Documentation: HTTP needs update-server-info.
[git/git.git] / Documentation / tutorial.txt
1 A short git tutorial
2 ====================
3
4 Introduction
5 ------------
6
7 This is trying to be a short tutorial on setting up and using a git
8 repository, mainly because being hands-on and using explicit examples is
9 often the best way of explaining what is going on.
10
11 In normal life, most people wouldn't use the "core" git programs
12 directly, but rather script around them to make them more palatable.
13 Understanding the core git stuff may help some people get those scripts
14 done, though, and it may also be instructive in helping people
15 understand what it is that the higher-level helper scripts are actually
16 doing.
17
18 The core git is often called "plumbing", with the prettier user
19 interfaces on top of it called "porcelain". You may not want to use the
20 plumbing directly very often, but it can be good to know what the
21 plumbing does for when the porcelain isn't flushing.
22
23 The material presented here often goes deep describing how things
24 work internally. If you are mostly interested in using git as a
25 SCM, you can skip them during your first pass.
26
27 [NOTE]
28 And those "too deep" descriptions are often marked as Note.
29
30 [NOTE]
31 If you are already familiar with another version control system,
32 like CVS, you may want to take a look at
33 link:everyday.html[Everyday GIT in 20 commands or so] first
34 before reading this.
35
36
37 Creating a git repository
38 -------------------------
39
40 Creating a new git repository couldn't be easier: all git repositories start
41 out empty, and the only thing you need to do is find yourself a
42 subdirectory that you want to use as a working tree - either an empty
43 one for a totally new project, or an existing working tree that you want
44 to import into git.
45
46 For our first example, we're going to start a totally new repository from
47 scratch, with no pre-existing files, and we'll call it `git-tutorial`.
48 To start up, create a subdirectory for it, change into that
49 subdirectory, and initialize the git infrastructure with `git-init-db`:
50
51 ------------------------------------------------
52 $ mkdir git-tutorial
53 $ cd git-tutorial
54 $ git-init-db
55 ------------------------------------------------
56
57 to which git will reply
58
59 ----------------
60 defaulting to local storage area
61 ----------------
62
63 which is just git's way of saying that you haven't been doing anything
64 strange, and that it will have created a local `.git` directory setup for
65 your new project. You will now have a `.git` directory, and you can
66 inspect that with `ls`. For your new empty project, it should show you
67 three entries, among other things:
68
69 - a symlink called `HEAD`, pointing to `refs/heads/master` (if your
70 platform does not have native symlinks, it is a file containing the
71 line "ref: refs/heads/master")
72 +
73 Don't worry about the fact that the file that the `HEAD` link points to
74 doesn't even exist yet -- you haven't created the commit that will
75 start your `HEAD` development branch yet.
76
77 - a subdirectory called `objects`, which will contain all the
78 objects of your project. You should never have any real reason to
79 look at the objects directly, but you might want to know that these
80 objects are what contains all the real 'data' in your repository.
81
82 - a subdirectory called `refs`, which contains references to objects.
83
84 In particular, the `refs` subdirectory will contain two other
85 subdirectories, named `heads` and `tags` respectively. They do
86 exactly what their names imply: they contain references to any number
87 of different 'heads' of development (aka 'branches'), and to any
88 'tags' that you have created to name specific versions in your
89 repository.
90
91 One note: the special `master` head is the default branch, which is
92 why the `.git/HEAD` file was created as a symlink to it even if it
93 doesn't yet exist. Basically, the `HEAD` link is supposed to always
94 point to the branch you are working on right now, and you always
95 start out expecting to work on the `master` branch.
96
97 However, this is only a convention, and you can name your branches
98 anything you want, and don't have to ever even 'have' a `master`
99 branch. A number of the git tools will assume that `.git/HEAD` is
100 valid, though.
101
102 [NOTE]
103 An 'object' is identified by its 160-bit SHA1 hash, aka 'object name',
104 and a reference to an object is always the 40-byte hex
105 representation of that SHA1 name. The files in the `refs`
106 subdirectory are expected to contain these hex references
107 (usually with a final `\'\n\'` at the end), and you should thus
108 expect to see a number of 41-byte files containing these
109 references in these `refs` subdirectories when you actually start
110 populating your tree.
111
112 [NOTE]
113 An advanced user may want to take a look at the
114 link:repository-layout.html[repository layout] document
115 after finishing this tutorial.
116
117 You have now created your first git repository. Of course, since it's
118 empty, that's not very useful, so let's start populating it with data.
119
120
121 Populating a git repository
122 ---------------------------
123
124 We'll keep this simple and stupid, so we'll start off with populating a
125 few trivial files just to get a feel for it.
126
127 Start off with just creating any random files that you want to maintain
128 in your git repository. We'll start off with a few bad examples, just to
129 get a feel for how this works:
130
131 ------------------------------------------------
132 $ echo "Hello World" >hello
133 $ echo "Silly example" >example
134 ------------------------------------------------
135
136 you have now created two files in your working tree (aka 'working directory'), but to
137 actually check in your hard work, you will have to go through two steps:
138
139 - fill in the 'index' file (aka 'cache') with the information about your
140 working tree state.
141
142 - commit that index file as an object.
143
144 The first step is trivial: when you want to tell git about any changes
145 to your working tree, you use the `git-update-index` program. That
146 program normally just takes a list of filenames you want to update, but
147 to avoid trivial mistakes, it refuses to add new entries to the index
148 (or remove existing ones) unless you explicitly tell it that you're
149 adding a new entry with the `\--add` flag (or removing an entry with the
150 `\--remove`) flag.
151
152 So to populate the index with the two files you just created, you can do
153
154 ------------------------------------------------
155 $ git-update-index --add hello example
156 ------------------------------------------------
157
158 and you have now told git to track those two files.
159
160 In fact, as you did that, if you now look into your object directory,
161 you'll notice that git will have added two new objects to the object
162 database. If you did exactly the steps above, you should now be able to do
163
164
165 ----------------
166 $ ls .git/objects/??/*
167 ----------------
168
169 and see two files:
170
171 ----------------
172 .git/objects/55/7db03de997c86a4a028e1ebd3a1ceb225be238
173 .git/objects/f2/4c74a2e500f5ee1332c86b94199f52b1d1d962
174 ----------------
175
176 which correspond with the objects with names of 557db... and f24c7..
177 respectively.
178
179 If you want to, you can use `git-cat-file` to look at those objects, but
180 you'll have to use the object name, not the filename of the object:
181
182 ----------------
183 $ git-cat-file -t 557db03de997c86a4a028e1ebd3a1ceb225be238
184 ----------------
185
186 where the `-t` tells `git-cat-file` to tell you what the "type" of the
187 object is. git will tell you that you have a "blob" object (ie just a
188 regular file), and you can see the contents with
189
190 ----------------
191 $ git-cat-file "blob" 557db03
192 ----------------
193
194 which will print out "Hello World". The object 557db03 is nothing
195 more than the contents of your file `hello`.
196
197 [NOTE]
198 Don't confuse that object with the file `hello` itself. The
199 object is literally just those specific *contents* of the file, and
200 however much you later change the contents in file `hello`, the object
201 we just looked at will never change. Objects are immutable.
202
203 [NOTE]
204 The second example demonstrates that you can
205 abbreviate the object name to only the first several
206 hexadecimal digits in most places.
207
208 Anyway, as we mentioned previously, you normally never actually take a
209 look at the objects themselves, and typing long 40-character hex
210 names is not something you'd normally want to do. The above digression
211 was just to show that `git-update-index` did something magical, and
212 actually saved away the contents of your files into the git object
213 database.
214
215 Updating the index did something else too: it created a `.git/index`
216 file. This is the index that describes your current working tree, and
217 something you should be very aware of. Again, you normally never worry
218 about the index file itself, but you should be aware of the fact that
219 you have not actually really "checked in" your files into git so far,
220 you've only *told* git about them.
221
222 However, since git knows about them, you can now start using some of the
223 most basic git commands to manipulate the files or look at their status.
224
225 In particular, let's not even check in the two files into git yet, we'll
226 start off by adding another line to `hello` first:
227
228 ------------------------------------------------
229 $ echo "It's a new day for git" >>hello
230 ------------------------------------------------
231
232 and you can now, since you told git about the previous state of `hello`, ask
233 git what has changed in the tree compared to your old index, using the
234 `git-diff-files` command:
235
236 ------------
237 $ git-diff-files
238 ------------
239
240 Oops. That wasn't very readable. It just spit out its own internal
241 version of a `diff`, but that internal version really just tells you
242 that it has noticed that "hello" has been modified, and that the old object
243 contents it had have been replaced with something else.
244
245 To make it readable, we can tell git-diff-files to output the
246 differences as a patch, using the `-p` flag:
247
248 ------------
249 $ git-diff-files -p
250 diff --git a/hello b/hello
251 index 557db03..263414f 100644
252 --- a/hello
253 +++ b/hello
254 @@ -1 +1,2 @@
255 Hello World
256 +It's a new day for git
257 ----
258
259 i.e. the diff of the change we caused by adding another line to `hello`.
260
261 In other words, `git-diff-files` always shows us the difference between
262 what is recorded in the index, and what is currently in the working
263 tree. That's very useful.
264
265 A common shorthand for `git-diff-files -p` is to just write `git
266 diff`, which will do the same thing.
267
268 ------------
269 $ git diff
270 diff --git a/hello b/hello
271 index 557db03..263414f 100644
272 --- a/hello
273 +++ b/hello
274 @@ -1 +1,2 @@
275 Hello World
276 +It's a new day for git
277 ------------
278
279
280 Committing git state
281 --------------------
282
283 Now, we want to go to the next stage in git, which is to take the files
284 that git knows about in the index, and commit them as a real tree. We do
285 that in two phases: creating a 'tree' object, and committing that 'tree'
286 object as a 'commit' object together with an explanation of what the
287 tree was all about, along with information of how we came to that state.
288
289 Creating a tree object is trivial, and is done with `git-write-tree`.
290 There are no options or other input: git-write-tree will take the
291 current index state, and write an object that describes that whole
292 index. In other words, we're now tying together all the different
293 filenames with their contents (and their permissions), and we're
294 creating the equivalent of a git "directory" object:
295
296 ------------------------------------------------
297 $ git-write-tree
298 ------------------------------------------------
299
300 and this will just output the name of the resulting tree, in this case
301 (if you have done exactly as I've described) it should be
302
303 ----------------
304 8988da15d077d4829fc51d8544c097def6644dbb
305 ----------------
306
307 which is another incomprehensible object name. Again, if you want to,
308 you can use `git-cat-file -t 8988d\...` to see that this time the object
309 is not a "blob" object, but a "tree" object (you can also use
310 `git-cat-file` to actually output the raw object contents, but you'll see
311 mainly a binary mess, so that's less interesting).
312
313 However -- normally you'd never use `git-write-tree` on its own, because
314 normally you always commit a tree into a commit object using the
315 `git-commit-tree` command. In fact, it's easier to not actually use
316 `git-write-tree` on its own at all, but to just pass its result in as an
317 argument to `git-commit-tree`.
318
319 `git-commit-tree` normally takes several arguments -- it wants to know
320 what the 'parent' of a commit was, but since this is the first commit
321 ever in this new repository, and it has no parents, we only need to pass in
322 the object name of the tree. However, `git-commit-tree`
323 also wants to get a commit message
324 on its standard input, and it will write out the resulting object name for the
325 commit to its standard output.
326
327 And this is where we create the `.git/refs/heads/master` file
328 which is pointed at by `HEAD`. This file is supposed to contain
329 the reference to the top-of-tree of the master branch, and since
330 that's exactly what `git-commit-tree` spits out, we can do this
331 all with a sequence of simple shell commands:
332
333 ------------------------------------------------
334 $ tree=$(git-write-tree)
335 $ commit=$(echo 'Initial commit' | git-commit-tree $tree)
336 $ git-update-ref HEAD $commit
337 ------------------------------------------------
338
339 which will say:
340
341 ----------------
342 Committing initial tree 8988da15d077d4829fc51d8544c097def6644dbb
343 ----------------
344
345 just to warn you about the fact that it created a totally new commit
346 that is not related to anything else. Normally you do this only *once*
347 for a project ever, and all later commits will be parented on top of an
348 earlier commit, and you'll never see this "Committing initial tree"
349 message ever again.
350
351 Again, normally you'd never actually do this by hand. There is a
352 helpful script called `git commit` that will do all of this for you. So
353 you could have just written `git commit`
354 instead, and it would have done the above magic scripting for you.
355
356
357 Making a change
358 ---------------
359
360 Remember how we did the `git-update-index` on file `hello` and then we
361 changed `hello` afterward, and could compare the new state of `hello` with the
362 state we saved in the index file?
363
364 Further, remember how I said that `git-write-tree` writes the contents
365 of the *index* file to the tree, and thus what we just committed was in
366 fact the *original* contents of the file `hello`, not the new ones. We did
367 that on purpose, to show the difference between the index state, and the
368 state in the working tree, and how they don't have to match, even
369 when we commit things.
370
371 As before, if we do `git-diff-files -p` in our git-tutorial project,
372 we'll still see the same difference we saw last time: the index file
373 hasn't changed by the act of committing anything. However, now that we
374 have committed something, we can also learn to use a new command:
375 `git-diff-index`.
376
377 Unlike `git-diff-files`, which showed the difference between the index
378 file and the working tree, `git-diff-index` shows the differences
379 between a committed *tree* and either the index file or the working
380 tree. In other words, `git-diff-index` wants a tree to be diffed
381 against, and before we did the commit, we couldn't do that, because we
382 didn't have anything to diff against.
383
384 But now we can do
385
386 ----------------
387 $ git-diff-index -p HEAD
388 ----------------
389
390 (where `-p` has the same meaning as it did in `git-diff-files`), and it
391 will show us the same difference, but for a totally different reason.
392 Now we're comparing the working tree not against the index file,
393 but against the tree we just wrote. It just so happens that those two
394 are obviously the same, so we get the same result.
395
396 Again, because this is a common operation, you can also just shorthand
397 it with
398
399 ----------------
400 $ git diff HEAD
401 ----------------
402
403 which ends up doing the above for you.
404
405 In other words, `git-diff-index` normally compares a tree against the
406 working tree, but when given the `\--cached` flag, it is told to
407 instead compare against just the index cache contents, and ignore the
408 current working tree state entirely. Since we just wrote the index
409 file to HEAD, doing `git-diff-index \--cached -p HEAD` should thus return
410 an empty set of differences, and that's exactly what it does.
411
412 [NOTE]
413 ================
414 `git-diff-index` really always uses the index for its
415 comparisons, and saying that it compares a tree against the working
416 tree is thus not strictly accurate. In particular, the list of
417 files to compare (the "meta-data") *always* comes from the index file,
418 regardless of whether the `\--cached` flag is used or not. The `\--cached`
419 flag really only determines whether the file *contents* to be compared
420 come from the working tree or not.
421
422 This is not hard to understand, as soon as you realize that git simply
423 never knows (or cares) about files that it is not told about
424 explicitly. git will never go *looking* for files to compare, it
425 expects you to tell it what the files are, and that's what the index
426 is there for.
427 ================
428
429 However, our next step is to commit the *change* we did, and again, to
430 understand what's going on, keep in mind the difference between "working
431 tree contents", "index file" and "committed tree". We have changes
432 in the working tree that we want to commit, and we always have to
433 work through the index file, so the first thing we need to do is to
434 update the index cache:
435
436 ------------------------------------------------
437 $ git-update-index hello
438 ------------------------------------------------
439
440 (note how we didn't need the `\--add` flag this time, since git knew
441 about the file already).
442
443 Note what happens to the different `git-diff-\*` versions here. After
444 we've updated `hello` in the index, `git-diff-files -p` now shows no
445 differences, but `git-diff-index -p HEAD` still *does* show that the
446 current state is different from the state we committed. In fact, now
447 `git-diff-index` shows the same difference whether we use the `--cached`
448 flag or not, since now the index is coherent with the working tree.
449
450 Now, since we've updated `hello` in the index, we can commit the new
451 version. We could do it by writing the tree by hand again, and
452 committing the tree (this time we'd have to use the `-p HEAD` flag to
453 tell commit that the HEAD was the *parent* of the new commit, and that
454 this wasn't an initial commit any more), but you've done that once
455 already, so let's just use the helpful script this time:
456
457 ------------------------------------------------
458 $ git commit
459 ------------------------------------------------
460
461 which starts an editor for you to write the commit message and tells you
462 a bit about what you have done.
463
464 Write whatever message you want, and all the lines that start with '#'
465 will be pruned out, and the rest will be used as the commit message for
466 the change. If you decide you don't want to commit anything after all at
467 this point (you can continue to edit things and update the index), you
468 can just leave an empty message. Otherwise `git commit` will commit
469 the change for you.
470
471 You've now made your first real git commit. And if you're interested in
472 looking at what `git commit` really does, feel free to investigate:
473 it's a few very simple shell scripts to generate the helpful (?) commit
474 message headers, and a few one-liners that actually do the
475 commit itself (`git-commit`).
476
477
478 Inspecting Changes
479 ------------------
480
481 While creating changes is useful, it's even more useful if you can tell
482 later what changed. The most useful command for this is another of the
483 `diff` family, namely `git-diff-tree`.
484
485 `git-diff-tree` can be given two arbitrary trees, and it will tell you the
486 differences between them. Perhaps even more commonly, though, you can
487 give it just a single commit object, and it will figure out the parent
488 of that commit itself, and show the difference directly. Thus, to get
489 the same diff that we've already seen several times, we can now do
490
491 ----------------
492 $ git-diff-tree -p HEAD
493 ----------------
494
495 (again, `-p` means to show the difference as a human-readable patch),
496 and it will show what the last commit (in `HEAD`) actually changed.
497
498 [NOTE]
499 ============
500 Here is an ASCII art by Jon Loeliger that illustrates how
501 various diff-\* commands compare things.
502
503 diff-tree
504 +----+
505 | |
506 | |
507 V V
508 +-----------+
509 | Object DB |
510 | Backing |
511 | Store |
512 +-----------+
513 ^ ^
514 | |
515 | | diff-index --cached
516 | |
517 diff-index | V
518 | +-----------+
519 | | Index |
520 | | "cache" |
521 | +-----------+
522 | ^
523 | |
524 | | diff-files
525 | |
526 V V
527 +-----------+
528 | Working |
529 | Directory |
530 +-----------+
531 ============
532
533 More interestingly, you can also give `git-diff-tree` the `-v` flag, which
534 tells it to also show the commit message and author and date of the
535 commit, and you can tell it to show a whole series of diffs.
536 Alternatively, you can tell it to be "silent", and not show the diffs at
537 all, but just show the actual commit message.
538
539 In fact, together with the `git-rev-list` program (which generates a
540 list of revisions), `git-diff-tree` ends up being a veritable fount of
541 changes. A trivial (but very useful) script called `git-whatchanged` is
542 included with git which does exactly this, and shows a log of recent
543 activities.
544
545 To see the whole history of our pitiful little git-tutorial project, you
546 can do
547
548 ----------------
549 $ git log
550 ----------------
551
552 which shows just the log messages, or if we want to see the log together
553 with the associated patches use the more complex (and much more
554 powerful)
555
556 ----------------
557 $ git-whatchanged -p --root
558 ----------------
559
560 and you will see exactly what has changed in the repository over its
561 short history.
562
563 [NOTE]
564 The `\--root` flag is a flag to `git-diff-tree` to tell it to
565 show the initial aka 'root' commit too. Normally you'd probably not
566 want to see the initial import diff, but since the tutorial project
567 was started from scratch and is so small, we use it to make the result
568 a bit more interesting.
569
570 With that, you should now be having some inkling of what git does, and
571 can explore on your own.
572
573 [NOTE]
574 Most likely, you are not directly using the core
575 git Plumbing commands, but using Porcelain like Cogito on top
576 of it. Cogito works a bit differently and you usually do not
577 have to run `git-update-index` yourself for changed files (you
578 do tell underlying git about additions and removals via
579 `cg-add` and `cg-rm` commands). Just before you make a commit
580 with `cg-commit`, Cogito figures out which files you modified,
581 and runs `git-update-index` on them for you.
582
583
584 Tagging a version
585 -----------------
586
587 In git, there are two kinds of tags, a "light" one, and an "annotated tag".
588
589 A "light" tag is technically nothing more than a branch, except we put
590 it in the `.git/refs/tags/` subdirectory instead of calling it a `head`.
591 So the simplest form of tag involves nothing more than
592
593 ------------------------------------------------
594 $ git tag my-first-tag
595 ------------------------------------------------
596
597 which just writes the current `HEAD` into the `.git/refs/tags/my-first-tag`
598 file, after which point you can then use this symbolic name for that
599 particular state. You can, for example, do
600
601 ----------------
602 $ git diff my-first-tag
603 ----------------
604
605 to diff your current state against that tag (which at this point will
606 obviously be an empty diff, but if you continue to develop and commit
607 stuff, you can use your tag as an "anchor-point" to see what has changed
608 since you tagged it.
609
610 An "annotated tag" is actually a real git object, and contains not only a
611 pointer to the state you want to tag, but also a small tag name and
612 message, along with optionally a PGP signature that says that yes,
613 you really did
614 that tag. You create these annotated tags with either the `-a` or
615 `-s` flag to `git tag`:
616
617 ----------------
618 $ git tag -s <tagname>
619 ----------------
620
621 which will sign the current `HEAD` (but you can also give it another
622 argument that specifies the thing to tag, ie you could have tagged the
623 current `mybranch` point by using `git tag <tagname> mybranch`).
624
625 You normally only do signed tags for major releases or things
626 like that, while the light-weight tags are useful for any marking you
627 want to do -- any time you decide that you want to remember a certain
628 point, just create a private tag for it, and you have a nice symbolic
629 name for the state at that point.
630
631
632 Copying repositories
633 --------------------
634
635 git repositories are normally totally self-sufficient and relocatable
636 Unlike CVS, for example, there is no separate notion of
637 "repository" and "working tree". A git repository normally *is* the
638 working tree, with the local git information hidden in the `.git`
639 subdirectory. There is nothing else. What you see is what you got.
640
641 [NOTE]
642 You can tell git to split the git internal information from
643 the directory that it tracks, but we'll ignore that for now: it's not
644 how normal projects work, and it's really only meant for special uses.
645 So the mental model of "the git information is always tied directly to
646 the working tree that it describes" may not be technically 100%
647 accurate, but it's a good model for all normal use.
648
649 This has two implications:
650
651 - if you grow bored with the tutorial repository you created (or you've
652 made a mistake and want to start all over), you can just do simple
653 +
654 ----------------
655 $ rm -rf git-tutorial
656 ----------------
657 +
658 and it will be gone. There's no external repository, and there's no
659 history outside the project you created.
660
661 - if you want to move or duplicate a git repository, you can do so. There
662 is `git clone` command, but if all you want to do is just to
663 create a copy of your repository (with all the full history that
664 went along with it), you can do so with a regular
665 `cp -a git-tutorial new-git-tutorial`.
666 +
667 Note that when you've moved or copied a git repository, your git index
668 file (which caches various information, notably some of the "stat"
669 information for the files involved) will likely need to be refreshed.
670 So after you do a `cp -a` to create a new copy, you'll want to do
671 +
672 ----------------
673 $ git-update-index --refresh
674 ----------------
675 +
676 in the new repository to make sure that the index file is up-to-date.
677
678 Note that the second point is true even across machines. You can
679 duplicate a remote git repository with *any* regular copy mechanism, be it
680 `scp`, `rsync` or `wget`.
681
682 When copying a remote repository, you'll want to at a minimum update the
683 index cache when you do this, and especially with other peoples'
684 repositories you often want to make sure that the index cache is in some
685 known state (you don't know *what* they've done and not yet checked in),
686 so usually you'll precede the `git-update-index` with a
687
688 ----------------
689 $ git-read-tree --reset HEAD
690 $ git-update-index --refresh
691 ----------------
692
693 which will force a total index re-build from the tree pointed to by `HEAD`.
694 It resets the index contents to `HEAD`, and then the `git-update-index`
695 makes sure to match up all index entries with the checked-out files.
696 If the original repository had uncommitted changes in its
697 working tree, `git-update-index --refresh` notices them and
698 tells you they need to be updated.
699
700 The above can also be written as simply
701
702 ----------------
703 $ git reset
704 ----------------
705
706 and in fact a lot of the common git command combinations can be scripted
707 with the `git xyz` interfaces. You can learn things by just looking
708 at what the various git scripts do. For example, `git reset` is the
709 above two lines implemented in `git-reset`, but some things like
710 `git status` and `git commit` are slightly more complex scripts around
711 the basic git commands.
712
713 Many (most?) public remote repositories will not contain any of
714 the checked out files or even an index file, and will *only* contain the
715 actual core git files. Such a repository usually doesn't even have the
716 `.git` subdirectory, but has all the git files directly in the
717 repository.
718
719 To create your own local live copy of such a "raw" git repository, you'd
720 first create your own subdirectory for the project, and then copy the
721 raw repository contents into the `.git` directory. For example, to
722 create your own copy of the git repository, you'd do the following
723
724 ----------------
725 $ mkdir my-git
726 $ cd my-git
727 $ rsync -rL rsync://rsync.kernel.org/pub/scm/git/git.git/ .git
728 ----------------
729
730 followed by
731
732 ----------------
733 $ git-read-tree HEAD
734 ----------------
735
736 to populate the index. However, now you have populated the index, and
737 you have all the git internal files, but you will notice that you don't
738 actually have any of the working tree files to work on. To get
739 those, you'd check them out with
740
741 ----------------
742 $ git-checkout-index -u -a
743 ----------------
744
745 where the `-u` flag means that you want the checkout to keep the index
746 up-to-date (so that you don't have to refresh it afterward), and the
747 `-a` flag means "check out all files" (if you have a stale copy or an
748 older version of a checked out tree you may also need to add the `-f`
749 flag first, to tell git-checkout-index to *force* overwriting of any old
750 files).
751
752 Again, this can all be simplified with
753
754 ----------------
755 $ git clone rsync://rsync.kernel.org/pub/scm/git/git.git/ my-git
756 $ cd my-git
757 $ git checkout
758 ----------------
759
760 which will end up doing all of the above for you.
761
762 You have now successfully copied somebody else's (mine) remote
763 repository, and checked it out.
764
765
766 Creating a new branch
767 ---------------------
768
769 Branches in git are really nothing more than pointers into the git
770 object database from within the `.git/refs/` subdirectory, and as we
771 already discussed, the `HEAD` branch is nothing but a symlink to one of
772 these object pointers.
773
774 You can at any time create a new branch by just picking an arbitrary
775 point in the project history, and just writing the SHA1 name of that
776 object into a file under `.git/refs/heads/`. You can use any filename you
777 want (and indeed, subdirectories), but the convention is that the
778 "normal" branch is called `master`. That's just a convention, though,
779 and nothing enforces it.
780
781 To show that as an example, let's go back to the git-tutorial repository we
782 used earlier, and create a branch in it. You do that by simply just
783 saying that you want to check out a new branch:
784
785 ------------
786 $ git checkout -b mybranch
787 ------------
788
789 will create a new branch based at the current `HEAD` position, and switch
790 to it.
791
792 [NOTE]
793 ================================================
794 If you make the decision to start your new branch at some
795 other point in the history than the current `HEAD`, you can do so by
796 just telling `git checkout` what the base of the checkout would be.
797 In other words, if you have an earlier tag or branch, you'd just do
798
799 ------------
800 $ git checkout -b mybranch earlier-commit
801 ------------
802
803 and it would create the new branch `mybranch` at the earlier commit,
804 and check out the state at that time.
805 ================================================
806
807 You can always just jump back to your original `master` branch by doing
808
809 ------------
810 $ git checkout master
811 ------------
812
813 (or any other branch-name, for that matter) and if you forget which
814 branch you happen to be on, a simple
815
816 ------------
817 $ ls -l .git/HEAD
818 ------------
819
820 will tell you where it's pointing (Note that on platforms with bad or no
821 symlink support, you have to execute
822
823 ------------
824 $ cat .git/HEAD
825 ------------
826
827 instead). To get the list of branches you have, you can say
828
829 ------------
830 $ git branch
831 ------------
832
833 which is nothing more than a simple script around `ls .git/refs/heads`.
834 There will be asterisk in front of the branch you are currently on.
835
836 Sometimes you may wish to create a new branch _without_ actually
837 checking it out and switching to it. If so, just use the command
838
839 ------------
840 $ git branch <branchname> [startingpoint]
841 ------------
842
843 which will simply _create_ the branch, but will not do anything further.
844 You can then later -- once you decide that you want to actually develop
845 on that branch -- switch to that branch with a regular `git checkout`
846 with the branchname as the argument.
847
848
849 Merging two branches
850 --------------------
851
852 One of the ideas of having a branch is that you do some (possibly
853 experimental) work in it, and eventually merge it back to the main
854 branch. So assuming you created the above `mybranch` that started out
855 being the same as the original `master` branch, let's make sure we're in
856 that branch, and do some work there.
857
858 ------------------------------------------------
859 $ git checkout mybranch
860 $ echo "Work, work, work" >>hello
861 $ git commit -m 'Some work.' hello
862 ------------------------------------------------
863
864 Here, we just added another line to `hello`, and we used a shorthand for
865 doing both `git-update-index hello` and `git commit` by just giving the
866 filename directly to `git commit`. The `-m` flag is to give the
867 commit log message from the command line.
868
869 Now, to make it a bit more interesting, let's assume that somebody else
870 does some work in the original branch, and simulate that by going back
871 to the master branch, and editing the same file differently there:
872
873 ------------
874 $ git checkout master
875 ------------
876
877 Here, take a moment to look at the contents of `hello`, and notice how they
878 don't contain the work we just did in `mybranch` -- because that work
879 hasn't happened in the `master` branch at all. Then do
880
881 ------------
882 $ echo "Play, play, play" >>hello
883 $ echo "Lots of fun" >>example
884 $ git commit -m 'Some fun.' hello example
885 ------------
886
887 since the master branch is obviously in a much better mood.
888
889 Now, you've got two branches, and you decide that you want to merge the
890 work done. Before we do that, let's introduce a cool graphical tool that
891 helps you view what's going on:
892
893 ----------------
894 $ gitk --all
895 ----------------
896
897 will show you graphically both of your branches (that's what the `\--all`
898 means: normally it will just show you your current `HEAD`) and their
899 histories. You can also see exactly how they came to be from a common
900 source.
901
902 Anyway, let's exit `gitk` (`^Q` or the File menu), and decide that we want
903 to merge the work we did on the `mybranch` branch into the `master`
904 branch (which is currently our `HEAD` too). To do that, there's a nice
905 script called `git merge`, which wants to know which branches you want
906 to resolve and what the merge is all about:
907
908 ------------
909 $ git merge "Merge work in mybranch" HEAD mybranch
910 ------------
911
912 where the first argument is going to be used as the commit message if
913 the merge can be resolved automatically.
914
915 Now, in this case we've intentionally created a situation where the
916 merge will need to be fixed up by hand, though, so git will do as much
917 of it as it can automatically (which in this case is just merge the `example`
918 file, which had no differences in the `mybranch` branch), and say:
919
920 ----------------
921 Trying really trivial in-index merge...
922 fatal: Merge requires file-level merging
923 Nope.
924 ...
925 Auto-merging hello
926 CONFLICT (content): Merge conflict in hello
927 Automatic merge failed/prevented; fix up by hand
928 ----------------
929
930 which is way too verbose, but it basically tells you that it failed the
931 really trivial merge ("Simple merge") and did an "Automatic merge"
932 instead, but that too failed due to conflicts in `hello`.
933
934 Not to worry. It left the (trivial) conflict in `hello` in the same form you
935 should already be well used to if you've ever used CVS, so let's just
936 open `hello` in our editor (whatever that may be), and fix it up somehow.
937 I'd suggest just making it so that `hello` contains all four lines:
938
939 ------------
940 Hello World
941 It's a new day for git
942 Play, play, play
943 Work, work, work
944 ------------
945
946 and once you're happy with your manual merge, just do a
947
948 ------------
949 $ git commit hello
950 ------------
951
952 which will very loudly warn you that you're now committing a merge
953 (which is correct, so never mind), and you can write a small merge
954 message about your adventures in git-merge-land.
955
956 After you're done, start up `gitk \--all` to see graphically what the
957 history looks like. Notice that `mybranch` still exists, and you can
958 switch to it, and continue to work with it if you want to. The
959 `mybranch` branch will not contain the merge, but next time you merge it
960 from the `master` branch, git will know how you merged it, so you'll not
961 have to do _that_ merge again.
962
963 Another useful tool, especially if you do not always work in X-Window
964 environment, is `git show-branch`.
965
966 ------------------------------------------------
967 $ git show-branch master mybranch
968 * [master] Merge work in mybranch
969 ! [mybranch] Some work.
970 --
971 + [master] Merge work in mybranch
972 ++ [mybranch] Some work.
973 ------------------------------------------------
974
975 The first two lines indicate that it is showing the two branches
976 and the first line of the commit log message from their
977 top-of-the-tree commits, you are currently on `master` branch
978 (notice the asterisk `*` character), and the first column for
979 the later output lines is used to show commits contained in the
980 `master` branch, and the second column for the `mybranch`
981 branch. Three commits are shown along with their log messages.
982 All of them have plus `+` characters in the first column, which
983 means they are now part of the `master` branch. Only the "Some
984 work" commit has the plus `+` character in the second column,
985 because `mybranch` has not been merged to incorporate these
986 commits from the master branch. The string inside brackets
987 before the commit log message is a short name you can use to
988 name the commit. In the above example, 'master' and 'mybranch'
989 are branch heads. 'master~1' is the first parent of 'master'
990 branch head. Please see 'git-rev-parse' documentation if you
991 see more complex cases.
992
993 Now, let's pretend you are the one who did all the work in
994 `mybranch`, and the fruit of your hard work has finally been merged
995 to the `master` branch. Let's go back to `mybranch`, and run
996 resolve to get the "upstream changes" back to your branch.
997
998 ------------
999 $ git checkout mybranch
1000 $ git merge "Merge upstream changes." HEAD master
1001 ------------
1002
1003 This outputs something like this (the actual commit object names
1004 would be different)
1005
1006 ----------------
1007 Updating from ae3a2da... to a80b4aa....
1008 example | 1 +
1009 hello | 1 +
1010 2 files changed, 2 insertions(+), 0 deletions(-)
1011 ----------------
1012
1013 Because your branch did not contain anything more than what are
1014 already merged into the `master` branch, the resolve operation did
1015 not actually do a merge. Instead, it just updated the top of
1016 the tree of your branch to that of the `master` branch. This is
1017 often called 'fast forward' merge.
1018
1019 You can run `gitk \--all` again to see how the commit ancestry
1020 looks like, or run `show-branch`, which tells you this.
1021
1022 ------------------------------------------------
1023 $ git show-branch master mybranch
1024 ! [master] Merge work in mybranch
1025 * [mybranch] Merge work in mybranch
1026 --
1027 ++ [master] Merge work in mybranch
1028 ------------------------------------------------
1029
1030
1031 Merging external work
1032 ---------------------
1033
1034 It's usually much more common that you merge with somebody else than
1035 merging with your own branches, so it's worth pointing out that git
1036 makes that very easy too, and in fact, it's not that different from
1037 doing a `git merge`. In fact, a remote merge ends up being nothing
1038 more than "fetch the work from a remote repository into a temporary tag"
1039 followed by a `git merge`.
1040
1041 Fetching from a remote repository is done by, unsurprisingly,
1042 `git fetch`:
1043
1044 ----------------
1045 $ git fetch <remote-repository>
1046 ----------------
1047
1048 One of the following transports can be used to name the
1049 repository to download from:
1050
1051 Rsync::
1052 `rsync://remote.machine/path/to/repo.git/`
1053 +
1054 Rsync transport is usable for both uploading and downloading,
1055 but is completely unaware of what git does, and can produce
1056 unexpected results when you download from the public repository
1057 while the repository owner is uploading into it via `rsync`
1058 transport. Most notably, it could update the files under
1059 `refs/` which holds the object name of the topmost commits
1060 before uploading the files in `objects/` -- the downloader would
1061 obtain head commit object name while that object itself is still
1062 not available in the repository. For this reason, it is
1063 considered deprecated.
1064
1065 SSH::
1066 `remote.machine:/path/to/repo.git/` or
1067 +
1068 `ssh://remote.machine/path/to/repo.git/`
1069 +
1070 This transport can be used for both uploading and downloading,
1071 and requires you to have a log-in privilege over `ssh` to the
1072 remote machine. It finds out the set of objects the other side
1073 lacks by exchanging the head commits both ends have and
1074 transfers (close to) minimum set of objects. It is by far the
1075 most efficient way to exchange git objects between repositories.
1076
1077 Local directory::
1078 `/path/to/repo.git/`
1079 +
1080 This transport is the same as SSH transport but uses `sh` to run
1081 both ends on the local machine instead of running other end on
1082 the remote machine via `ssh`.
1083
1084 git Native::
1085 `git://remote.machine/path/to/repo.git/`
1086 +
1087 This transport was designed for anonymous downloading. Like SSH
1088 transport, it finds out the set of objects the downstream side
1089 lacks and transfers (close to) minimum set of objects.
1090
1091 HTTP(S)::
1092 `http://remote.machine/path/to/repo.git/`
1093 +
1094 Downloader from http and https URL
1095 first obtains the topmost commit object name from the remote site
1096 by looking at the specified refname under `repo.git/refs/` directory,
1097 and then tries to obtain the
1098 commit object by downloading from `repo.git/objects/xx/xxx\...`
1099 using the object name of that commit object. Then it reads the
1100 commit object to find out its parent commits and the associate
1101 tree object; it repeats this process until it gets all the
1102 necessary objects. Because of this behaviour, they are
1103 sometimes also called 'commit walkers'.
1104 +
1105 The 'commit walkers' are sometimes also called 'dumb
1106 transports', because they do not require any git aware smart
1107 server like git Native transport does. Any stock HTTP server
1108 that does not even support directory index would suffice. But
1109 you must prepare your repository with `git-update-server-info`
1110 to help dumb transport downloaders.
1111 +
1112 There are (confusingly enough) `git-ssh-fetch` and `git-ssh-upload`
1113 programs, which are 'commit walkers'; they outlived their
1114 usefulness when git Native and SSH transports were introduced,
1115 and not used by `git pull` or `git push` scripts.
1116
1117 Once you fetch from the remote repository, you `resolve` that
1118 with your current branch.
1119
1120 However -- it's such a common thing to `fetch` and then
1121 immediately `resolve`, that it's called `git pull`, and you can
1122 simply do
1123
1124 ----------------
1125 $ git pull <remote-repository>
1126 ----------------
1127
1128 and optionally give a branch-name for the remote end as a second
1129 argument.
1130
1131 [NOTE]
1132 You could do without using any branches at all, by
1133 keeping as many local repositories as you would like to have
1134 branches, and merging between them with `git pull`, just like
1135 you merge between branches. The advantage of this approach is
1136 that it lets you keep set of files for each `branch` checked
1137 out and you may find it easier to switch back and forth if you
1138 juggle multiple lines of development simultaneously. Of
1139 course, you will pay the price of more disk usage to hold
1140 multiple working trees, but disk space is cheap these days.
1141
1142 [NOTE]
1143 You could even pull from your own repository by
1144 giving '.' as <remote-repository> parameter to `git pull`. This
1145 is useful when you want to merge a local branch (or more, if you
1146 are making an Octopus) into the current branch.
1147
1148 It is likely that you will be pulling from the same remote
1149 repository from time to time. As a short hand, you can store
1150 the remote repository URL in a file under .git/remotes/
1151 directory, like this:
1152
1153 ------------------------------------------------
1154 $ mkdir -p .git/remotes/
1155 $ cat >.git/remotes/linus <<\EOF
1156 URL: http://www.kernel.org/pub/scm/git/git.git/
1157 EOF
1158 ------------------------------------------------
1159
1160 and use the filename to `git pull` instead of the full URL.
1161 The URL specified in such file can even be a prefix
1162 of a full URL, like this:
1163
1164 ------------------------------------------------
1165 $ cat >.git/remotes/jgarzik <<\EOF
1166 URL: http://www.kernel.org/pub/scm/linux/git/jgarzik/
1167 EOF
1168 ------------------------------------------------
1169
1170
1171 Examples.
1172
1173 . `git pull linus`
1174 . `git pull linus tag v0.99.1`
1175 . `git pull jgarzik/netdev-2.6.git/ e100`
1176
1177 the above are equivalent to:
1178
1179 . `git pull http://www.kernel.org/pub/scm/git/git.git/ HEAD`
1180 . `git pull http://www.kernel.org/pub/scm/git/git.git/ tag v0.99.1`
1181 . `git pull http://www.kernel.org/pub/.../jgarzik/netdev-2.6.git e100`
1182
1183
1184 How does the merge work?
1185 ------------------------
1186
1187 We said this tutorial shows what plumbing does to help you cope
1188 with the porcelain that isn't flushing, but we so far did not
1189 talk about how the merge really works. If you are following
1190 this tutorial the first time, I'd suggest to skip to "Publishing
1191 your work" section and come back here later.
1192
1193 OK, still with me? To give us an example to look at, let's go
1194 back to the earlier repository with "hello" and "example" file,
1195 and bring ourselves back to the pre-merge state:
1196
1197 ------------
1198 $ git show-branch --more=3 master mybranch
1199 ! [master] Merge work in mybranch
1200 * [mybranch] Merge work in mybranch
1201 --
1202 ++ [master] Merge work in mybranch
1203 ++ [master^2] Some work.
1204 ++ [master^] Some fun.
1205 ------------
1206
1207 Remember, before running `git merge`, our `master` head was at
1208 "Some fun." commit, while our `mybranch` head was at "Some
1209 work." commit.
1210
1211 ------------
1212 $ git checkout mybranch
1213 $ git reset --hard master^2
1214 $ git checkout master
1215 $ git reset --hard master^
1216 ------------
1217
1218 After rewinding, the commit structure should look like this:
1219
1220 ------------
1221 $ git show-branch
1222 * [master] Some fun.
1223 ! [mybranch] Some work.
1224 --
1225 + [mybranch] Some work.
1226 + [master] Some fun.
1227 ++ [mybranch^] New day.
1228 ------------
1229
1230 Now we are ready to experiment with the merge by hand.
1231
1232 `git merge` command, when merging two branches, uses 3-way merge
1233 algorithm. First, it finds the common ancestor between them.
1234 The command it uses is `git-merge-base`:
1235
1236 ------------
1237 $ mb=$(git-merge-base HEAD mybranch)
1238 ------------
1239
1240 The command writes the commit object name of the common ancestor
1241 to the standard output, so we captured its output to a variable,
1242 because we will be using it in the next step. BTW, the common
1243 ancestor commit is the "New day." commit in this case. You can
1244 tell it by:
1245
1246 ------------
1247 $ git-name-rev $mb
1248 my-first-tag
1249 ------------
1250
1251 After finding out a common ancestor commit, the second step is
1252 this:
1253
1254 ------------
1255 $ git-read-tree -m -u $mb HEAD mybranch
1256 ------------
1257
1258 This is the same `git-read-tree` command we have already seen,
1259 but it takes three trees, unlike previous examples. This reads
1260 the contents of each tree into different 'stage' in the index
1261 file (the first tree goes to stage 1, the second stage 2,
1262 etc.). After reading three trees into three stages, the paths
1263 that are the same in all three stages are 'collapsed' into stage
1264 0. Also paths that are the same in two of three stages are
1265 collapsed into stage 0, taking the SHA1 from either stage 2 or
1266 stage 3, whichever is different from stage 1 (i.e. only one side
1267 changed from the common ancestor).
1268
1269 After 'collapsing' operation, paths that are different in three
1270 trees are left in non-zero stages. At this point, you can
1271 inspect the index file with this command:
1272
1273 ------------
1274 $ git-ls-files --stage
1275 100644 7f8b141b65fdcee47321e399a2598a235a032422 0 example
1276 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello
1277 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello
1278 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello
1279 ------------
1280
1281 In our example of only two files, we did not have unchanged
1282 files so only 'example' resulted in collapsing, but in real-life
1283 large projects, only small number of files change in one commit,
1284 and this 'collapsing' tends to trivially merge most of the paths
1285 fairly quickly, leaving only a handful the real changes in non-zero
1286 stages.
1287
1288 To look at only non-zero stages, use `\--unmerged` flag:
1289
1290 ------------
1291 $ git-ls-files --unmerged
1292 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello
1293 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello
1294 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello
1295 ------------
1296
1297 The next step of merging is to merge these three versions of the
1298 file, using 3-way merge. This is done by giving
1299 `git-merge-one-file` command as one of the arguments to
1300 `git-merge-index` command:
1301
1302 ------------
1303 $ git-merge-index git-merge-one-file hello
1304 Auto-merging hello.
1305 merge: warning: conflicts during merge
1306 ERROR: Merge conflict in hello.
1307 fatal: merge program failed
1308 ------------
1309
1310 `git-merge-one-file` script is called with parameters to
1311 describe those three versions, and is responsible to leave the
1312 merge results in the working tree.
1313 It is a fairly straightforward shell script, and
1314 eventually calls `merge` program from RCS suite to perform a
1315 file-level 3-way merge. In this case, `merge` detects
1316 conflicts, and the merge result with conflict marks is left in
1317 the working tree.. This can be seen if you run `ls-files
1318 --stage` again at this point:
1319
1320 ------------
1321 $ git-ls-files --stage
1322 100644 7f8b141b65fdcee47321e399a2598a235a032422 0 example
1323 100644 263414f423d0e4d70dae8fe53fa34614ff3e2860 1 hello
1324 100644 06fa6a24256dc7e560efa5687fa84b51f0263c3a 2 hello
1325 100644 cc44c73eb783565da5831b4d820c962954019b69 3 hello
1326 ------------
1327
1328 This is the state of the index file and the working file after
1329 `git merge` returns control back to you, leaving the conflicting
1330 merge for you to resolve. Notice that the path `hello` is still
1331 unmerged, and what you see with `git diff` at this point is
1332 differences since stage 2 (i.e. your version).
1333
1334
1335 Publishing your work
1336 --------------------
1337
1338 So we can use somebody else's work from a remote repository; but
1339 how can *you* prepare a repository to let other people pull from
1340 it?
1341
1342 Your do your real work in your working tree that has your
1343 primary repository hanging under it as its `.git` subdirectory.
1344 You *could* make that repository accessible remotely and ask
1345 people to pull from it, but in practice that is not the way
1346 things are usually done. A recommended way is to have a public
1347 repository, make it reachable by other people, and when the
1348 changes you made in your primary working tree are in good shape,
1349 update the public repository from it. This is often called
1350 'pushing'.
1351
1352 [NOTE]
1353 This public repository could further be mirrored, and that is
1354 how git repositories at `kernel.org` are managed.
1355
1356 Publishing the changes from your local (private) repository to
1357 your remote (public) repository requires a write privilege on
1358 the remote machine. You need to have an SSH account there to
1359 run a single command, `git-receive-pack`.
1360
1361 First, you need to create an empty repository on the remote
1362 machine that will house your public repository. This empty
1363 repository will be populated and be kept up-to-date by pushing
1364 into it later. Obviously, this repository creation needs to be
1365 done only once.
1366
1367 [NOTE]
1368 `git push` uses a pair of programs,
1369 `git-send-pack` on your local machine, and `git-receive-pack`
1370 on the remote machine. The communication between the two over
1371 the network internally uses an SSH connection.
1372
1373 Your private repository's git directory is usually `.git`, but
1374 your public repository is often named after the project name,
1375 i.e. `<project>.git`. Let's create such a public repository for
1376 project `my-git`. After logging into the remote machine, create
1377 an empty directory:
1378
1379 ------------
1380 $ mkdir my-git.git
1381 ------------
1382
1383 Then, make that directory into a git repository by running
1384 `git init-db`, but this time, since its name is not the usual
1385 `.git`, we do things slightly differently:
1386
1387 ------------
1388 $ GIT_DIR=my-git.git git-init-db
1389 ------------
1390
1391 Make sure this directory is available for others you want your
1392 changes to be pulled by via the transport of your choice. Also
1393 you need to make sure that you have the `git-receive-pack`
1394 program on the `$PATH`.
1395
1396 [NOTE]
1397 Many installations of sshd do not invoke your shell as the login
1398 shell when you directly run programs; what this means is that if
1399 your login shell is `bash`, only `.bashrc` is read and not
1400 `.bash_profile`. As a workaround, make sure `.bashrc` sets up
1401 `$PATH` so that you can run `git-receive-pack` program.
1402
1403 [NOTE]
1404 If you plan to publish this repository to be accessed over http,
1405 you should do `chmod +x my-git.git/hooks/post-update` at this
1406 point. This makes sure that every time you push into this
1407 repository, `git-update-server-info` is run.
1408
1409 Your "public repository" is now ready to accept your changes.
1410 Come back to the machine you have your private repository. From
1411 there, run this command:
1412
1413 ------------
1414 $ git push <public-host>:/path/to/my-git.git master
1415 ------------
1416
1417 This synchronizes your public repository to match the named
1418 branch head (i.e. `master` in this case) and objects reachable
1419 from them in your current repository.
1420
1421 As a real example, this is how I update my public git
1422 repository. Kernel.org mirror network takes care of the
1423 propagation to other publicly visible machines:
1424
1425 ------------
1426 $ git push master.kernel.org:/pub/scm/git/git.git/
1427 ------------
1428
1429
1430 Packing your repository
1431 -----------------------
1432
1433 Earlier, we saw that one file under `.git/objects/??/` directory
1434 is stored for each git object you create. This representation
1435 is efficient to create atomically and safely, but
1436 not so convenient to transport over the network. Since git objects are
1437 immutable once they are created, there is a way to optimize the
1438 storage by "packing them together". The command
1439
1440 ------------
1441 $ git repack
1442 ------------
1443
1444 will do it for you. If you followed the tutorial examples, you
1445 would have accumulated about 17 objects in `.git/objects/??/`
1446 directories by now. `git repack` tells you how many objects it
1447 packed, and stores the packed file in `.git/objects/pack`
1448 directory.
1449
1450 [NOTE]
1451 You will see two files, `pack-\*.pack` and `pack-\*.idx`,
1452 in `.git/objects/pack` directory. They are closely related to
1453 each other, and if you ever copy them by hand to a different
1454 repository for whatever reason, you should make sure you copy
1455 them together. The former holds all the data from the objects
1456 in the pack, and the latter holds the index for random
1457 access.
1458
1459 If you are paranoid, running `git-verify-pack` command would
1460 detect if you have a corrupt pack, but do not worry too much.
1461 Our programs are always perfect ;-).
1462
1463 Once you have packed objects, you do not need to leave the
1464 unpacked objects that are contained in the pack file anymore.
1465
1466 ------------
1467 $ git prune-packed
1468 ------------
1469
1470 would remove them for you.
1471
1472 You can try running `find .git/objects -type f` before and after
1473 you run `git prune-packed` if you are curious. Also `git
1474 count-objects` would tell you how many unpacked objects are in
1475 your repository and how much space they are consuming.
1476
1477 [NOTE]
1478 `git pull` is slightly cumbersome for HTTP transport, as a
1479 packed repository may contain relatively few objects in a
1480 relatively large pack. If you expect many HTTP pulls from your
1481 public repository you might want to repack & prune often, or
1482 never.
1483
1484 If you run `git repack` again at this point, it will say
1485 "Nothing to pack". Once you continue your development and
1486 accumulate the changes, running `git repack` again will create a
1487 new pack, that contains objects created since you packed your
1488 repository the last time. We recommend that you pack your project
1489 soon after the initial import (unless you are starting your
1490 project from scratch), and then run `git repack` every once in a
1491 while, depending on how active your project is.
1492
1493 When a repository is synchronized via `git push` and `git pull`
1494 objects packed in the source repository are usually stored
1495 unpacked in the destination, unless rsync transport is used.
1496 While this allows you to use different packing strategies on
1497 both ends, it also means you may need to repack both
1498 repositories every once in a while.
1499
1500
1501 Working with Others
1502 -------------------
1503
1504 Although git is a truly distributed system, it is often
1505 convenient to organize your project with an informal hierarchy
1506 of developers. Linux kernel development is run this way. There
1507 is a nice illustration (page 17, "Merges to Mainline") in Randy
1508 Dunlap's presentation (`http://tinyurl.com/a2jdg`).
1509
1510 It should be stressed that this hierarchy is purely *informal*.
1511 There is nothing fundamental in git that enforces the "chain of
1512 patch flow" this hierarchy implies. You do not have to pull
1513 from only one remote repository.
1514
1515 A recommended workflow for a "project lead" goes like this:
1516
1517 1. Prepare your primary repository on your local machine. Your
1518 work is done there.
1519
1520 2. Prepare a public repository accessible to others.
1521 +
1522 If other people are pulling from your repository over dumb
1523 transport protocols (HTTP), you need to keep this repository
1524 'dumb transport friendly'. After `git init-db`,
1525 `$GIT_DIR/hooks/post-update` copied from the standard templates
1526 would contain a call to `git-update-server-info` but the
1527 `post-update` hook itself is disabled by default -- enable it
1528 with `chmod +x post-update`. This makes sure `git-update-server-info`
1529 keeps the necessary files up-to-date.
1530
1531 3. Push into the public repository from your primary
1532 repository.
1533
1534 4. `git repack` the public repository. This establishes a big
1535 pack that contains the initial set of objects as the
1536 baseline, and possibly `git prune` if the transport
1537 used for pulling from your repository supports packed
1538 repositories.
1539
1540 5. Keep working in your primary repository. Your changes
1541 include modifications of your own, patches you receive via
1542 e-mails, and merges resulting from pulling the "public"
1543 repositories of your "subsystem maintainers".
1544 +
1545 You can repack this private repository whenever you feel like.
1546
1547 6. Push your changes to the public repository, and announce it
1548 to the public.
1549
1550 7. Every once in a while, "git repack" the public repository.
1551 Go back to step 5. and continue working.
1552
1553
1554 A recommended work cycle for a "subsystem maintainer" who works
1555 on that project and has an own "public repository" goes like this:
1556
1557 1. Prepare your work repository, by `git clone` the public
1558 repository of the "project lead". The URL used for the
1559 initial cloning is stored in `.git/remotes/origin`.
1560
1561 2. Prepare a public repository accessible to others, just like
1562 the "project lead" person does.
1563
1564 3. Copy over the packed files from "project lead" public
1565 repository to your public repository, unless the "project
1566 lead" repository lives on the same machine as yours. In the
1567 latter case, you can use `objects/info/alternates` file to
1568 point at the repository you are borrowing from.
1569
1570 4. Push into the public repository from your primary
1571 repository. Run `git repack`, and possibly `git prune` if the
1572 transport used for pulling from your repository supports
1573 packed repositories.
1574
1575 5. Keep working in your primary repository. Your changes
1576 include modifications of your own, patches you receive via
1577 e-mails, and merges resulting from pulling the "public"
1578 repositories of your "project lead" and possibly your
1579 "sub-subsystem maintainers".
1580 +
1581 You can repack this private repository whenever you feel
1582 like.
1583
1584 6. Push your changes to your public repository, and ask your
1585 "project lead" and possibly your "sub-subsystem
1586 maintainers" to pull from it.
1587
1588 7. Every once in a while, `git repack` the public repository.
1589 Go back to step 5. and continue working.
1590
1591
1592 A recommended work cycle for an "individual developer" who does
1593 not have a "public" repository is somewhat different. It goes
1594 like this:
1595
1596 1. Prepare your work repository, by `git clone` the public
1597 repository of the "project lead" (or a "subsystem
1598 maintainer", if you work on a subsystem). The URL used for
1599 the initial cloning is stored in `.git/remotes/origin`.
1600
1601 2. Do your work in your repository on 'master' branch.
1602
1603 3. Run `git fetch origin` from the public repository of your
1604 upstream every once in a while. This does only the first
1605 half of `git pull` but does not merge. The head of the
1606 public repository is stored in `.git/refs/heads/origin`.
1607
1608 4. Use `git cherry origin` to see which ones of your patches
1609 were accepted, and/or use `git rebase origin` to port your
1610 unmerged changes forward to the updated upstream.
1611
1612 5. Use `git format-patch origin` to prepare patches for e-mail
1613 submission to your upstream and send it out. Go back to
1614 step 2. and continue.
1615
1616
1617 Working with Others, Shared Repository Style
1618 --------------------------------------------
1619
1620 If you are coming from CVS background, the style of cooperation
1621 suggested in the previous section may be new to you. You do not
1622 have to worry. git supports "shared public repository" style of
1623 cooperation you are probably more familiar with as well.
1624
1625 For this, set up a public repository on a machine that is
1626 reachable via SSH by people with "commit privileges". Put the
1627 committers in the same user group and make the repository
1628 writable by that group.
1629
1630 You, as an individual committer, then:
1631
1632 - First clone the shared repository to a local repository:
1633 ------------------------------------------------
1634 $ git clone repo.shared.xz:/pub/scm/project.git/ my-project
1635 $ cd my-project
1636 $ hack away
1637 ------------------------------------------------
1638
1639 - Merge the work others might have done while you were hacking
1640 away:
1641 ------------------------------------------------
1642 $ git pull origin
1643 $ test the merge result
1644 ------------------------------------------------
1645 [NOTE]
1646 ================================
1647 The first `git clone` would have placed the following in
1648 `my-project/.git/remotes/origin` file, and that's why this and
1649 the next step work.
1650 ------------
1651 URL: repo.shared.xz:/pub/scm/project.git/ my-project
1652 Pull: master:origin
1653 ------------
1654 ================================
1655
1656 - push your work as the new head of the shared
1657 repository.
1658 ------------------------------------------------
1659 $ git push origin master
1660 ------------------------------------------------
1661 If somebody else pushed into the same shared repository while
1662 you were working locally, `git push` in the last step would
1663 complain, telling you that the remote `master` head does not
1664 fast forward. You need to pull and merge those other changes
1665 back before you push your work when it happens.
1666
1667
1668 Advanced Shared Repository Management
1669 -------------------------------------
1670
1671 Being able to push into a shared repository means being able to
1672 write into it. If your developers are coming over the network,
1673 this means you, as the repository administrator, need to give
1674 each of them an SSH access to the shared repository machine.
1675
1676 In some cases, though, you may not want to give a normal shell
1677 account to them, but want to restrict them to be able to only
1678 do `git push` into the repository and nothing else.
1679
1680 You can achieve this by setting the login shell of your
1681 developers on the shared repository host to `git-shell` program.
1682
1683 [NOTE]
1684 Most likely you would also need to list `git-shell` program in
1685 `/etc/shells` file.
1686
1687 This restricts the set of commands that can be run from incoming
1688 SSH connection for these users to only `receive-pack` and
1689 `upload-pack`, so the only thing they can do are `git fetch` and
1690 `git push`.
1691
1692 You still need to create UNIX user accounts for each developer,
1693 and put them in the same group. Make sure that the repository
1694 shared among these developers is writable by that group.
1695
1696 You can implement finer grained branch policies using update
1697 hooks. There is a document ("control access to branches") in
1698 Documentation/howto by Carl Baldwin and JC outlining how to (1)
1699 limit access to branch per user, (2) forbid overwriting existing
1700 tags.
1701
1702
1703 Bundling your work together
1704 ---------------------------
1705
1706 It is likely that you will be working on more than one thing at
1707 a time. It is easy to manage those more-or-less independent tasks
1708 using branches with git.
1709
1710 We have already seen how branches work previously,
1711 with "fun and work" example using two branches. The idea is the
1712 same if there are more than two branches. Let's say you started
1713 out from "master" head, and have some new code in the "master"
1714 branch, and two independent fixes in the "commit-fix" and
1715 "diff-fix" branches:
1716
1717 ------------
1718 $ git show-branch
1719 ! [commit-fix] Fix commit message normalization.
1720 ! [diff-fix] Fix rename detection.
1721 * [master] Release candidate #1
1722 ---
1723 + [diff-fix] Fix rename detection.
1724 + [diff-fix~1] Better common substring algorithm.
1725 + [commit-fix] Fix commit message normalization.
1726 + [master] Release candidate #1
1727 +++ [diff-fix~2] Pretty-print messages.
1728 ------------
1729
1730 Both fixes are tested well, and at this point, you want to merge
1731 in both of them. You could merge in 'diff-fix' first and then
1732 'commit-fix' next, like this:
1733
1734 ------------
1735 $ git merge 'Merge fix in diff-fix' master diff-fix
1736 $ git merge 'Merge fix in commit-fix' master commit-fix
1737 ------------
1738
1739 Which would result in:
1740
1741 ------------
1742 $ git show-branch
1743 ! [commit-fix] Fix commit message normalization.
1744 ! [diff-fix] Fix rename detection.
1745 * [master] Merge fix in commit-fix
1746 ---
1747 + [master] Merge fix in commit-fix
1748 + + [commit-fix] Fix commit message normalization.
1749 + [master~1] Merge fix in diff-fix
1750 ++ [diff-fix] Fix rename detection.
1751 ++ [diff-fix~1] Better common substring algorithm.
1752 + [master~2] Release candidate #1
1753 +++ [master~3] Pretty-print messages.
1754 ------------
1755
1756 However, there is no particular reason to merge in one branch
1757 first and the other next, when what you have are a set of truly
1758 independent changes (if the order mattered, then they are not
1759 independent by definition). You could instead merge those two
1760 branches into the current branch at once. First let's undo what
1761 we just did and start over. We would want to get the master
1762 branch before these two merges by resetting it to 'master~2':
1763
1764 ------------
1765 $ git reset --hard master~2
1766 ------------
1767
1768 You can make sure 'git show-branch' matches the state before
1769 those two 'git merge' you just did. Then, instead of running
1770 two 'git merge' commands in a row, you would pull these two
1771 branch heads (this is known as 'making an Octopus'):
1772
1773 ------------
1774 $ git pull . commit-fix diff-fix
1775 $ git show-branch
1776 ! [commit-fix] Fix commit message normalization.
1777 ! [diff-fix] Fix rename detection.
1778 * [master] Octopus merge of branches 'diff-fix' and 'commit-fix'
1779 ---
1780 + [master] Octopus merge of branches 'diff-fix' and 'commit-fix'
1781 + + [commit-fix] Fix commit message normalization.
1782 ++ [diff-fix] Fix rename detection.
1783 ++ [diff-fix~1] Better common substring algorithm.
1784 + [master~1] Release candidate #1
1785 +++ [master~2] Pretty-print messages.
1786 ------------
1787
1788 Note that you should not do Octopus because you can. An octopus
1789 is a valid thing to do and often makes it easier to view the
1790 commit history if you are pulling more than two independent
1791 changes at the same time. However, if you have merge conflicts
1792 with any of the branches you are merging in and need to hand
1793 resolve, that is an indication that the development happened in
1794 those branches were not independent after all, and you should
1795 merge two at a time, documenting how you resolved the conflicts,
1796 and the reason why you preferred changes made in one side over
1797 the other. Otherwise it would make the project history harder
1798 to follow, not easier.
1799
1800 [ to be continued.. cvsimports ]