Merge branch 'maint'
[git/git.git] / Documentation / git-read-tree.txt
1 git-read-tree(1)
2 ================
3
4 NAME
5 ----
6 git-read-tree - Reads tree information into the index
7
8
9 SYNOPSIS
10 --------
11 [verse]
12 'git read-tree' [[-m [--trivial] [--aggressive] | --reset | --prefix=<prefix>]
13 [-u [--exclude-per-directory=<gitignore>] | -i]]
14 [--index-output=<file>] [--no-sparse-checkout]
15 (--empty | <tree-ish1> [<tree-ish2> [<tree-ish3>]])
16
17
18 DESCRIPTION
19 -----------
20 Reads the tree information given by <tree-ish> into the index,
21 but does not actually *update* any of the files it "caches". (see:
22 linkgit:git-checkout-index[1])
23
24 Optionally, it can merge a tree into the index, perform a
25 fast-forward (i.e. 2-way) merge, or a 3-way merge, with the `-m`
26 flag. When used with `-m`, the `-u` flag causes it to also update
27 the files in the work tree with the result of the merge.
28
29 Trivial merges are done by 'git read-tree' itself. Only conflicting paths
30 will be in unmerged state when 'git read-tree' returns.
31
32 OPTIONS
33 -------
34 -m::
35 Perform a merge, not just a read. The command will
36 refuse to run if your index file has unmerged entries,
37 indicating that you have not finished previous merge you
38 started.
39
40 --reset::
41 Same as -m, except that unmerged entries are discarded
42 instead of failing.
43
44 -u::
45 After a successful merge, update the files in the work
46 tree with the result of the merge.
47
48 -i::
49 Usually a merge requires the index file as well as the
50 files in the working tree to be up to date with the
51 current head commit, in order not to lose local
52 changes. This flag disables the check with the working
53 tree and is meant to be used when creating a merge of
54 trees that are not directly related to the current
55 working tree status into a temporary index file.
56
57 -n::
58 --dry-run::
59 Check if the command would error out, without updating the index
60 nor the files in the working tree for real.
61
62 -v::
63 Show the progress of checking files out.
64
65 --trivial::
66 Restrict three-way merge by 'git read-tree' to happen
67 only if there is no file-level merging required, instead
68 of resolving merge for trivial cases and leaving
69 conflicting files unresolved in the index.
70
71 --aggressive::
72 Usually a three-way merge by 'git read-tree' resolves
73 the merge for really trivial cases and leaves other
74 cases unresolved in the index, so that porcelains can
75 implement different merge policies. This flag makes the
76 command resolve a few more cases internally:
77 +
78 * when one side removes a path and the other side leaves the path
79 unmodified. The resolution is to remove that path.
80 * when both sides remove a path. The resolution is to remove that path.
81 * when both sides add a path identically. The resolution
82 is to add that path.
83
84 --prefix=<prefix>/::
85 Keep the current index contents, and read the contents
86 of the named tree-ish under the directory at `<prefix>`. The
87 original index file cannot have anything at the path
88 `<prefix>` itself, nor anything in the `<prefix>/`
89 directory. Note that the `<prefix>/` value must end
90 with a slash.
91
92 --exclude-per-directory=<gitignore>::
93 When running the command with `-u` and `-m` options, the
94 merge result may need to overwrite paths that are not
95 tracked in the current branch. The command usually
96 refuses to proceed with the merge to avoid losing such a
97 path. However this safety valve sometimes gets in the
98 way. For example, it often happens that the other
99 branch added a file that used to be a generated file in
100 your branch, and the safety valve triggers when you try
101 to switch to that branch after you ran `make` but before
102 running `make clean` to remove the generated file. This
103 option tells the command to read per-directory exclude
104 file (usually '.gitignore') and allows such an untracked
105 but explicitly ignored file to be overwritten.
106
107 --index-output=<file>::
108 Instead of writing the results out to `$GIT_INDEX_FILE`,
109 write the resulting index in the named file. While the
110 command is operating, the original index file is locked
111 with the same mechanism as usual. The file must allow
112 to be rename(2)ed into from a temporary file that is
113 created next to the usual index file; typically this
114 means it needs to be on the same filesystem as the index
115 file itself, and you need write permission to the
116 directories the index file and index output file are
117 located in.
118
119 --no-sparse-checkout::
120 Disable sparse checkout support even if `core.sparseCheckout`
121 is true.
122
123 --empty::
124 Instead of reading tree object(s) into the index, just empty
125 it.
126
127 <tree-ish#>::
128 The id of the tree object(s) to be read/merged.
129
130
131 Merging
132 -------
133 If `-m` is specified, 'git read-tree' can perform 3 kinds of
134 merge, a single tree merge if only 1 tree is given, a
135 fast-forward merge with 2 trees, or a 3-way merge if 3 trees are
136 provided.
137
138
139 Single Tree Merge
140 ~~~~~~~~~~~~~~~~~
141 If only 1 tree is specified, 'git read-tree' operates as if the user did not
142 specify `-m`, except that if the original index has an entry for a
143 given pathname, and the contents of the path match with the tree
144 being read, the stat info from the index is used. (In other words, the
145 index's stat()s take precedence over the merged tree's).
146
147 That means that if you do a `git read-tree -m <newtree>` followed by a
148 `git checkout-index -f -u -a`, the 'git checkout-index' only checks out
149 the stuff that really changed.
150
151 This is used to avoid unnecessary false hits when 'git diff-files' is
152 run after 'git read-tree'.
153
154
155 Two Tree Merge
156 ~~~~~~~~~~~~~~
157
158 Typically, this is invoked as `git read-tree -m $H $M`, where $H
159 is the head commit of the current repository, and $M is the head
160 of a foreign tree, which is simply ahead of $H (i.e. we are in a
161 fast-forward situation).
162
163 When two trees are specified, the user is telling 'git read-tree'
164 the following:
165
166 1. The current index and work tree is derived from $H, but
167 the user may have local changes in them since $H.
168
169 2. The user wants to fast-forward to $M.
170
171 In this case, the `git read-tree -m $H $M` command makes sure
172 that no local change is lost as the result of this "merge".
173 Here are the "carry forward" rules, where "I" denotes the index,
174 "clean" means that index and work tree coincide, and "exists"/"nothing"
175 refer to the presence of a path in the specified commit:
176
177 I H M Result
178 -------------------------------------------------------
179 0 nothing nothing nothing (does not happen)
180 1 nothing nothing exists use M
181 2 nothing exists nothing remove path from index
182 3 nothing exists exists, use M if "initial checkout",
183 H == M keep index otherwise
184 exists, fail
185 H != M
186
187 clean I==H I==M
188 ------------------
189 4 yes N/A N/A nothing nothing keep index
190 5 no N/A N/A nothing nothing keep index
191
192 6 yes N/A yes nothing exists keep index
193 7 no N/A yes nothing exists keep index
194 8 yes N/A no nothing exists fail
195 9 no N/A no nothing exists fail
196
197 10 yes yes N/A exists nothing remove path from index
198 11 no yes N/A exists nothing fail
199 12 yes no N/A exists nothing fail
200 13 no no N/A exists nothing fail
201
202 clean (H==M)
203 ------
204 14 yes exists exists keep index
205 15 no exists exists keep index
206
207 clean I==H I==M (H!=M)
208 ------------------
209 16 yes no no exists exists fail
210 17 no no no exists exists fail
211 18 yes no yes exists exists keep index
212 19 no no yes exists exists keep index
213 20 yes yes no exists exists use M
214 21 no yes no exists exists fail
215
216 In all "keep index" cases, the index entry stays as in the
217 original index file. If the entry is not up to date,
218 'git read-tree' keeps the copy in the work tree intact when
219 operating under the -u flag.
220
221 When this form of 'git read-tree' returns successfully, you can
222 see which of the "local changes" that you made were carried forward by running
223 `git diff-index --cached $M`. Note that this does not
224 necessarily match what `git diff-index --cached $H` would have
225 produced before such a two tree merge. This is because of cases
226 18 and 19 --- if you already had the changes in $M (e.g. maybe
227 you picked it up via e-mail in a patch form), `git diff-index
228 --cached $H` would have told you about the change before this
229 merge, but it would not show in `git diff-index --cached $M`
230 output after the two-tree merge.
231
232 Case 3 is slightly tricky and needs explanation. The result from this
233 rule logically should be to remove the path if the user staged the removal
234 of the path and then switching to a new branch. That however will prevent
235 the initial checkout from happening, so the rule is modified to use M (new
236 tree) only when the content of the index is empty. Otherwise the removal
237 of the path is kept as long as $H and $M are the same.
238
239 3-Way Merge
240 ~~~~~~~~~~~
241 Each "index" entry has two bits worth of "stage" state. stage 0 is the
242 normal one, and is the only one you'd see in any kind of normal use.
243
244 However, when you do 'git read-tree' with three trees, the "stage"
245 starts out at 1.
246
247 This means that you can do
248
249 ----------------
250 $ git read-tree -m <tree1> <tree2> <tree3>
251 ----------------
252
253 and you will end up with an index with all of the <tree1> entries in
254 "stage1", all of the <tree2> entries in "stage2" and all of the
255 <tree3> entries in "stage3". When performing a merge of another
256 branch into the current branch, we use the common ancestor tree
257 as <tree1>, the current branch head as <tree2>, and the other
258 branch head as <tree3>.
259
260 Furthermore, 'git read-tree' has special-case logic that says: if you see
261 a file that matches in all respects in the following states, it
262 "collapses" back to "stage0":
263
264 - stage 2 and 3 are the same; take one or the other (it makes no
265 difference - the same work has been done on our branch in
266 stage 2 and their branch in stage 3)
267
268 - stage 1 and stage 2 are the same and stage 3 is different; take
269 stage 3 (our branch in stage 2 did not do anything since the
270 ancestor in stage 1 while their branch in stage 3 worked on
271 it)
272
273 - stage 1 and stage 3 are the same and stage 2 is different take
274 stage 2 (we did something while they did nothing)
275
276 The 'git write-tree' command refuses to write a nonsensical tree, and it
277 will complain about unmerged entries if it sees a single entry that is not
278 stage 0.
279
280 OK, this all sounds like a collection of totally nonsensical rules,
281 but it's actually exactly what you want in order to do a fast
282 merge. The different stages represent the "result tree" (stage 0, aka
283 "merged"), the original tree (stage 1, aka "orig"), and the two trees
284 you are trying to merge (stage 2 and 3 respectively).
285
286 The order of stages 1, 2 and 3 (hence the order of three
287 <tree-ish> command line arguments) are significant when you
288 start a 3-way merge with an index file that is already
289 populated. Here is an outline of how the algorithm works:
290
291 - if a file exists in identical format in all three trees, it will
292 automatically collapse to "merged" state by 'git read-tree'.
293
294 - a file that has _any_ difference what-so-ever in the three trees
295 will stay as separate entries in the index. It's up to "porcelain
296 policy" to determine how to remove the non-0 stages, and insert a
297 merged version.
298
299 - the index file saves and restores with all this information, so you
300 can merge things incrementally, but as long as it has entries in
301 stages 1/2/3 (i.e., "unmerged entries") you can't write the result. So
302 now the merge algorithm ends up being really simple:
303
304 * you walk the index in order, and ignore all entries of stage 0,
305 since they've already been done.
306
307 * if you find a "stage1", but no matching "stage2" or "stage3", you
308 know it's been removed from both trees (it only existed in the
309 original tree), and you remove that entry.
310
311 * if you find a matching "stage2" and "stage3" tree, you remove one
312 of them, and turn the other into a "stage0" entry. Remove any
313 matching "stage1" entry if it exists too. .. all the normal
314 trivial rules ..
315
316 You would normally use 'git merge-index' with supplied
317 'git merge-one-file' to do this last step. The script updates
318 the files in the working tree as it merges each path and at the
319 end of a successful merge.
320
321 When you start a 3-way merge with an index file that is already
322 populated, it is assumed that it represents the state of the
323 files in your work tree, and you can even have files with
324 changes unrecorded in the index file. It is further assumed
325 that this state is "derived" from the stage 2 tree. The 3-way
326 merge refuses to run if it finds an entry in the original index
327 file that does not match stage 2.
328
329 This is done to prevent you from losing your work-in-progress
330 changes, and mixing your random changes in an unrelated merge
331 commit. To illustrate, suppose you start from what has been
332 committed last to your repository:
333
334 ----------------
335 $ JC=`git rev-parse --verify "HEAD^0"`
336 $ git checkout-index -f -u -a $JC
337 ----------------
338
339 You do random edits, without running 'git update-index'. And then
340 you notice that the tip of your "upstream" tree has advanced
341 since you pulled from him:
342
343 ----------------
344 $ git fetch git://.... linus
345 $ LT=`cat .git/FETCH_HEAD`
346 ----------------
347
348 Your work tree is still based on your HEAD ($JC), but you have
349 some edits since. Three-way merge makes sure that you have not
350 added or modified index entries since $JC, and if you haven't,
351 then does the right thing. So with the following sequence:
352
353 ----------------
354 $ git read-tree -m -u `git merge-base $JC $LT` $JC $LT
355 $ git merge-index git-merge-one-file -a
356 $ echo "Merge with Linus" | \
357 git commit-tree `git write-tree` -p $JC -p $LT
358 ----------------
359
360 what you would commit is a pure merge between $JC and $LT without
361 your work-in-progress changes, and your work tree would be
362 updated to the result of the merge.
363
364 However, if you have local changes in the working tree that
365 would be overwritten by this merge, 'git read-tree' will refuse
366 to run to prevent your changes from being lost.
367
368 In other words, there is no need to worry about what exists only
369 in the working tree. When you have local changes in a part of
370 the project that is not involved in the merge, your changes do
371 not interfere with the merge, and are kept intact. When they
372 *do* interfere, the merge does not even start ('git read-tree'
373 complains loudly and fails without modifying anything). In such
374 a case, you can simply continue doing what you were in the
375 middle of doing, and when your working tree is ready (i.e. you
376 have finished your work-in-progress), attempt the merge again.
377
378
379 Sparse checkout
380 ---------------
381
382 "Sparse checkout" allows populating the working directory sparsely.
383 It uses the skip-worktree bit (see linkgit:git-update-index[1]) to tell
384 Git whether a file in the working directory is worth looking at.
385
386 'git read-tree' and other merge-based commands ('git merge', 'git
387 checkout'...) can help maintaining the skip-worktree bitmap and working
388 directory update. `$GIT_DIR/info/sparse-checkout` is used to
389 define the skip-worktree reference bitmap. When 'git read-tree' needs
390 to update the working directory, it resets the skip-worktree bit in the index
391 based on this file, which uses the same syntax as .gitignore files.
392 If an entry matches a pattern in this file, skip-worktree will not be
393 set on that entry. Otherwise, skip-worktree will be set.
394
395 Then it compares the new skip-worktree value with the previous one. If
396 skip-worktree turns from set to unset, it will add the corresponding
397 file back. If it turns from unset to set, that file will be removed.
398
399 While `$GIT_DIR/info/sparse-checkout` is usually used to specify what
400 files are in, you can also specify what files are _not_ in, using
401 negate patterns. For example, to remove the file `unwanted`:
402
403 ----------------
404 /*
405 !unwanted
406 ----------------
407
408 Another tricky thing is fully repopulating the working directory when you
409 no longer want sparse checkout. You cannot just disable "sparse
410 checkout" because skip-worktree bits are still in the index and your working
411 directory is still sparsely populated. You should re-populate the working
412 directory with the `$GIT_DIR/info/sparse-checkout` file content as
413 follows:
414
415 ----------------
416 /*
417 ----------------
418
419 Then you can disable sparse checkout. Sparse checkout support in 'git
420 read-tree' and similar commands is disabled by default. You need to
421 turn `core.sparseCheckout` on in order to have sparse checkout
422 support.
423
424
425 SEE ALSO
426 --------
427 linkgit:git-write-tree[1]; linkgit:git-ls-files[1];
428 linkgit:gitignore[5]
429
430 GIT
431 ---
432 Part of the linkgit:git[1] suite