 ------------
764
765 will create a new branch based at the current `HEAD` position, and switch
766 to it.
767
768 [NOTE]
769 ================================================
770 If you make the decision to start your new branch at some
771 other point in the history than the current `HEAD`, you can do so by
772 just telling `git checkout` what the base of the checkout would be.
773 In other words, if you have an earlier tag or branch, you'd just do
774
775 ------------
776 $ git checkout -b mybranch earlier-commit
777 ------------
778
779 and it would create the new branch `mybranch` at the earlier commit,
780 and check out the state at that time.
781 ================================================
782
783 You can always just jump back to your original `master` branch by doing
784
785 ------------
786 $ git checkout master
787 ------------
788
789 (or any other branch-name, for that matter) and if you forget which
790 branch you happen to be on, a simple
791
792 ------------
793 $ cat .git/HEAD
794 ------------
795
796 will tell you where it's pointing. To get the list of branches
797 you have, you can say
798
799 ------------
800 $ git branch
801 ------------
802
803 which used to be nothing more than a simple script around `ls .git/refs/heads`.
804 There will be an asterisk in front of the branch you are currently on.
805
806 Sometimes you may wish to create a new branch _without_ actually
807 checking it out and switching to it. If so, just use the command
808
809 ------------
810 $ git branch <branchname> [startingpoint]
811 ------------
812
813 which will simply _create_ the branch, but will not do anything further.
814 You can then later -- once you decide that you want to actually develop
815 on that branch -- switch to that branch with a regular `git checkout`
816 with the branchname as the argument.
817
818
819 Merging two branches
820 --------------------
821
822 One of the ideas of having a branch is that you do some (possibly
823 experimental) work in it, and eventually merge it back to the main
824 branch. So assuming you created the above `mybranch` that started out
825 being the same as the original `master` branch, let's make sure we're in
826 that branch, and do some work there.
827
828 ------------------------------------------------
829 $ git checkout mybranch
830 $ echo "Work, work, work" >>hello
831 $ git commit -m "Some work." -i hello
832 ------------------------------------------------
833
834 Here, we just added another line to `hello`, and we used a shorthand for
835 doing both `git-update-index hello` and `git commit` by just giving the
836 filename directly to `git commit`, with an `-i` flag (it tells
837 git to 'include' that file in addition to what you have done to
838 the index file so far when making the commit). The `-m` flag is to give the
839 commit log message from the command line.
840
841 Now, to make it a bit more interesting, let's assume that somebody else
842 does some work in the original branch, and simulate that by going back
843 to the master branch, and editing the same file differently there:
844
845 ------------
846 $ git checkout master
847 ------------
848
849 Here, take a moment to look at the contents of `hello`, and notice how they
850 don't contain the work we just did in `mybranch` -- because that work
851 hasn't happened in the `master` branch at all. Then do
852
853 ------------
854 $ echo "Play, play, play" >>hello
855 $ echo "Lots of fun" >>example
856 $ git commit -m "Some fun." -i hello example
857 ------------
858
859 since the master branch is obviously in a much better mood.
860
861 Now, you've got two branches, and you decide that you want to merge the
862 work done. Before we do that, let's introduce a cool graphical tool that
863 helps you view what's going on:
864
865 ----------------
866 $ gitk --all
867 ----------------
868
869 will show you graphically both of your branches (that's what the `\--all`
870 means: normally it will just show you your current `HEAD`) and their
871 histories. You can also see exactly how they came to be from a common
872 source.
873
874 Anyway, let's exit `gitk` (`^Q` or the File menu), and decide that we want
875 to merge the work we did on the `mybranch` branch into the `master`
876 branch (which is currently our `HEAD` too). To do that, there's a nice
877 script called `git merge`, which wants to know which branches you want
878 to resolve and what the merge is all about:
879
880 ------------
881 $ git merge -m "Merge work in mybranch" mybranch
882 ------------
883
884 where the first argument is going to be used as the commit message if
885 the merge can be resolved automatically.
886
887 Now, in this case we've intentionally created a situation where the
888 merge will need to be fixed up by hand, though, so git will do as much
889 of it as it can automatically (which in this case is just merge the `example`
890 file, which had no differences in the `mybranch` branch), and say:
891
892 ----------------
893 Auto-merging hello
894 CONFLICT (content): Merge conflict in hello
895 Automatic merge failed; fix up by hand
896 ----------------
897
898 It tells you that it did an "Automatic merge", which
899 failed due to conflicts in `hello`.
900
901 Not to worry. It left the (trivial) conflict in `hello` in the same form you
902 should already be well used to if you've ever used CVS, so let's just
903 open `hello` in our editor (whatever that may be), and fix it up somehow.
904 I'd suggest just making it so that `hello` contains all four lines:
905
906 ------------
907 Hello World
908 It's a new day for git
909 Play, play, play
910 Work, work, work
911 ------------
912
913 and once you're happy with your manual merge, just do a
914
915 ------------
916 $ git commit -i hello
917 ------------
918
919 which will very loudly warn you that you're now committing a merge
920 (which is correct, so never mind), and you can write a small merge
921 message about your adventures in git-merge-land.
922
923 After you're done, start up `gitk \--all` to see graphically what the
924 history looks like. Notice that `mybranch` still exists, and you can
925 switch to it, and continue to work with it if you want to. The
926 `mybranch` branch will not contain the merge, but next time you merge it
927 from the `master` branch, git will know how you merged it, so you'll not
928 have to do _that_ merge again.
929
930 Another useful tool, especially if you do not always work in X-Window
931 environment, is `git show-branch`.
932
933 ------------------------------------------------
934 $ git-show-branch --topo-order --more=1 master mybranch
935 * [master] Merge work in mybranch
936 ! [mybranch] Some work.
937 --
938 - [master] Merge work in mybranch
939 *+ [mybranch] Some work.
940 * [master^] Some fun.
941 ------------------------------------------------
942
943 The first two lines indicate that it is showing the two branches
944 and the first line of the commit log message from their
945 top-of-the-tree commits, you are currently on `master` branch
946 (notice the asterisk `\*` character), and the first column for
947 the later output lines is used to show commits contained in the
948 `master` branch, and the second column for the `mybranch`
949 branch. Three commits are shown along with their log messages.
950 All of them have non blank characters in the first column (`*`
951 shows an ordinary commit on the current branch, `-` is a merge commit), which
952 means they are now part of the `master` branch. Only the "Some
953 work" commit has the plus `+` character in the second column,
954 because `mybranch` has not been merged to incorporate these
955 commits from the master branch. The string inside brackets
956 before the commit log message is a short name you can use to
957 name the commit. In the above example, 'master' and 'mybranch'
958 are branch heads. 'master^' is the first parent of 'master'
959 branch head. Please see 'git-rev-parse' documentation if you
960 see more complex cases.
961
962 [NOTE]
963 Without the '--more=1' option, 'git-show-branch' would not output the
964 '[master^]' commit, as '[mybranch]' commit is a common ancestor of
965 both 'master' and 'mybranch' tips. Please see 'git-show-branch'
966 documentation for details.
967
968 [NOTE]
969 If there were more commits on the 'master' branch after the merge, the
970 merge commit itself would not be shown by 'git-show-branch' by
971 default. You would need to provide '--sparse' option to make the
972 merge commit visible in this case.
973
974 Now, let's pretend you are the one who did all the work in
975 `mybranch`, and the fruit of your hard work has finally been merged
976 to the `master` branch. Let's go back to `mybranch`, and run
977 `git merge` to get the "upstream changes" back to your branch.
978
979 ------------
980 $ git checkout mybranch
981 $ git merge -m "Merge upstream changes." master
982 ------------
983
984 This outputs something like this (the actual commit object names
985 would be different)
986
987 ----------------
988 Updating from ae3a2da... to a80b4aa....
989 Fast forward
990 example | 1 +
991 hello | 1 +
992 2 files changed, 2 insertions(+), 0 deletions(-)
993 ----------------
994
995 Because your branch did not contain anything more than what are
996 already merged into the `master` branch, the merge operation did
997 not actually do a merge. Instead, it just updated the top of
998 the tree of your branch to that of the `master` branch. This is
999 often called 'fast forward' merge.
1000
1001 You can run `gitk \--all` again to see how the commit ancestry
1002 looks like, or run `show-branch`, which tells you this.
1003
1004 ------------------------------------------------
1005 $ git show-branch master mybranch
1006 ! [master] Merge work in mybranch
1007 * [mybranch] Merge work in mybranch
1008 --
1009 -- [master] Merge work in mybranch
1010 ------------------------------------------------
1011
1012
1013 Merging external work
1014 ---------------------
1015
1016 It's usually much more common that you merge with somebody else than
1017 merging with your own branches, so it's worth pointing out that git
1018 makes that very easy too, and in fact, it's not that different from
1019 doing a `git merge`. In fact, a remote merge ends up being nothing
1020 more than "fetch the work from a remote repository into a temporary tag"
1021 followed by a `git merge`.
1022
1023 Fetching from a remote repository is done by, unsurprisingly,
1024 `git fetch`:
1025
1026 ----------------
1027 $ git fetch <remote-repository>
1028 ----------------
1029
1030 One of the following transports can be used to name the
1031 repository to download from:
1032
1033 Rsync::
1034 `rsync://remote.machine/path/to/repo.git/`
1035 +
1036 Rsync transport is usable for both uploading and downloading,
1037 but is completely unaware of what git does, and can produce
1038 unexpected results when you download from the public repository
1039 while the repository owner is uploading into it via `rsync`
1040 transport. Most notably, it could update the files under
1041 `refs/` which holds the object name of the topmost commits
1042 before uploading the files in `objects/` -- the downloader would
1043 obtain head commit object name while that object itself is still
1044 not available in the repository. For this reason, it is
1045 considered deprecated.
1046
1047 SSH::
1048 `remote.machine:/path/to/repo.git/` or
1049 +
1050 `ssh://remote.machine/path/to/repo.git/`
1051 +
1052 This transport can be used for both uploading and downloading,
1053 and requires you to have a log-in privilege over `ssh` to the
1054 remote machine. It finds out the set of objects the other side
1055 lacks by exchanging the head commits both ends have and
1056 transfers (close to) minimum set of objects. It is by far the
1057 most efficient way to exchange git objects between repositories.
1058
1059 Local directory::
1060 `/path/to/repo.git/`
1061 +
1062 This transport is the same as SSH transport but uses `sh` to run
1063 both ends on the local machine instead of running other end on
1064 the remote machine via `ssh`.
1065
1066 git Native::
1067 `git://remote.machine/path/to/repo.git/`
1068 +
1069 This transport was designed for anonymous downloading. Like SSH
1070 transport, it finds out the set of objects the downstream side
1071 lacks and transfers (close to) minimum set of objects.
1072
1073 HTTP(S)::
1074 `http://remote.machine/path/to/repo.git/`
1075 +
1076 Downloader from http and https URL
1077 first obtains the topmost commit object name from the remote site
1078 by looking at the specified refname under `repo.git/refs/` directory,
1079 and then tries to obtain the
1080 commit object by downloading from `repo.git/objects/xx/xxx\...`
1081 using the object name of that commit object. Then it reads the
1082 commit object to find out its parent commits and the associate
1083 tree object; it repeats this process until it gets all the
1084 necessary objects. Because of this behavior, they are
1085 sometimes also called 'commit walkers'.
1086 +
1087 The 'commit walkers' are sometimes also called 'dumb
1088 transports', because they do not require any git aware smart
1089 server like git Native transport does. Any stock HTTP server
1090 that does not even support directory index would suffice. But
1091 you must prepare your repository with `git-update-server-info`
1092 to help dumb transport downloaders.
1093
1094 Once you fetch from the remote repository, you `merge` that
1095 with your current branch.
1096
1097 However -- it's such a common thing to `fetch` and then
1098 immediately `merge`, that it's called `git pull`, and you can
1099 simply do
1100
1101 ----------------
1102 $ git pull <remote-repository>
1103 ----------------
1104
1105 and optionally give a branch-name for the remote end as a second
1106 argument.
1107
1108 [NOTE]
1109 You could do without using any branches at all, by
1110 keeping as many local repositories as you would like to have
1111 branches, and merging between them with `git pull`, just like
1112 you merge between branches. The advantage of this approach is
1113 that it lets you keep a set of files for each `branch` checked
1114 out and you may find it easier to switch back and forth if you
1115 juggle multiple lines of development simultaneously. Of
1116 course, you will pay the price of more disk usage to hold
1117 multiple working trees, but disk space is cheap these days.
1118
1119 It is likely that you will be pulling from the same remote
1120 repository from time to time. As a short hand, you can store
1121 the remote repository URL in the local repository's config file
1122 like this:
1123
1124 ------------------------------------------------
1125 $ git config remote.linus.url http://www.kernel.org/pub/scm/git/git.git/
1126 ------------------------------------------------
1127
1128 and use the "linus" keyword with `git pull` instead of the full URL.
1129
1130 Examples.
1131
1132 . `git pull linus`
1133 . `git pull linus tag v0.99.1`
1134
1135 the above are equivalent to:
1136
1137 . `git pull http://www.kernel.org/pub/scm/git/git.git/ HEAD`
1138 . `git pull http://www.kernel.org/pub/scm/git/git.git/ tag v0.99.1`
1139
1140
1141 How does the merge work?
1142 ------------------------
1143
1144 We said this tutorial shows what plumbing does to help you cope
1145 with the porcelain that isn't flushing, but we so far did not
1146 talk about how the merge really works. If you are following
1147 this tutorial the first time, I'd suggest to skip to "Publishing
1148 your work" section and come back here later.
1149
1150 OK, still with me? To give us an example to look at, let's go
1151 back to the earlier repository with "hello" and "example" file,
1152 and bring ourselves back to the pre-merge state:
1153
1154 ------------
1155 $ git show-branch --more=3 master mybranch
1156 ! [master] Merge work in mybranch
1157 * [mybranch] Merge work in mybranch
1158 --
1159 -- [master] Merge work in mybranch
1160 +* [master^2] Some work.
1161 +* [master^] Some fun.
1162 ------------
1163
1164 Remember, before running `git merge`, our `master` head was at
1165 "Some fun." commit, while our `mybranch` head was at "Some
1166 work." commit.
1167
1168 ------------
1169 $ git checkout mybranch
1170 $ git reset --hard master^2
1171 $ git checkout master
1172 $ git reset --hard master^
1173 ------------
1174
1175 After rewinding, the commit structure should look like this:
1176
1177 ------------
1178 $ git show-branch
1179 * [master] Some fun.
1180 ! [mybranch] Some work.
1181 --
1182 + [mybranch] Some work.
1183 * [master] Some fun.
1184 *+ [mybranch^] New day.
1185 ------------
1186
1187 Now we are ready to experiment with the merge by hand.
1188
1189 `git merge` command, when merging two branches, uses 3-way merge
1190 algorithm. First, it finds the common ancestor between them.
1191 The command it uses is `git-merge-base`:
1192
1193 ------------
1194 $ mb=$(git-merge-base HEAD mybranch)
1195 ------------
1196
1197 The command writes the commit object name of the common ancestor
1198 to the standard output, so we captured its output to a variable,
1199 because we will be using it in the next step. By the way, the common
1200 ancestor commit is the "New day." commit in this case. You can
1201 tell it by:
1202
1203 ------------
1204 $ git-name-rev $mb
1205 my-first-tag
1206 ------------
1207
1208 After finding out a common ancestor commit, the second step is
1209 this:
1210
1211 ------------
1212 $ git-read-tree -m -u $mb HEAD mybranch
1213 ------------
1214
1215 This is the same `git-read-tree` command we have already seen,
1216 but it takes three trees, unlike previous examples. This reads
1217 the contents of each tree into different 'stage' in the index
1218 file (the first tree goes to stage 1, the second stage 2,
1219 etc.). After reading three trees into three stages, the paths
1220 that are the same in all three stages are 'collapsed' into stage
1221 0. Also paths that are the same in two of three stages are
1222 collapsed into stage 0, taking the SHA1 from either stage 2 or
1223 stage 3, whichever is different from stage 1 (i.e. only one side
1224 changed from the common ancestor).
1225
1226 After 'collapsing' operation, paths that are different in three
1227 trees are left in non-zero stages. At this point, you can
1228 inspect the index file with this command:
1229
1230 ------------
1231 $ git-ls-files --stage
1232 100644 7f8b141b65fdcee47321e399a2598a235a032422 0 example
1233 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello
1234 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello
1235 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello
1236 ------------
1237
1238 In our example of only two files, we did not have unchanged
1239 files so only 'example' resulted in collapsing, but in real-life
1240 large projects, only small number of files change in one commit,
1241 and this 'collapsing' tends to trivially merge most of the paths
1242 fairly quickly, leaving only a handful the real changes in non-zero
1243 stages.
1244
1245 To look at only non-zero stages, use `\--unmerged` flag:
1246
1247 ------------
1248 $ git-ls-files --unmerged
1249 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello
1250 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello
1251 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello
1252 ------------
1253
1254 The next step of merging is to merge these three versions of the
1255 file, using 3-way merge. This is done by giving
1256 `git-merge-one-file` command as one of the arguments to
1257 `git-merge-index` command:
1258
1259 ------------
1260 $ git-merge-index git-merge-one-file hello
1261 Auto-merging hello.
1262 merge: warning: conflicts during merge
1263 ERROR: Merge conflict in hello.
1264 fatal: merge program failed
1265 ------------
1266
1267 `git-merge-one-file` script is called with parameters to
1268 describe those three versions, and is responsible to leave the
1269 merge results in the working tree.
1270 It is a fairly straightforward shell script, and
1271 eventually calls `merge` program from RCS suite to perform a
1272 file-level 3-way merge. In this case, `merge` detects
1273 conflicts, and the merge result with conflict marks is left in
1274 the working tree.. This can be seen if you run `ls-files
1275 --stage` again at this point:
1276
1277 ------------
1278 $ git-ls-files --stage
1279 100644 7f8b141b65fdcee47321e399a2598a235a032422 0 example
1280 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello
1281 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello
1282 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello
1283 ------------
1284
1285 This is the state of the index file and the working file after
1286 `git merge` returns control back to you, leaving the conflicting
1287 merge for you to resolve. Notice that the path `hello` is still
1288 unmerged, and what you see with `git diff` at this point is
1289 differences since stage 2 (i.e. your version).
1290
1291
1292 Publishing your work
1293 --------------------
1294
1295 So, we can use somebody else's work from a remote repository, but
1296 how can *you* prepare a repository to let other people pull from
1297 it?
1298
1299 You do your real work in your working tree that has your
1300 primary repository hanging under it as its `.git` subdirectory.
1301 You *could* make that repository accessible remotely and ask
1302 people to pull from it, but in practice that is not the way
1303 things are usually done. A recommended way is to have a public
1304 repository, make it reachable by other people, and when the
1305 changes you made in your primary working tree are in good shape,
1306 update the public repository from it. This is often called
1307 'pushing'.
1308
1309 [NOTE]
1310 This public repository could further be mirrored, and that is
1311 how git repositories at `kernel.org` are managed.
1312
1313 Publishing the changes from your local (private) repository to
1314 your remote (public) repository requires a write privilege on
1315 the remote machine. You need to have an SSH account there to
1316 run a single command, `git-receive-pack`.
1317
1318 First, you need to create an empty repository on the remote
1319 machine that will house your public repository. This empty
1320 repository will be populated and be kept up-to-date by pushing
1321 into it later. Obviously, this repository creation needs to be
1322 done only once.
1323
1324 [NOTE]
1325 `git push` uses a pair of programs,
1326 `git-send-pack` on your local machine, and `git-receive-pack`
1327 on the remote machine. The communication between the two over
1328 the network internally uses an SSH connection.
1329
1330 Your private repository's git directory is usually `.git`, but
1331 your public repository is often named after the project name,
1332 i.e. `<project>.git`. Let's create such a public repository for
1333 project `my-git`. After logging into the remote machine, create
1334 an empty directory:
1335
1336 ------------
1337 $ mkdir my-git.git
1338 ------------
1339
1340 Then, make that directory into a git repository by running
1341 `git init`, but this time, since its name is not the usual
1342 `.git`, we do things slightly differently:
1343
1344 ------------
1345 $ GIT_DIR=my-git.git git-init
1346 ------------
1347
1348 Make sure this directory is available for others you want your
1349 changes to be pulled by via the transport of your choice. Also
1350 you need to make sure that you have the `git-receive-pack`
1351 program on the `$PATH`.
1352
1353 [NOTE]
1354 Many installations of sshd do not invoke your shell as the login
1355 shell when you directly run programs; what this means is that if
1356 your login shell is `bash`, only `.bashrc` is read and not
1357 `.bash_profile`. As a workaround, make sure `.bashrc` sets up
1358 `$PATH` so that you can run `git-receive-pack` program.
1359
1360 [NOTE]
1361 If you plan to publish this repository to be accessed over http,
1362 you should do `chmod +x my-git.git/hooks/post-update` at this
1363 point. This makes sure that every time you push into this
1364 repository, `git-update-server-info` is run.
1365
1366 Your "public repository" is now ready to accept your changes.
1367 Come back to the machine you have your private repository. From
1368 there, run this command:
1369
1370 ------------
1371 $ git push <public-host>:/path/to/my-git.git master
1372 ------------
1373
1374 This synchronizes your public repository to match the named
1375 branch head (i.e. `master` in this case) and objects reachable
1376 from them in your current repository.
1377
1378 As a real example, this is how I update my public git
1379 repository. Kernel.org mirror network takes care of the
1380 propagation to other publicly visible machines:
1381
1382 ------------
1383 $ git push master.kernel.org:/pub/scm/git/git.git/
1384 ------------
1385
1386
1387 Packing your repository
1388 -----------------------
1389
1390 Earlier, we saw that one file under `.git/objects/??/` directory
1391 is stored for each git object you create. This representation
1392 is efficient to create atomically and safely, but
1393 not so convenient to transport over the network. Since git objects are
1394 immutable once they are created, there is a way to optimize the
1395 storage by "packing them together". The command
1396
1397 ------------
1398 $ git repack
1399 ------------
1400
1401 will do it for you. If you followed the tutorial examples, you
1402 would have accumulated about 17 objects in `.git/objects/??/`
1403 directories by now. `git repack` tells you how many objects it
1404 packed, and stores the packed file in `.git/objects/pack`
1405 directory.
1406
1407 [NOTE]
1408 You will see two files, `pack-\*.pack` and `pack-\*.idx`,
1409 in `.git/objects/pack` directory. They are closely related to
1410 each other, and if you ever copy them by hand to a different
1411 repository for whatever reason, you should make sure you copy
1412 them together. The former holds all the data from the objects
1413 in the pack, and the latter holds the index for random
1414 access.
1415
1416 If you are paranoid, running `git-verify-pack` command would
1417 detect if you have a corrupt pack, but do not worry too much.
1418 Our programs are always perfect ;-).
1419
1420 Once you have packed objects, you do not need to leave the
1421 unpacked objects that are contained in the pack file anymore.
1422
1423 ------------
1424 $ git prune-packed
1425 ------------
1426
1427 would remove them for you.
1428
1429 You can try running `find .git/objects -type f` before and after
1430 you run `git prune-packed` if you are curious. Also `git
1431 count-objects` would tell you how many unpacked objects are in
1432 your repository and how much space they are consuming.
1433
1434 [NOTE]
1435 `git pull` is slightly cumbersome for HTTP transport, as a
1436 packed repository may contain relatively few objects in a
1437 relatively large pack. If you expect many HTTP pulls from your
1438 public repository you might want to repack & prune often, or
1439 never.
1440
1441 If you run `git repack` again at this point, it will say
1442 "Nothing to pack". Once you continue your development and
1443 accumulate the changes, running `git repack` again will create a
1444 new pack, that contains objects created since you packed your
1445 repository the last time. We recommend that you pack your project
1446 soon after the initial import (unless you are starting your
1447 project from scratch), and then run `git repack` every once in a
1448 while, depending on how active your project is.
1449
1450 When a repository is synchronized via `git push` and `git pull`
1451 objects packed in the source repository are usually stored
1452 unpacked in the destination, unless rsync transport is used.
1453 While this allows you to use different packing strategies on
1454 both ends, it also means you may need to repack both
1455 repositories every once in a while.
1456
1457
1458 Working with Others
1459 -------------------
1460
1461 Although git is a truly distributed system, it is often
1462 convenient to organize your project with an informal hierarchy
1463 of developers. Linux kernel development is run this way. There
1464 is a nice illustration (page 17, "Merges to Mainline") in
1465 link:http://www.xenotime.net/linux/mentor/linux-mentoring-2006.pdf[Randy Dunlap's presentation].
1466
1467 It should be stressed that this hierarchy is purely *informal*.
1468 There is nothing fundamental in git that enforces the "chain of
1469 patch flow" this hierarchy implies. You do not have to pull
1470 from only one remote repository.
1471
1472 A recommended workflow for a "project lead" goes like this:
1473
1474 1. Prepare your primary repository on your local machine. Your
1475 work is done there.
1476
1477 2. Prepare a public repository accessible to others.
1478 +
1479 If other people are pulling from your repository over dumb
1480 transport protocols (HTTP), you need to keep this repository
1481 'dumb transport friendly'. After `git init`,
1482 `$GIT_DIR/hooks/post-update` copied from the standard templates
1483 would contain a call to `git-update-server-info` but the
1484 `post-update` hook itself is disabled by default -- enable it
1485 with `chmod +x post-update`. This makes sure `git-update-server-info`
1486 keeps the necessary files up-to-date.
1487
1488 3. Push into the public repository from your primary
1489 repository.
1490
1491 4. `git repack` the public repository. This establishes a big
1492 pack that contains the initial set of objects as the
1493 baseline, and possibly `git prune` if the transport
1494 used for pulling from your repository supports packed
1495 repositories.
1496
1497 5. Keep working in your primary repository. Your changes
1498 include modifications of your own, patches you receive via
1499 e-mails, and merges resulting from pulling the "public"
1500 repositories of your "subsystem maintainers".
1501 +
1502 You can repack this private repository whenever you feel like.
1503
1504 6. Push your changes to the public repository, and announce it
1505 to the public.
1506
1507 7. Every once in a while, "git repack" the public repository.
1508 Go back to step 5. and continue working.
1509
1510
1511 A recommended work cycle for a "subsystem maintainer" who works
1512 on that project and has an own "public repository" goes like this:
1513
1514 1. Prepare your work repository, by `git clone` the public
1515 repository of the "project lead". The URL used for the
1516 initial cloning is stored in the remote.origin.url
1517 configuration variable.
1518
1519 2. Prepare a public repository accessible to others, just like
1520 the "project lead" person does.
1521
1522 3. Copy over the packed files from "project lead" public
1523 repository to your public repository, unless the "project
1524 lead" repository lives on the same machine as yours. In the
1525 latter case, you can use `objects/info/alternates` file to
1526 point at the repository you are borrowing from.
1527
1528 4. Push into the public repository from your primary
1529 repository. Run `git repack`, and possibly `git prune` if the
1530 transport used for pulling from your repository supports
1531 packed repositories.
1532
1533 5. Keep working in your primary repository. Your changes
1534 include modifications of your own, patches you receive via
1535 e-mails, and merges resulting from pulling the "public"
1536 repositories of your "project lead" and possibly your
1537 "sub-subsystem maintainers".
1538 +
1539 You can repack this private repository whenever you feel
1540 like.
1541
1542 6. Push your changes to your public repository, and ask your
1543 "project lead" and possibly your "sub-subsystem
1544 maintainers" to pull from it.
1545
1546 7. Every once in a while, `git repack` the public repository.
1547 Go back to step 5. and continue working.
1548
1549
1550 A recommended work cycle for an "individual developer" who does
1551 not have a "public" repository is somewhat different. It goes
1552 like this:
1553
1554 1. Prepare your work repository, by `git clone` the public
1555 repository of the "project lead" (or a "subsystem
1556 maintainer", if you work on a subsystem). The URL used for
1557 the initial cloning is stored in the remote.origin.url
1558 configuration variable.
1559
1560 2. Do your work in your repository on 'master' branch.
1561
1562 3. Run `git fetch origin` from the public repository of your
1563 upstream every once in a while. This does only the first
1564 half of `git pull` but does not merge. The head of the
1565 public repository is stored in `.git/refs/remotes/origin/master`.
1566
1567 4. Use `git cherry origin` to see which ones of your patches
1568 were accepted, and/or use `git rebase origin` to port your
1569 unmerged changes forward to the updated upstream.
1570
1571 5. Use `git format-patch origin` to prepare patches for e-mail
1572 submission to your upstream and send it out. Go back to
1573 step 2. and continue.
1574
1575
1576 Working with Others, Shared Repository Style
1577 --------------------------------------------
1578
1579 If you are coming from CVS background, the style of cooperation
1580 suggested in the previous section may be new to you. You do not
1581 have to worry. git supports "shared public repository" style of
1582 cooperation you are probably more familiar with as well.
1583
1584 See link:cvs-migration.html[git for CVS users] for the details.
1585
1586 Bundling your work together
1587 ---------------------------
1588
1589 It is likely that you will be working on more than one thing at
1590 a time. It is easy to manage those more-or-less independent tasks
1591 using branches with git.
1592
1593 We have already seen how branches work previously,
1594 with "fun and work" example using two branches. The idea is the
1595 same if there are more than two branches. Let's say you started
1596 out from "master" head, and have some new code in the "master"
1597 branch, and two independent fixes in the "commit-fix" and
1598 "diff-fix" branches:
1599
1600 ------------
1601 $ git show-branch
1602 ! [commit-fix] Fix commit message normalization.
1603 ! [diff-fix] Fix rename detection.
1604 * [master] Release candidate #1
1605 ---
1606 + [diff-fix] Fix rename detection.
1607 + [diff-fix~1] Better common substring algorithm.
1608 + [commit-fix] Fix commit message normalization.
1609 * [master] Release candidate #1
1610 ++* [diff-fix~2] Pretty-print messages.
1611 ------------
1612
1613 Both fixes are tested well, and at this point, you want to merge
1614 in both of them. You could merge in 'diff-fix' first and then
1615 'commit-fix' next, like this:
1616
1617 ------------
1618 $ git merge -m "Merge fix in diff-fix" diff-fix
1619 $ git merge -m "Merge fix in commit-fix" commit-fix
1620 ------------
1621
1622 Which would result in:
1623
1624 ------------
1625 $ git show-branch
1626 ! [commit-fix] Fix commit message normalization.
1627 ! [diff-fix] Fix rename detection.
1628 * [master] Merge fix in commit-fix
1629 ---
1630 - [master] Merge fix in commit-fix
1631 + * [commit-fix] Fix commit message normalization.
1632 - [master~1] Merge fix in diff-fix
1633 +* [diff-fix] Fix rename detection.
1634 +* [diff-fix~1] Better common substring algorithm.
1635 * [master~2] Release candidate #1
1636 ++* [master~3] Pretty-print messages.
1637 ------------
1638
1639 However, there is no particular reason to merge in one branch
1640 first and the other next, when what you have are a set of truly
1641 independent changes (if the order mattered, then they are not
1642 independent by definition). You could instead merge those two
1643 branches into the current branch at once. First let's undo what
1644 we just did and start over. We would want to get the master
1645 branch before these two merges by resetting it to 'master~2':
1646
1647 ------------
1648 $ git reset --hard master~2
1649 ------------
1650
1651 You can make sure 'git show-branch' matches the state before
1652 those two 'git merge' you just did. Then, instead of running
1653 two 'git merge' commands in a row, you would merge these two
1654 branch heads (this is known as 'making an Octopus'):
1655
1656 ------------
1657 $ git merge commit-fix diff-fix
1658 $ git show-branch
1659 ! [commit-fix] Fix commit message normalization.
1660 ! [diff-fix] Fix rename detection.
1661 * [master] Octopus merge of branches 'diff-fix' and 'commit-fix'
1662 ---
1663 - [master] Octopus merge of branches 'diff-fix' and 'commit-fix'
1664 + * [commit-fix] Fix commit message normalization.
1665 +* [diff-fix] Fix rename detection.
1666 +* [diff-fix~1] Better common substring algorithm.
1667 * [master~1] Release candidate #1
1668 ++* [master~2] Pretty-print messages.
1669 ------------
1670
1671 Note that you should not do Octopus because you can. An octopus
1672 is a valid thing to do and often makes it easier to view the
1673 commit history if you are merging more than two independent
1674 changes at the same time. However, if you have merge conflicts
1675 with any of the branches you are merging in and need to hand
1676 resolve, that is an indication that the development happened in
1677 those branches were not independent after all, and you should
1678 merge two at a time, documenting how you resolved the conflicts,
1679 and the reason why you preferred changes made in one side over
1680 the other. Otherwise it would make the project history harder
1681 to follow, not easier.