Documentation / git-read-tree.txton commit Tidy up git mergetool's backup file behaviour (44c36d1)
   1git-read-tree(1)
   2================
   3
   4NAME
   5----
   6git-read-tree - Reads tree information into the index
   7
   8
   9SYNOPSIS
  10--------
  11'git-read-tree' (<tree-ish> | [[-m [--trivial] [--aggressive] | --reset | --prefix=<prefix>] [-u | -i]] [--exclude-per-directory=<gitignore>] [--index-output=<file>] <tree-ish1> [<tree-ish2> [<tree-ish3>]])
  12
  13
  14DESCRIPTION
  15-----------
  16Reads the tree information given by <tree-ish> into the index,
  17but does not actually *update* any of the files it "caches". (see:
  18linkgit:git-checkout-index[1])
  19
  20Optionally, it can merge a tree into the index, perform a
  21fast-forward (i.e. 2-way) merge, or a 3-way merge, with the `-m`
  22flag.  When used with `-m`, the `-u` flag causes it to also update
  23the files in the work tree with the result of the merge.
  24
  25Trivial merges are done by `git-read-tree` itself.  Only conflicting paths
  26will be in unmerged state when `git-read-tree` returns.
  27
  28OPTIONS
  29-------
  30-m::
  31        Perform a merge, not just a read.  The command will
  32        refuse to run if your index file has unmerged entries,
  33        indicating that you have not finished previous merge you
  34        started.
  35
  36--reset::
  37        Same as -m, except that unmerged entries are discarded
  38        instead of failing.
  39
  40-u::
  41        After a successful merge, update the files in the work
  42        tree with the result of the merge.
  43
  44-i::
  45        Usually a merge requires the index file as well as the
  46        files in the working tree are up to date with the
  47        current head commit, in order not to lose local
  48        changes.  This flag disables the check with the working
  49        tree and is meant to be used when creating a merge of
  50        trees that are not directly related to the current
  51        working tree status into a temporary index file.
  52
  53--trivial::
  54        Restrict three-way merge by `git-read-tree` to happen
  55        only if there is no file-level merging required, instead
  56        of resolving merge for trivial cases and leaving
  57        conflicting files unresolved in the index.
  58
  59--aggressive::
  60        Usually a three-way merge by `git-read-tree` resolves
  61        the merge for really trivial cases and leaves other
  62        cases unresolved in the index, so that Porcelains can
  63        implement different merge policies.  This flag makes the
  64        command to resolve a few more cases internally:
  65+
  66* when one side removes a path and the other side leaves the path
  67  unmodified.  The resolution is to remove that path.
  68* when both sides remove a path.  The resolution is to remove that path.
  69* when both sides adds a path identically.  The resolution
  70  is to add that path.
  71
  72--prefix=<prefix>/::
  73        Keep the current index contents, and read the contents
  74        of named tree-ish under directory at `<prefix>`.  The
  75        original index file cannot have anything at the path
  76        `<prefix>` itself, and have nothing in `<prefix>/`
  77        directory.  Note that the `<prefix>/` value must end
  78        with a slash.
  79
  80--exclude-per-directory=<gitignore>::
  81        When running the command with `-u` and `-m` options, the
  82        merge result may need to overwrite paths that are not
  83        tracked in the current branch.  The command usually
  84        refuses to proceed with the merge to avoid losing such a
  85        path.  However this safety valve sometimes gets in the
  86        way.  For example, it often happens that the other
  87        branch added a file that used to be a generated file in
  88        your branch, and the safety valve triggers when you try
  89        to switch to that branch after you ran `make` but before
  90        running `make clean` to remove the generated file.  This
  91        option tells the command to read per-directory exclude
  92        file (usually '.gitignore') and allows such an untracked
  93        but explicitly ignored file to be overwritten.
  94
  95--index-output=<file>::
  96        Instead of writing the results out to `$GIT_INDEX_FILE`,
  97        write the resulting index in the named file.  While the
  98        command is operating, the original index file is locked
  99        with the same mechanism as usual.  The file must allow
 100        to be rename(2)ed into from a temporary file that is
 101        created next to the usual index file; typically this
 102        means it needs to be on the same filesystem as the index
 103        file itself, and you need write permission to the
 104        directories the index file and index output file are
 105        located in.
 106
 107<tree-ish#>::
 108        The id of the tree object(s) to be read/merged.
 109
 110
 111Merging
 112-------
 113If `-m` is specified, `git-read-tree` can perform 3 kinds of
 114merge, a single tree merge if only 1 tree is given, a
 115fast-forward merge with 2 trees, or a 3-way merge if 3 trees are
 116provided.
 117
 118
 119Single Tree Merge
 120~~~~~~~~~~~~~~~~~
 121If only 1 tree is specified, git-read-tree operates as if the user did not
 122specify `-m`, except that if the original index has an entry for a
 123given pathname, and the contents of the path matches with the tree
 124being read, the stat info from the index is used. (In other words, the
 125index's stat()s take precedence over the merged tree's).
 126
 127That means that if you do a `git-read-tree -m <newtree>` followed by a
 128`git-checkout-index -f -u -a`, the `git-checkout-index` only checks out
 129the stuff that really changed.
 130
 131This is used to avoid unnecessary false hits when `git-diff-files` is
 132run after `git-read-tree`.
 133
 134
 135Two Tree Merge
 136~~~~~~~~~~~~~~
 137
 138Typically, this is invoked as `git-read-tree -m $H $M`, where $H
 139is the head commit of the current repository, and $M is the head
 140of a foreign tree, which is simply ahead of $H (i.e. we are in a
 141fast forward situation).
 142
 143When two trees are specified, the user is telling git-read-tree
 144the following:
 145
 146     1. The current index and work tree is derived from $H, but
 147        the user may have local changes in them since $H;
 148
 149     2. The user wants to fast-forward to $M.
 150
 151In this case, the `git-read-tree -m $H $M` command makes sure
 152that no local change is lost as the result of this "merge".
 153Here are the "carry forward" rules:
 154
 155        I (index)           H        M        Result
 156       -------------------------------------------------------
 157      0 nothing             nothing  nothing  (does not happen)
 158      1 nothing             nothing  exists   use M
 159      2 nothing             exists   nothing  remove path from index
 160      3 nothing             exists   exists   use M
 161
 162        clean I==H  I==M
 163       ------------------
 164      4 yes   N/A   N/A     nothing  nothing  keep index
 165      5 no    N/A   N/A     nothing  nothing  keep index
 166
 167      6 yes   N/A   yes     nothing  exists   keep index
 168      7 no    N/A   yes     nothing  exists   keep index
 169      8 yes   N/A   no      nothing  exists   fail
 170      9 no    N/A   no      nothing  exists   fail
 171
 172     10 yes   yes   N/A     exists   nothing  remove path from index
 173     11 no    yes   N/A     exists   nothing  fail
 174     12 yes   no    N/A     exists   nothing  fail
 175     13 no    no    N/A     exists   nothing  fail
 176
 177        clean (H=M)
 178       ------
 179     14 yes                 exists   exists   keep index
 180     15 no                  exists   exists   keep index
 181
 182        clean I==H  I==M (H!=M)
 183       ------------------
 184     16 yes   no    no      exists   exists   fail
 185     17 no    no    no      exists   exists   fail
 186     18 yes   no    yes     exists   exists   keep index
 187     19 no    no    yes     exists   exists   keep index
 188     20 yes   yes   no      exists   exists   use M
 189     21 no    yes   no      exists   exists   fail
 190
 191In all "keep index" cases, the index entry stays as in the
 192original index file.  If the entry were not up to date,
 193git-read-tree keeps the copy in the work tree intact when
 194operating under the -u flag.
 195
 196When this form of git-read-tree returns successfully, you can
 197see what "local changes" you made are carried forward by running
 198`git-diff-index --cached $M`.  Note that this does not
 199necessarily match `git-diff-index --cached $H` would have
 200produced before such a two tree merge.  This is because of cases
 20118 and 19 --- if you already had the changes in $M (e.g. maybe
 202you picked it up via e-mail in a patch form), `git-diff-index
 203--cached $H` would have told you about the change before this
 204merge, but it would not show in `git-diff-index --cached $M`
 205output after two-tree merge.
 206
 207
 2083-Way Merge
 209~~~~~~~~~~~
 210Each "index" entry has two bits worth of "stage" state. stage 0 is the
 211normal one, and is the only one you'd see in any kind of normal use.
 212
 213However, when you do `git-read-tree` with three trees, the "stage"
 214starts out at 1.
 215
 216This means that you can do
 217
 218----------------
 219$ git-read-tree -m <tree1> <tree2> <tree3>
 220----------------
 221
 222and you will end up with an index with all of the <tree1> entries in
 223"stage1", all of the <tree2> entries in "stage2" and all of the
 224<tree3> entries in "stage3".  When performing a merge of another
 225branch into the current branch, we use the common ancestor tree
 226as <tree1>, the current branch head as <tree2>, and the other
 227branch head as <tree3>.
 228
 229Furthermore, `git-read-tree` has special-case logic that says: if you see
 230a file that matches in all respects in the following states, it
 231"collapses" back to "stage0":
 232
 233   - stage 2 and 3 are the same; take one or the other (it makes no
 234     difference - the same work has been done on our branch in
 235     stage 2 and their branch in stage 3)
 236
 237   - stage 1 and stage 2 are the same and stage 3 is different; take
 238     stage 3 (our branch in stage 2 did not do anything since the
 239     ancestor in stage 1 while their branch in stage 3 worked on
 240     it)
 241
 242   - stage 1 and stage 3 are the same and stage 2 is different take
 243     stage 2 (we did something while they did nothing)
 244
 245The `git-write-tree` command refuses to write a nonsensical tree, and it
 246will complain about unmerged entries if it sees a single entry that is not
 247stage 0.
 248
 249OK, this all sounds like a collection of totally nonsensical rules,
 250but it's actually exactly what you want in order to do a fast
 251merge. The different stages represent the "result tree" (stage 0, aka
 252"merged"), the original tree (stage 1, aka "orig"), and the two trees
 253you are trying to merge (stage 2 and 3 respectively).
 254
 255The order of stages 1, 2 and 3 (hence the order of three
 256<tree-ish> command line arguments) are significant when you
 257start a 3-way merge with an index file that is already
 258populated.  Here is an outline of how the algorithm works:
 259
 260- if a file exists in identical format in all three trees, it will
 261  automatically collapse to "merged" state by git-read-tree.
 262
 263- a file that has _any_ difference what-so-ever in the three trees
 264  will stay as separate entries in the index. It's up to "porcelain
 265  policy" to determine how to remove the non-0 stages, and insert a
 266  merged version.
 267
 268- the index file saves and restores with all this information, so you
 269  can merge things incrementally, but as long as it has entries in
 270  stages 1/2/3 (i.e., "unmerged entries") you can't write the result. So
 271  now the merge algorithm ends up being really simple:
 272
 273  * you walk the index in order, and ignore all entries of stage 0,
 274    since they've already been done.
 275
 276  * if you find a "stage1", but no matching "stage2" or "stage3", you
 277    know it's been removed from both trees (it only existed in the
 278    original tree), and you remove that entry.
 279
 280  * if you find a matching "stage2" and "stage3" tree, you remove one
 281    of them, and turn the other into a "stage0" entry. Remove any
 282    matching "stage1" entry if it exists too.  .. all the normal
 283    trivial rules ..
 284
 285You would normally use `git-merge-index` with supplied
 286`git-merge-one-file` to do this last step.  The script updates
 287the files in the working tree as it merges each path and at the
 288end of a successful merge.
 289
 290When you start a 3-way merge with an index file that is already
 291populated, it is assumed that it represents the state of the
 292files in your work tree, and you can even have files with
 293changes unrecorded in the index file.  It is further assumed
 294that this state is "derived" from the stage 2 tree.  The 3-way
 295merge refuses to run if it finds an entry in the original index
 296file that does not match stage 2.
 297
 298This is done to prevent you from losing your work-in-progress
 299changes, and mixing your random changes in an unrelated merge
 300commit.  To illustrate, suppose you start from what has been
 301committed last to your repository:
 302
 303----------------
 304$ JC=`git-rev-parse --verify "HEAD^0"`
 305$ git-checkout-index -f -u -a $JC
 306----------------
 307
 308You do random edits, without running git-update-index.  And then
 309you notice that the tip of your "upstream" tree has advanced
 310since you pulled from him:
 311
 312----------------
 313$ git-fetch git://.... linus
 314$ LT=`cat .git/FETCH_HEAD`
 315----------------
 316
 317Your work tree is still based on your HEAD ($JC), but you have
 318some edits since.  Three-way merge makes sure that you have not
 319added or modified index entries since $JC, and if you haven't,
 320then does the right thing.  So with the following sequence:
 321
 322----------------
 323$ git-read-tree -m -u `git-merge-base $JC $LT` $JC $LT
 324$ git-merge-index git-merge-one-file -a
 325$ echo "Merge with Linus" | \
 326  git-commit-tree `git-write-tree` -p $JC -p $LT
 327----------------
 328
 329what you would commit is a pure merge between $JC and $LT without
 330your work-in-progress changes, and your work tree would be
 331updated to the result of the merge.
 332
 333However, if you have local changes in the working tree that
 334would be overwritten by this merge,`git-read-tree` will refuse
 335to run to prevent your changes from being lost.
 336
 337In other words, there is no need to worry about what exists only
 338in the working tree.  When you have local changes in a part of
 339the project that is not involved in the merge, your changes do
 340not interfere with the merge, and are kept intact.  When they
 341*do* interfere, the merge does not even start (`git-read-tree`
 342complains loudly and fails without modifying anything).  In such
 343a case, you can simply continue doing what you were in the
 344middle of doing, and when your working tree is ready (i.e. you
 345have finished your work-in-progress), attempt the merge again.
 346
 347
 348See Also
 349--------
 350linkgit:git-write-tree[1]; linkgit:git-ls-files[1];
 351linkgit:gitignore[5]
 352
 353
 354Author
 355------
 356Written by Linus Torvalds <torvalds@osdl.org>
 357
 358Documentation
 359--------------
 360Documentation by David Greaves, Junio C Hamano and the git-list <git@vger.kernel.org>.
 361
 362GIT
 363---
 364Part of the linkgit:git[7] suite