Add Git official document to help

[TortoiseGit.git] / doc / source / en / TortoiseGit / git_doc / git-read-tree.html.xml

<?xml version="1.0" encoding="UTF-8"?>\r
<!DOCTYPE article PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">\r
\r
<article lang="en" id="git-read-tree(1)">\r
<articleinfo>\r
    <title>git-read-tree(1)</title>\r
        <indexterm>\r
                <primary>git-read-tree(1)</primary>\r
        </indexterm>\r
</articleinfo>\r
<simplesect id="_name">\r
<title>NAME</title>\r
<simpara>git-read-tree - Reads tree information into the index</simpara>\r
</simplesect>\r
<simplesect id="_synopsis">\r
<title>SYNOPSIS</title>\r
<simpara><emphasis>git read-tree</emphasis> (&lt;tree-ish&gt; | [[-m [--trivial] [--aggressive] | --reset | --prefix=&lt;prefix&gt;] [-u | -i]] [--exclude-per-directory=&lt;gitignore&gt;] [--index-output=&lt;file&gt;] &lt;tree-ish1&gt; [&lt;tree-ish2&gt; [&lt;tree-ish3&gt;]])</simpara>\r
</simplesect>\r
<simplesect id="_description">\r
<title>DESCRIPTION</title>\r
<simpara>Reads the tree information given by &lt;tree-ish&gt; into the index,\r
but does not actually <emphasis role="strong">update</emphasis> any of the files it "caches". (see:\r
<xref linkend="git-checkout-index(1)"/>)</simpara>\r
<simpara>Optionally, it can merge a tree into the index, perform a\r
fast-forward (i.e. 2-way) merge, or a 3-way merge, with the <literal>-m</literal>\r
flag.  When used with <literal>-m</literal>, the <literal>-u</literal> flag causes it to also update\r
the files in the work tree with the result of the merge.</simpara>\r
<simpara>Trivial merges are done by <emphasis>git-read-tree</emphasis> itself.  Only conflicting paths\r
will be in unmerged state when <emphasis>git-read-tree</emphasis> returns.</simpara>\r
</simplesect>\r
<simplesect id="_options">\r
<title>OPTIONS</title>\r
<variablelist>\r
<varlistentry>\r
<term>\r
-m\r
</term>\r
<listitem>\r
<simpara>\r
        Perform a merge, not just a read.  The command will\r
        refuse to run if your index file has unmerged entries,\r
        indicating that you have not finished previous merge you\r
        started.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
--reset\r
</term>\r
<listitem>\r
<simpara>\r
        Same as -m, except that unmerged entries are discarded\r
        instead of failing.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
-u\r
</term>\r
<listitem>\r
<simpara>\r
        After a successful merge, update the files in the work\r
        tree with the result of the merge.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
-i\r
</term>\r
<listitem>\r
<simpara>\r
        Usually a merge requires the index file as well as the\r
        files in the working tree are up to date with the\r
        current head commit, in order not to lose local\r
        changes.  This flag disables the check with the working\r
        tree and is meant to be used when creating a merge of\r
        trees that are not directly related to the current\r
        working tree status into a temporary index file.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
-v\r
</term>\r
<listitem>\r
<simpara>\r
        Show the progress of checking files out.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
--trivial\r
</term>\r
<listitem>\r
<simpara>\r
        Restrict three-way merge by <emphasis>git-read-tree</emphasis> to happen\r
        only if there is no file-level merging required, instead\r
        of resolving merge for trivial cases and leaving\r
        conflicting files unresolved in the index.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
--aggressive\r
</term>\r
<listitem>\r
<simpara>\r
        Usually a three-way merge by <emphasis>git-read-tree</emphasis> resolves\r
        the merge for really trivial cases and leaves other\r
        cases unresolved in the index, so that Porcelains can\r
        implement different merge policies.  This flag makes the\r
        command to resolve a few more cases internally:\r
</simpara>\r
<itemizedlist>\r
<listitem>\r
<simpara>\r
when one side removes a path and the other side leaves the path\r
  unmodified.  The resolution is to remove that path.\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
when both sides remove a path.  The resolution is to remove that path.\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
when both sides adds a path identically.  The resolution\r
  is to add that path.\r
</simpara>\r
</listitem>\r
</itemizedlist>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
--prefix=&lt;prefix&gt;/\r
</term>\r
<listitem>\r
<simpara>\r
        Keep the current index contents, and read the contents\r
        of named tree-ish under directory at <literal>&lt;prefix&gt;</literal>.  The\r
        original index file cannot have anything at the path\r
        <literal>&lt;prefix&gt;</literal> itself, and have nothing in <literal>&lt;prefix&gt;/</literal>\r
        directory.  Note that the <literal>&lt;prefix&gt;/</literal> value must end\r
        with a slash.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
--exclude-per-directory=&lt;gitignore&gt;\r
</term>\r
<listitem>\r
<simpara>\r
        When running the command with <literal>-u</literal> and <literal>-m</literal> options, the\r
        merge result may need to overwrite paths that are not\r
        tracked in the current branch.  The command usually\r
        refuses to proceed with the merge to avoid losing such a\r
        path.  However this safety valve sometimes gets in the\r
        way.  For example, it often happens that the other\r
        branch added a file that used to be a generated file in\r
        your branch, and the safety valve triggers when you try\r
        to switch to that branch after you ran <literal>make</literal> but before\r
        running <literal>make clean</literal> to remove the generated file.  This\r
        option tells the command to read per-directory exclude\r
        file (usually <emphasis>.gitignore</emphasis>) and allows such an untracked\r
        but explicitly ignored file to be overwritten.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
--index-output=&lt;file&gt;\r
</term>\r
<listitem>\r
<simpara>\r
        Instead of writing the results out to <literal>$GIT_INDEX_FILE</literal>,\r
        write the resulting index in the named file.  While the\r
        command is operating, the original index file is locked\r
        with the same mechanism as usual.  The file must allow\r
        to be rename(2)ed into from a temporary file that is\r
        created next to the usual index file; typically this\r
        means it needs to be on the same filesystem as the index\r
        file itself, and you need write permission to the\r
        directories the index file and index output file are\r
        located in.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
<varlistentry>\r
<term>\r
&lt;tree-ish#&gt;\r
</term>\r
<listitem>\r
<simpara>\r
        The id of the tree object(s) to be read/merged.\r
</simpara>\r
</listitem>\r
</varlistentry>\r
</variablelist>\r
</simplesect>\r
<simplesect id="_merging">\r
<title>Merging</title>\r
<simpara>If <literal>-m</literal> is specified, <emphasis>git-read-tree</emphasis> can perform 3 kinds of\r
merge, a single tree merge if only 1 tree is given, a\r
fast-forward merge with 2 trees, or a 3-way merge if 3 trees are\r
provided.</simpara>\r
<simplesect id="_single_tree_merge">\r
<title>Single Tree Merge</title>\r
<simpara>If only 1 tree is specified, <emphasis>git-read-tree</emphasis> operates as if the user did not\r
specify <literal>-m</literal>, except that if the original index has an entry for a\r
given pathname, and the contents of the path matches with the tree\r
being read, the stat info from the index is used. (In other words, the\r
index&#8217;s stat()s take precedence over the merged tree&#8217;s).</simpara>\r
<simpara>That means that if you do a <literal>git read-tree -m &lt;newtree&gt;</literal> followed by a\r
<literal>git checkout-index -f -u -a</literal>, the <emphasis>git-checkout-index</emphasis> only checks out\r
the stuff that really changed.</simpara>\r
<simpara>This is used to avoid unnecessary false hits when <emphasis>git-diff-files</emphasis> is\r
run after <emphasis>git-read-tree</emphasis>.</simpara>\r
</simplesect>\r
<simplesect id="_two_tree_merge">\r
<title>Two Tree Merge</title>\r
<simpara>Typically, this is invoked as <literal>git read-tree -m $H $M</literal>, where $H\r
is the head commit of the current repository, and $M is the head\r
of a foreign tree, which is simply ahead of $H (i.e. we are in a\r
fast forward situation).</simpara>\r
<simpara>When two trees are specified, the user is telling <emphasis>git-read-tree</emphasis>\r
the following:</simpara>\r
<orderedlist numeration="arabic">\r
<listitem>\r
<simpara>\r
The current index and work tree is derived from $H, but\r
        the user may have local changes in them since $H;\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
The user wants to fast-forward to $M.\r
</simpara>\r
</listitem>\r
</orderedlist>\r
<simpara>In this case, the <literal>git read-tree -m $H $M</literal> command makes sure\r
that no local change is lost as the result of this "merge".\r
Here are the "carry forward" rules:</simpara>\r
<literallayout class="monospaced">  I (index)           H        M        Result\r
 -------------------------------------------------------\r
0 nothing             nothing  nothing  (does not happen)\r
1 nothing             nothing  exists   use M\r
2 nothing             exists   nothing  remove path from index\r
3 nothing             exists   exists,  use M if "initial checkout"\r
                               H == M   keep index otherwise\r
                               exists   fail\r
                               H != M</literallayout>\r
<literallayout class="monospaced">  clean I==H  I==M\r
 ------------------\r
4 yes   N/A   N/A     nothing  nothing  keep index\r
5 no    N/A   N/A     nothing  nothing  keep index</literallayout>\r
<literallayout class="monospaced">6 yes   N/A   yes     nothing  exists   keep index\r
7 no    N/A   yes     nothing  exists   keep index\r
8 yes   N/A   no      nothing  exists   fail\r
9 no    N/A   no      nothing  exists   fail</literallayout>\r
<literallayout class="monospaced">10 yes   yes   N/A     exists   nothing  remove path from index\r
11 no    yes   N/A     exists   nothing  fail\r
12 yes   no    N/A     exists   nothing  fail\r
13 no    no    N/A     exists   nothing  fail</literallayout>\r
<literallayout class="monospaced">   clean (H=M)\r
  ------\r
14 yes                 exists   exists   keep index\r
15 no                  exists   exists   keep index</literallayout>\r
<literallayout class="monospaced">   clean I==H  I==M (H!=M)\r
  ------------------\r
16 yes   no    no      exists   exists   fail\r
17 no    no    no      exists   exists   fail\r
18 yes   no    yes     exists   exists   keep index\r
19 no    no    yes     exists   exists   keep index\r
20 yes   yes   no      exists   exists   use M\r
21 no    yes   no      exists   exists   fail</literallayout>\r
<simpara>In all "keep index" cases, the index entry stays as in the\r
original index file.  If the entry were not up to date,\r
<emphasis>git-read-tree</emphasis> keeps the copy in the work tree intact when\r
operating under the -u flag.</simpara>\r
<simpara>When this form of <emphasis>git-read-tree</emphasis> returns successfully, you can\r
see what "local changes" you made are carried forward by running\r
<literal>git diff-index --cached $M</literal>.  Note that this does not\r
necessarily match <literal>git diff-index --cached $H</literal> would have\r
produced before such a two tree merge.  This is because of cases\r
18 and 19 --- if you already had the changes in $M (e.g. maybe\r
you picked it up via e-mail in a patch form), <literal>git diff-index\r
--cached $H</literal> would have told you about the change before this\r
merge, but it would not show in <literal>git diff-index --cached $M</literal>\r
output after two-tree merge.</simpara>\r
<simpara>Case #3 is slightly tricky and needs explanation.  The result from this\r
rule logically should be to remove the path if the user staged the removal\r
of the path and then switching to a new branch.  That however will prevent\r
the initial checkout from happening, so the rule is modified to use M (new\r
tree) only when the contents of the index is empty.  Otherwise the removal\r
of the path is kept as long as $H and $M are the same.</simpara>\r
</simplesect>\r
<simplesect id="_3_way_merge">\r
<title>3-Way Merge</title>\r
<simpara>Each "index" entry has two bits worth of "stage" state. stage 0 is the\r
normal one, and is the only one you&#8217;d see in any kind of normal use.</simpara>\r
<simpara>However, when you do <emphasis>git-read-tree</emphasis> with three trees, the "stage"\r
starts out at 1.</simpara>\r
<simpara>This means that you can do</simpara>\r
<literallayout>$ git read-tree -m &lt;tree1&gt; &lt;tree2&gt; &lt;tree3&gt;</literallayout>\r
<simpara>and you will end up with an index with all of the &lt;tree1&gt; entries in\r
"stage1", all of the &lt;tree2&gt; entries in "stage2" and all of the\r
&lt;tree3&gt; entries in "stage3".  When performing a merge of another\r
branch into the current branch, we use the common ancestor tree\r
as &lt;tree1&gt;, the current branch head as &lt;tree2&gt;, and the other\r
branch head as &lt;tree3&gt;.</simpara>\r
<simpara>Furthermore, <emphasis>git-read-tree</emphasis> has special-case logic that says: if you see\r
a file that matches in all respects in the following states, it\r
"collapses" back to "stage0":</simpara>\r
<itemizedlist>\r
<listitem>\r
<simpara>\r
stage 2 and 3 are the same; take one or the other (it makes no\r
     difference - the same work has been done on our branch in\r
     stage 2 and their branch in stage 3)\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
stage 1 and stage 2 are the same and stage 3 is different; take\r
     stage 3 (our branch in stage 2 did not do anything since the\r
     ancestor in stage 1 while their branch in stage 3 worked on\r
     it)\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
stage 1 and stage 3 are the same and stage 2 is different take\r
     stage 2 (we did something while they did nothing)\r
</simpara>\r
</listitem>\r
</itemizedlist>\r
<simpara>The <emphasis>git-write-tree</emphasis> command refuses to write a nonsensical tree, and it\r
will complain about unmerged entries if it sees a single entry that is not\r
stage 0.</simpara>\r
<simpara>OK, this all sounds like a collection of totally nonsensical rules,\r
but it&#8217;s actually exactly what you want in order to do a fast\r
merge. The different stages represent the "result tree" (stage 0, aka\r
"merged"), the original tree (stage 1, aka "orig"), and the two trees\r
you are trying to merge (stage 2 and 3 respectively).</simpara>\r
<simpara>The order of stages 1, 2 and 3 (hence the order of three\r
&lt;tree-ish&gt; command line arguments) are significant when you\r
start a 3-way merge with an index file that is already\r
populated.  Here is an outline of how the algorithm works:</simpara>\r
<itemizedlist>\r
<listitem>\r
<simpara>\r
if a file exists in identical format in all three trees, it will\r
  automatically collapse to "merged" state by <emphasis>git-read-tree</emphasis>.\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
a file that has <emphasis>any</emphasis> difference what-so-ever in the three trees\r
  will stay as separate entries in the index. It&#8217;s up to "porcelain\r
  policy" to determine how to remove the non-0 stages, and insert a\r
  merged version.\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
the index file saves and restores with all this information, so you\r
  can merge things incrementally, but as long as it has entries in\r
  stages 1/2/3 (i.e., "unmerged entries") you can&#8217;t write the result. So\r
  now the merge algorithm ends up being really simple:\r
</simpara>\r
<itemizedlist>\r
<listitem>\r
<simpara>\r
you walk the index in order, and ignore all entries of stage 0,\r
    since they&#8217;ve already been done.\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
if you find a "stage1", but no matching "stage2" or "stage3", you\r
    know it&#8217;s been removed from both trees (it only existed in the\r
    original tree), and you remove that entry.\r
</simpara>\r
</listitem>\r
<listitem>\r
<simpara>\r
if you find a matching "stage2" and "stage3" tree, you remove one\r
    of them, and turn the other into a "stage0" entry. Remove any\r
    matching "stage1" entry if it exists too.  .. all the normal\r
    trivial rules ..\r
</simpara>\r
</listitem>\r
</itemizedlist>\r
</listitem>\r
</itemizedlist>\r
<simpara>You would normally use <emphasis>git-merge-index</emphasis> with supplied\r
<emphasis>git-merge-one-file</emphasis> to do this last step.  The script updates\r
the files in the working tree as it merges each path and at the\r
end of a successful merge.</simpara>\r
<simpara>When you start a 3-way merge with an index file that is already\r
populated, it is assumed that it represents the state of the\r
files in your work tree, and you can even have files with\r
changes unrecorded in the index file.  It is further assumed\r
that this state is "derived" from the stage 2 tree.  The 3-way\r
merge refuses to run if it finds an entry in the original index\r
file that does not match stage 2.</simpara>\r
<simpara>This is done to prevent you from losing your work-in-progress\r
changes, and mixing your random changes in an unrelated merge\r
commit.  To illustrate, suppose you start from what has been\r
committed last to your repository:</simpara>\r
<literallayout>$ JC=`git rev-parse --verify "HEAD^0"`\r
$ git checkout-index -f -u -a $JC</literallayout>\r
<simpara>You do random edits, without running <emphasis>git-update-index</emphasis>.  And then\r
you notice that the tip of your "upstream" tree has advanced\r
since you pulled from him:</simpara>\r
<literallayout>$ git fetch git://.... linus\r
$ LT=`cat .git/FETCH_HEAD`</literallayout>\r
<simpara>Your work tree is still based on your HEAD ($JC), but you have\r
some edits since.  Three-way merge makes sure that you have not\r
added or modified index entries since $JC, and if you haven&#8217;t,\r
then does the right thing.  So with the following sequence:</simpara>\r
<literallayout>$ git read-tree -m -u `git merge-base $JC $LT` $JC $LT\r
$ git merge-index git-merge-one-file -a\r
$ echo "Merge with Linus" | \\r
  git commit-tree `git write-tree` -p $JC -p $LT</literallayout>\r
<simpara>what you would commit is a pure merge between $JC and $LT without\r
your work-in-progress changes, and your work tree would be\r
updated to the result of the merge.</simpara>\r
<simpara>However, if you have local changes in the working tree that\r
would be overwritten by this merge, <emphasis>git-read-tree</emphasis> will refuse\r
to run to prevent your changes from being lost.</simpara>\r
<simpara>In other words, there is no need to worry about what exists only\r
in the working tree.  When you have local changes in a part of\r
the project that is not involved in the merge, your changes do\r
not interfere with the merge, and are kept intact.  When they\r
<emphasis role="strong">do</emphasis> interfere, the merge does not even start (<emphasis>git-read-tree</emphasis>\r
complains loudly and fails without modifying anything).  In such\r
a case, you can simply continue doing what you were in the\r
middle of doing, and when your working tree is ready (i.e. you\r
have finished your work-in-progress), attempt the merge again.</simpara>\r
</simplesect>\r
</simplesect>\r
<simplesect id="_see_also">\r
<title>SEE ALSO</title>\r
<simpara><xref linkend="git-write-tree(1)"/>; <xref linkend="git-ls-files(1)"/>;\r
<xref linkend="gitignore(5)"/></simpara>\r
</simplesect>\r
<simplesect id="_author">\r
<title>Author</title>\r
<simpara>Written by Linus Torvalds &lt;<ulink url="mailto:torvalds@osdl.org">torvalds@osdl.org</ulink>&gt;</simpara>\r
</simplesect>\r
<simplesect id="_documentation">\r
<title>Documentation</title>\r
<simpara>Documentation by David Greaves, Junio C Hamano and the git-list &lt;<ulink url="mailto:git@vger.kernel.org">git@vger.kernel.org</ulink>&gt;.</simpara>\r
</simplesect>\r
<simplesect id="_git">\r
<title>GIT</title>\r
<simpara>Part of the <xref linkend="git(1)"/> suite</simpara>\r
</simplesect>\r
</article>\r